JP2011015018A - Automatic sound volume controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve a smooth conversation as usual without requiring a driver or an occupant to operate a sound volume adjusting switch or a mute switch or to speak in a loud voice.SOLUTION: The automatic sound volume controller 1 includes: voice signal extraction means 2, 3 for extracting acoustic signals by applying band limitation of voice band components and adaptive algorithm to voice signals collected by microphones M1 and M2; a sound volume correction means 4 for reducing the variation feeling of a sound volume in music signals; a voice analysis means 6 for obtaining a voice analysis gain obtained by weighting a voice detection part in the acoustic signals; a voice detection means 7 for enhancing the voice detection part and obtaining voice detection signals by applying the voice analysis gain to the acoustic signals; and a sound volume control means 8 for reducing the output level of the music signals when detecting the voice signals on the basis of the voice detection signals.

Description

  The present invention relates to an automatic volume control apparatus, and more particularly, to realize smooth conversation by reducing the output level of music according to the utterance when conversation is performed in a space where music is played. The present invention relates to an automatic volume control device capable of performing the above.

  In the interior of a running vehicle, music and radio programs are often played during driving (running). In such a situation, when the driver and the passenger have a conversation, there is a possibility that a smooth conversation (listening of the conversation, etc.) may be hindered by the reproduced sound such as music.

  A general in-vehicle audio apparatus is provided with a volume control switch for adjusting the volume, a mute switch for temporarily reducing the volume, and the like (see, for example, Patent Document 1 and Patent Document 2). . For this reason, the driver or the like often reduces the playback volume of music or the like to such an extent that the conversation is not hindered by operating the volume control switch and the mute switch.

JP 2008-62906 A JP 2006-67490 A

  However, the method of operating the volume control switch every time a conversation is performed to reduce the reproduced sound has a problem that the operation becomes complicated and may hinder smooth conversation. On the other hand, with the method of reducing the music playback sound using the mute switch, the music playback sound is reduced even when the conversation is interrupted, and music or radio programs cannot be enjoyed. There was a problem.

  For this reason, a conversation is often performed in a state in which music or the like is played by a speaker speaking loudly without operating a volume control switch or a mute switch. If this continues, there is a problem that the conversation may be fatigued not only by the speaker but also by the conversation partner.

  The present invention has been made in view of the above problems, and a driver or a passenger does a smooth conversation in a normal state without operating a volume control switch or a mute switch and without generating a loud voice. It is an object of the present invention to provide an automatic sound volume control device that can be used.

  In order to solve the above problems, an automatic volume control device according to the present invention applies a band limiting process of a voice band component to a voice signal collected by a microphone and applies an adaptive algorithm to a voice band. Audio signal extraction means for extracting the audio signal as an acoustic signal and a music signal from a sound source are divided for each band, and when the signal level of the music signal is equal to or higher than a certain level, the signal level in each divided band is constant Volume correction means for correcting the volume of the music signal by maintaining the value, and voice analysis means for obtaining a voice analysis gain in which a voice detection portion is weighted from the acoustic signal extracted by the voice signal extraction means; Applying the voice analysis gain obtained by the voice analysis means to the acoustic signal extracted by the voice signal extraction means. Thus, the sound detection part that makes the sound detection part in the acoustic signal obvious and obtains the sound detection signal indicating the presence or absence of the sound detection, and the sound volume correction means based on the sound detection signal obtained by the sound detection means And a volume control means for reducing the output level of the music signal subjected to the volume correction when the voice is detected.

  According to the automatic volume control device of the present invention, the output level of the music signal whose volume is corrected by the volume correction means is reduced at the time of voice detection based on the voice detection signal obtained by the voice detection means. The signal level of the music signal can be automatically reduced according to the conversation (speech) of the talker. For this reason, it is possible to perform a smooth conversation without operating a volume control switch or a mute switch each time a conversation is performed.

  In particular, the volume correction means divides the music signal from the sound source for each band, and maintains the signal level in each divided band at a constant value when the signal level of the music signal is equal to or higher than a certain level. It is possible to reduce the fluctuation of the sound volume at. For the music signal with reduced volume fluctuation, the output level is reduced when the audio signal is detected, so the music does not depend on the source or genre of the sound source, and the music is accurate and comfortable. It is possible to reduce the sound volume.

  Furthermore, when obtaining a voice detection signal indicating the presence or absence of voice detection in the voice detection means, the voice detection portion is further weighted with respect to the voice signal related to the voice band extracted by the voice signal extraction means. By applying the voice analysis gain, the voice detection portion in the acoustic signal is revealed and the presence or absence of voice detection is detected, so that the detection accuracy of the voice detection signal can be increased.

  In the automatic sound volume control apparatus described above, the sound detection means further detects the sound at a predetermined time by integrating the sound detection value at a predetermined time in the sound detection portion based on the sound detection signal. A tempo detection means for calculating a tempo detection value for determining a state change and determining a utterance speed of the speaker based on the obtained integral value, and a tempo detection value obtained by the tempo detection means based on the tempo detection value obtained by the tempo detection means. Adaptive attack release filter means for determining an attack time corresponding to the detection time and a release time corresponding to the holding time of the voice detection, and setting the determined attack time and release time for the voice detection signal It may be a thing.

  Thus, the tempo detection means of the voice detection means calculates the tempo detection value based on the change in the detection state of the voice during a predetermined time, so that the utterance speed of the speaker can be determined based on the tempo detection value. It becomes. Therefore, the adaptive attack release filter means determines the attack time and release time based on the tempo detection value, so that the voice detection time (fade-in time) and holding time (fade-out time) according to the utterance speed (tempo) of the speaker. ) Can be varied, and the sound volume can be controlled without any sense of incongruity.

  Further, in the automatic volume control device described above, the tempo detection unit is configured to determine the speaking speed of the speaker by clearing the integration value obtained by the integration processing based on the input of the reset signal. The tempo detection value may be recalculated.

  Thus, by recalculating the tempo detection value for determining the utterance speed of the speaker based on the input of the reset signal, the voice detection time (fade-in time) and the holding time (based on the tempo detection value) are determined. Since the content of the determination of (fade-out time) is calculated (learned) again, it can be controlled so that the sound detection and holding time is appropriately changed, and it is possible to reduce the uncomfortable feeling of unexpected volume changes. It becomes.

  Further, in the above-described automatic volume control device, the adaptive attack release filter means includes a variable mode for setting the attack time and the release time based on the tempo detection value, and the attack time and the release time for the tempo detection. It may have a fixed mode that is set to a predetermined value regardless of the value.

  As described above, in the automatic volume control device according to the present invention, the voice detection time (fade-in time) and the holding time (fade-out time) determined by the adaptive attack release filter unit are changed according to the speaking speed of the speaker. Alternatively, whether the predetermined value is set regardless of the speaking speed of the speaker can be changed according to the preference of the user. For this reason, it is possible to set the optimum voice detection time (fade-in time) and holding time (fade-out time) according to the use situation and preference of the user. In particular, in the case of the fixed mode, the voice detection time (fade-in time) and the holding time (fade-out time) can be appropriately set according to the user's preference, so that the user's needs can be dealt with in detail. It becomes possible.

  Further, in the above-described automatic volume control device, the audio signal extraction unit performs an NLMS adaptive algorithm after performing a first band limiting process corresponding to an audio band component on the audio signal collected by the microphone. Applying a second band limiting process corresponding to a voice band component to the array microphone means for extracting a voice signal relating to the voice band by applying the voice signal relating to the voice band extracted by the array microphone means Audio canceller means for applying a multi-stage LMS adaptive algorithm to the audio signal subjected to the second band limitation processing using an adaptive filter cascade-connected in correspondence with the number of channels of the music signal May be provided.

  In this way, by applying the NLMS adaptation algorithm in the array microphone means, and further applying the LMS adaptation algorithm in multiple stages using the adaptive filters cascaded in correspondence with the number of music signal channels in the audio canceller means. Thus, signal components (noise components) other than the audio signal components in the audio signal can be reduced while ensuring effective and high convergence, and the detection accuracy of the audio signal in the audio band can be improved. .

  In the automatic sound volume control apparatus described above, the bandwidth limit width of the second bandwidth limit process in the audio canceller means includes an upper limit value and a lower limit value of the bandwidth limit width of the first bandwidth limit process in the array microphone means. The bandwidth may be set to be slightly wider than the bandwidth limit of the first bandwidth limitation process.

  As described above, the bandwidth limit width of the second bandwidth limit process includes the upper limit value and the lower limit value of the bandwidth limit width of the first bandwidth limit process, and is slightly wider than the bandwidth limit width of the first bandwidth limit process. By setting the bandwidth, the audio cancellation performance near the cutoff frequency of the band limitation can be improved.

  In the automatic volume control device described above, the volume control means may change an output level reduction amount in the music signal in accordance with a volume state of the sound source.

  Thus, by changing the output level reduction amount in the music signal in accordance with the volume state of the sound source, the volume control can be performed in accordance with the volume state of the sound source. (Volume) can be changed.

  According to the automatic volume control device of the present invention, the output level of the music signal is reduced when the voice is detected, so that the signal level of the music signal is automatically reduced according to the conversation (utterance) of the talker. Can do. For this reason, it is possible to perform a smooth conversation without operating a volume control switch or a mute switch each time a conversation is performed.

It is the block diagram which showed schematic structure of the automatic volume control apparatus which concerns on this Embodiment. It is the block diagram which showed schematic structure of the array microphone part which concerns on this Embodiment. It is the figure which showed the state in which the microphone and the speaker were installed in the vehicle interior. It is the figure which showed visually the emphasized directivity calculated | required by the directivity of microphone M1 and microphone M2 which concern on this Embodiment, and the directivity difference of each microphone. It is the block diagram which showed schematic structure of the audio canceller part which concerns on this Embodiment. The acoustic signal (omnidirectional microphone) collected by the microphone according to the present embodiment, the acoustic signal (array microphone unit) after the adaptive filter unit of the array microphone unit is applied, and the first in the audio canceller unit Acoustic signal after applying the adaptive filter unit (array microphone unit + audio canceller unit (L)) and acoustic signal after applying the second adaptive filter unit (array microphone unit + audio canceller unit (L + R)) It is the figure which showed the frequency characteristic. (A) shows the filter coefficient applied in the adaptive filter part of the array microphone part, (b) shows the filter coefficient applied in the first adaptive filter part of the audio canceller part, and (c) shows the audio coefficient. It is the figure which showed the filter coefficient applied in the 2nd adaptive filter part of a canceller part. It is the block diagram which showed schematic structure of the sound volume correction | amendment part which concerns on this Embodiment. It is the block diagram which showed the function part used for the process which frequency-divides L channel music signal L1 into three bands, a low region, a mid range, and a high region, in the 3 band band pass filter part which concerns on this Embodiment. . It is the figure which showed the filter characteristic of the 1st low pass filter part of the 3 band band pass filter part which concerns on this Embodiment, and a 2nd low pass filter part. It is the figure which showed the state which performed the maximum value detection for every 2 msec with respect to the input signal, and also hold | maintained the maximum value only for 16 msec in the maximum value detection and maximum value holding part which concerns on this Embodiment. It is the block diagram which showed schematic structure of the gain calculation part which concerns on this Embodiment. It is the figure which showed an example of the level conversion operation | movement of the 1st lookup table part-3rd lookup table part which concerns on this Embodiment. It is the block diagram which showed schematic structure of the gain setting part which concerns on this Embodiment. From the maximum value hold signal (control signal) output from the maximum value detection and maximum value hold unit on the basis of the low frequency signal of the volume correction unit according to the present embodiment, and from the first attack release filter unit of the gain calculation unit It is the figure which showed the output signal output and the low-pass control signal output from a gain calculation part. From the maximum value hold signal (control signal) output from the maximum value detection and maximum value hold unit and the second attack release filter unit of the gain calculation unit with reference to the mid-range signal of the volume correction unit according to the present embodiment It is the figure which showed the output signal output and the mid range control signal output from a gain calculation part. It is the figure which showed the content of FIG. 15 in case the signal level of a sound source is low. (A) shows the signal state when the volume correction is not performed in the volume correction unit according to the present embodiment, and (b) is the volume correction performed in the volume correction unit according to the present embodiment. It is the figure which showed the signal state in the case. It is the block diagram which showed schematic structure of the audio | voice analysis part which concerns on this Embodiment. (A) shows the music signal output from the sound source according to the present embodiment, (b) shows the audio signal collected from the microphone according to the present embodiment, and (c) shows the present embodiment. It is the figure which showed the music signal after volume control was performed in the automatic volume control apparatus which concerns on the form. (A) is a signal input from the audio canceller to the audio analyzer when the music signal shown in FIG. 20 (a) and the audio signal as shown in FIG. 20 (b) are input to the microphone. (B) shows the voice analysis gain output from the voice analysis unit in the same case. It is the block diagram which showed schematic structure of the audio | voice detection part which concerns on this Embodiment. When a music signal as shown in FIG. 20A and an audio signal as shown in FIG. 20B are input to the microphone, the input signal (audio signal × audio analysis gain) of the audio detection threshold unit. And a voice detection threshold set by a voice detection threshold unit. It is the block diagram which showed schematic structure of the tempo detection part which concerns on this Embodiment. (A) shows an audio detection signal binarized by detecting the audio detection threshold in the audio detection threshold unit according to the present embodiment, and (b) is based on the binarized audio detection signal, The signal output from the tempo gain unit is shown, and (c) shows the integrated value (integrated output) obtained by the low-pass filter unit according to the present embodiment and the tempo detection value output by the gain offset unit. Yes. (A) shows the audio | voice detection signal input from the audio | voice detection threshold part which concerns on this Embodiment in fixed mode, (b) is the audio | voice holding | maintenance filter in an adaptive attack release filter part, and the audio | voice holding | maintenance in an audio | voice holding threshold part It is the figure which showed the threshold. (A) shows the voice interval detection value obtained based on the signal states shown in FIGS. 26 (a) and (b), and (b) shows the attack release filter section for the voice detection signal shown in (a). (C) shows the volume control value in the level limiter unit. It is the block diagram which showed schematic structure of the volume control part which concerns on this Embodiment. In the level calculation part which concerns on this Embodiment, it is the figure which showed multiple types of volume control levels which change according to increase / decrease in volume information (volume adjustment amount). It is the figure which showed the relationship between attack time (voice detection time) and release time (holding time) determined according to a tempo detection value in the tempo detection part which concerns on this Embodiment. (A) shows the detection state of the voice detection signal when the voice speed is short and the voice interval is very long in the voice signal detection state, and (b) shows the voice holding of the adaptive attack release filter unit in the fixed mode. (C) is the figure which showed the example of adaptive operation of the audio | voice holding filter of the adaptive attack release filter part in a variable mode. (A) shows the detection state of the sound detection value when the sound speed is slightly short and the sound interval is slightly long in the sound signal detection state, and (b) is the sound holding of the adaptive attack release filter unit in the fixed mode. (C) is the figure which showed the example of adaptive operation of the audio | voice holding filter of the adaptive attack release filter part in a variable mode.

  Hereinafter, an automatic sound volume control apparatus according to the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of an automatic volume control apparatus 1 according to the present embodiment. In the present embodiment, the case where the automatic volume control device 1 is installed in a vehicle will be described as an example. By installing the automatic volume control device according to the present invention in a vehicle, it is possible to automatically reduce the volume of music output from the in-vehicle audio device according to the presence or absence of a conversation.

  As shown in FIG. 1, an automatic volume control apparatus 1 according to the present embodiment includes an array microphone unit (array microphone unit, audio signal extraction unit) 2 and an audio canceller unit (audio canceller unit, audio signal extraction unit) 3. A volume correction unit (volume correction unit) 4, a main volume unit 5, a voice analysis unit (speech analysis unit) 6, a voice detection unit (speech detection unit) 7, and a volume control unit (volume control unit) 8. And a power amplifier unit 9, microphones M1 and M2, and speakers 10a and 10b.

[Array microphone section]
FIG. 2 is a block diagram showing a schematic configuration of the array microphone unit 2. As shown in FIG. 2, the array microphone unit 2 includes a first band pass filter unit 21, a second band pass filter unit 22, a delay unit 23, and an adaptive filter unit 24.

  The first band-pass filter unit 21 and the second band-pass filter unit 22 have a role of performing band limitation of about 400 Hz to 2.4 kHz on an acoustic signal input via the microphone M1 and the microphone M2. . Therefore, the acoustic signals passing through the first bandpass filter unit 21 and the second bandpass filter unit 22 are only signals corresponding to the audio band among the acoustic signals input via the microphones M1 and M2.

  The delay unit 23 has a role of delaying the acoustic signal on the microphone M1 side so as to correspond to the signal subtraction processing in the adaptive filter unit 24. For this reason, the delay unit 23 is applied only to the acoustic signal of the microphone M <b> 1 whose band is limited by the first bandpass filter unit 21. The acoustic signal subjected to the delay process by the delay unit 23 is input to the adaptive filter unit 24.

  The adaptive filter unit 24 performs subtraction processing of the acoustic signal input from the microphone M2 from the acoustic signal input from the microphone M1 and subjected to delay processing by the delay unit 23.

  The adaptive filter unit 24 includes an FIR (Finite Impulse Response Filter) unit 25, an NLMS (Normalized Least Mean Square) unit 26, and an addition unit 27.

  The FIR unit 25 includes a finite impulse response filter, and has a function of performing filter processing on the acoustic signal collected by the microphone M <b> 2 based on the coefficient control performed by the NLMS unit 26. The adder 27 applies the acoustic signal from the microphone M2 that has been filtered by the FIR unit 25 to the acoustic signal from the microphone M1 that has been subjected to the delay processing by the delay unit 23 while inverting the phase. Addition (substantially subtracting the acoustic signal of the microphone M2 subjected to the filtering process from the acoustic signal of the microphone M1). The acoustic signal added by the adding unit 27 is output from the adaptive filter unit 24 and also output to the NLMS unit 26.

  The NLMS unit 26 converts the acoustic signal acquired from the adding unit 27 (a signal obtained by subtracting the acoustic signal of the microphone M2 subjected to the filtering process from the acoustic signal of the microphone M1) and the acoustic signal collected by the microphone M2 into the acoustic signal. Based on the least square algorithm, the filter coefficient control of the FIR unit 25 is performed. By providing the NLMS unit 26 in the adaptive filter unit 24 in this way, it is possible to apply an NLMS algorithm having a feature that the adaptive speed does not depend on the magnitude of the input signal.

  As shown in FIG. 3, the microphone M1 and the microphone M2 are installed in a sun visor provided above the driver seat 28a and the passenger seat 28b of the vehicle 28. The microphone M1 and the microphone M2 are used for acquiring a conversation in the vehicle interior. As shown in FIG. 4, a non-directional microphone is used for the microphone M1, and a single microphone M2 is used. A directional microphone is used. In this way, the sound collected by the omnidirectional microphone M1 and the sound collected by the directional microphone M2 are respectively input to the array microphone unit 2.

  When the sound collected by the omnidirectional microphone M1 and the sound collected by the directional microphone M2 are respectively input to the array microphone unit 2, the adaptive filter unit 24 of the array microphone unit 2 is used. , The acoustic signal of the microphone M2 is subtracted from the acoustic signal of the microphone M1, so that the subtraction result remains the null direction of the microphone M2 (the direction other than the directivity range in the microphone M2), and as a result, the directivity in the corresponding direction is It will be emphasized.

  Accordingly, by installing the microphone M1 and the microphone M2 so that the speaker is positioned in the direction in which the directivity is emphasized, it is possible to effectively acquire the voice of the speaker. By effectively acquiring the voice of the speaker in this way, the acquired voice is emphasized, so the acoustic signal required in the array microphone unit 2 is the voice of the speaker (desired signal D). The relative ratio to the music (undesired signal U) output from the in-vehicle audio signal, that is, the D / U is improved.

  Note that the configurations of the array microphone unit 2, the microphone M1, and the microphone M2 are not limited to the configurations described in the present embodiment, and the directivity of the speaker's voice is emphasized to improve D / U. A different configuration may be used as long as a possible method is realized.

  Further, as shown in FIG. 3, the vehicle 28 is provided with four speakers, specifically, a right front door, a left front door, a right rear door, and a left rear door, respectively. From the speaker provided in the right rear door (this speaker corresponds to the speaker 10a), a music signal (right music signal R5) in which the acoustic effect of the right component is emphasized in the power amplifier unit 9 is output, and the left front door and the left From the speaker provided in the rear door (this speaker corresponds to the speaker 10b), a music signal (left music signal L5) in which the acoustic effect of the left component is emphasized in the power amplifier unit 9 is output.

[Audio canceller]
Next, the audio canceller unit 3 will be described. FIG. 5 is a block diagram showing a schematic configuration of the audio canceller unit 3. As shown in FIG. 5, the audio canceller unit 3 includes a first band pass filter unit 31, a second band pass filter unit 32, a first delay unit 33, a second delay unit 34, and a first adaptive filter unit. 35 and a second adaptive filter unit 36.

  The first band-pass filter unit 31 and the second band-pass filter unit 32 are about 200 Hz to 2.6 kHz in the two-channel music signal that has passed through the volume correction unit 4, that is, the left music signal L2 and the right music signal R2. By performing the band limitation, it has a role of allowing only the sound band signal to pass mainly among the acoustic signals.

  Note that the audio canceller unit 3 uses the band limit width (about 200 Hz to 2.6 kHz) set in the first bandpass filter unit 31 and the second bandpass filter unit 32 as the first bandpass filter of the array microphone unit 2. By setting the bandwidth (including 400 Hz to 2.4 kHz) wider than the bandwidth limit (about 400 Hz to 2.4 kHz) set by the unit 21 and the second band pass filter unit 22, the array microphone unit The audio cancellation performance is improved in the vicinity of the band limit cutoff of 2, that is, at 400 Hz and 2.4 kHz.

  The first delay unit 33 and the second delay unit 34 have a role of performing a delay process on the acoustic signal subjected to the band limiting process by the first band pass filter unit 31 and the second band pass filter unit 32. Yes. By the delay processing by the first delay unit 33 and the second delay unit 34, it is possible to correct the propagation delay of the acoustic signal input through the array microphone unit 2.

  The first adaptive filter unit 35 is roughly configured by a first FIR unit 37, a first LMS unit 39, and a first adder unit 41. The second adaptive filter unit 36 includes a second FIR unit 38, a second LMS unit 40, a second LMS unit 40, and a second LMS unit 40. The adder 42 is roughly configured. The first adaptive filter unit 35 and the second adaptive filter unit 36 correspond to a configuration in which the NLMS unit 26 in the adaptive filter unit 24 of the array microphone unit 2 is replaced with a first LMS unit 39 and a second LMS unit 40.

  In the first adaptive filter unit 35 and the second adaptive filter unit 36, the first LMS unit 39 and the second LMS unit 40 use a general LMS (Least Mean Square) algorithm to input sound from the array microphone unit 2. A process of sequentially subtracting the music signal L2 and the music signal R2 whose volume has been corrected by the volume correction unit 4 from the signal is performed. The specific configurations of the first adaptive filter unit 35 and the second adaptive filter unit 36 are the same as those of the adaptive filter unit 24 of the array microphone unit 2 as shown in FIG. Description is omitted.

  In the audio canceller unit 3, the first adaptive filter unit 35 and the second adaptive filter unit 36 are cascade-connected as shown in FIG. Therefore, in the audio canceller unit 3, the first adaptive filter unit 35 subtracts the acoustic signal input from the array microphone unit 2 using the music signal L 2, and then the second adaptive filter unit 36 subtracts the first adaptive filter unit 35. The processed acoustic signal is subtracted by the music signal R2. In this case, since the first adaptive filter unit 35 needs to converge faster than the second adaptive filter unit 36, the adaptive speed is set to be large. When there are two or more sound sources, the same effect can be obtained by increasing the number of adaptive filter units installed according to the number of channels.

  FIG. 6 shows the frequency characteristics of the acoustic signal collected by the microphone M1 in the vehicle 28 as shown in FIG. 3, the adaptive filter unit 24 when the array microphone unit 2 and the audio canceller unit 3 are operated, FIG. 5 is a diagram illustrating frequency characteristics of output signals of an adaptive filter unit and a second adaptive filter unit. Specifically, FIG. 6 shows an acoustic signal (a graph indicated by “omnidirectional microphone” in FIG. 6) of the microphone M1 before applying the adaptive filter unit 24 of the array microphone unit 2, and an adaptive filter of the array microphone unit 2. 6 and the acoustic signal after applying the first adaptive filter unit 35 in the audio canceller unit 3 (“array microphone” in FIG. 6). Part + audio canceller part (L) ”and an acoustic signal after applying the second adaptive filter part 36 in the audio canceller part 3 (“ array microphone part + audio canceller part (L + R) ”in FIG. 6) Graph).

  In the case shown in FIG. 6, a long-period M-sequence signal is used for the music signal L1 and the music signal R1 output from the in-vehicle audio apparatus, and the music signal L1 and the music signal R1 are uncorrelated signals. It has become. 7A shows filter coefficients applied by the adaptive filter unit 24 of the array microphone unit 2, and FIG. 7B shows a filter applied by the first adaptive filter unit 35 of the audio canceller unit 3. FIG. 7C shows filter coefficients applied by the second adaptive filter unit 36 of the audio canceller unit 3. Specifically, the FIR filter length of the FIR unit 25 in the adaptive filter unit 24 of the array microphone unit 2 is 128 tap, the FIR filter length in the first adaptive filter unit 35 of the audio canceller unit 3, and the FIR filter of the second adaptive filter unit 36 The length is set to 192 taps, and the sampling frequency in each FIR unit is set to 6 kHz.

  Comparing the frequency characteristics before and after applying the adaptive filter units 24, 35, and 36, as shown in FIG. 6, the output signal level in the array microphone unit 2 is higher than the output signal level of the omnidirectional microphone M1. The output value is attenuated by about 10 dB. Further, the output level of the first adaptive filter unit 35 of the audio canceller unit 3 is attenuated by about 8 dB with respect to the output signal level of the array microphone unit 2, and the output signal level of the second adaptive filter unit 36 is About 1 dB is attenuated with respect to the output level of one adaptive filter unit 35.

  Thus, compared to the signal level until the adaptive filter unit 24 is applied, the signal level after all the adaptive filter unit 24, the first adaptive filter unit 35, and the second adaptive filter unit 36 are applied is 30 dB in total. The output value is attenuated nearby, and as a result, the D / U is greatly improved. In addition, in the filter coefficient of each adaptive filter shown in FIGS. 7A to 7C, since the response of the FIR filter continues, the D / U can be further improved by increasing the filter tap length. You can expect.

[Volume correction section]
Next, the volume correction unit 4 will be described. FIG. 8 is a block diagram showing a schematic configuration of the volume correction unit 4. As shown in FIG. 8, the volume correction unit 4 includes a three-band bandpass filter unit 51, a maximum value detection and maximum value hold unit 52, a gain calculation unit 53, a delay unit 54, and a gain setting unit 55. Have.

  The 3-band bandpass filter unit 51 frequency-divides the L-channel music signal L1 and the R-channel music signal R1 output from the in-vehicle audio signal into three bands, a low band, a middle band, and a high band, respectively. Have a role.

  FIG. 9 is a block diagram illustrating a functional unit used for frequency division of the L-channel music signal L1 into three bands, a low band, a middle band, and a high band, in the three-band bandpass filter unit 51. The functional units for frequency-dividing the music signal L1 of the three-band bandpass filter unit 51 include a first low-pass filter unit 56, a second low-pass filter unit 57, a delay unit 58, and two addition units 59 and 60. It is roughly composed. In FIG. 9, for convenience, only the functional unit of the three-band bandpass filter unit 51 for frequency-dividing the L-channel music signal L1 is shown, but the actual three-band bandpass filter unit 51 is In addition to the configuration for frequency division of the L channel music signal L1 shown in FIG. 9, the configuration for frequency division of the R channel music signal R1 is also provided.

  The first low-pass filter unit 56 and the second low-pass filter unit 57 are configured by FIR type low-pass filters, and each of the first low-pass filter unit 56 and the second low-pass filter unit 57 is as shown in FIG. Has filter characteristics. In the present embodiment, the sampling frequency is set to 48 kHz, the filter characteristic of the first low-pass filter unit 56 is that the FIR filter length is 128 tap, the cutoff frequency is 400 Hz, and the filter characteristic of the second low-pass filter unit 57 is FIR filter The length is set to 128 tap and the cut-off frequency is set to 4 kHz. The delay unit 58 is provided to adjust the delay by applying the first low-pass filter unit 56 and the second low-pass filter unit 57. The delay unit 58 includes the first low-pass filter unit 56 and the second low-pass filter unit 56. It is set to 64 tap, which is half the filter length of the low-pass filter unit 57.

  Based on the configuration shown in FIG. 9, the adder 60 subtracts the output signal from the second low-pass filter 57 from the music signal (all band (ALL) signal) subjected to delay processing by the delay unit 58. By doing so, a high frequency signal (High) of 4 kHz to 24 kHz is generated. Further, the adder 59 subtracts the output signal (low-frequency signal) from the first low-pass filter unit 56 from the output signal from the second low-pass filter unit 57 to generate a mid-range signal (Mid). Is done. Further, a low-frequency signal (Low) is generated by the filter processing in the first low-pass filter unit 56, and further, a signal of the entire band subjected to the delay processing in the delay unit 58 is generated. Therefore, the L-channel music signal L1 input in the 3-band bandpass filter unit 51 is separated into a low-frequency music signal, a mid-frequency music signal, a high-frequency music signal, and a full-band music signal. Will be output.

  Note that, as described above, the 3-band bandpass filter unit 51 includes not only the configuration for the music signal L1 (L channel) but also a functional unit that combines the configuration for the music signal R1 (R channel). The same operation is performed not only on the L channel side as described above but also on the R channel side.

  The maximum value detection and maximum value hold unit 52 synthesizes the music signal (ALL) of the entire band for the music signal L1 (L channel) and the music signal R1 (R channel) from the three-band bandpass filter unit 51. After that, the maximum value of the predetermined section is detected, and the maximum value is held for a predetermined time and is output as a control signal (maximum value hold signal).

  In the present embodiment, the predetermined interval is set to 2 msec, and the hold time is set to 8 times the predetermined interval. In the present embodiment, as shown in FIG. 11, the sampling frequency is 48 kHz, the maximum value of the 96-sample section is calculated and output every 2 msec, and the maximum 32 msec is held.

  Next, the gain calculation unit 53 will be described. FIG. 12 is a block diagram showing a schematic configuration of the gain calculation unit 53. As shown in FIG. 12, the gain calculation unit 53 includes a first attack release filter unit 63, a second attack release filter unit 64, a third attack release filter unit 65, a first lookup table unit 67, A second lookup table unit 68; a third lookup table unit 69; a first low-pass filter unit 71; a second low-pass filter unit 72; and a third low-pass filter unit 73.

  The first attack release filter unit 63 to the third attack release filter unit 65 have a role of setting the response speed of the control signal received from the maximum value detection and maximum value hold unit 52, and as a response speed setting method. Specifically, the setting is made by the attack time and the release time. The attack time and the release time can be arbitrarily set. In this embodiment, the attack time and release time for the high frequency signal and the low frequency signal are higher than those for the mid frequency signal. The attack time and release time are set so that the set time becomes longer (the control speed becomes slower).

  The first look-up table unit 67 to the third look-up table unit 69 have a role of performing level conversion of the control signal received from the maximum value detection and maximum value hold unit 52, and the inputted input signal is constant. The signal is output as an output signal after level conversion to the output value.

  FIG. 13 is a diagram illustrating an example of the level conversion operation of the first lookup table unit 67 to the third lookup table unit 69 according to the present embodiment. When the signal level of the input signal (control signal) received from the maximum value detection and maximum value hold unit 52 is −30 dB to 0 dB, the output level is −16 dB (low frequency control signal) in the first lookup table unit 67. In the second lookup table unit 68, the output level is controlled to -20 dB (middle range control signal), and in the third lookup table unit 69, the output level is controlled to -18 dB (high range control signal). Is done. Further, when the signal level of the input signal (control signal) received from the maximum value detection and maximum value hold unit 52 is −30 dB or less, the first lookup table unit 67 to the third lookup table unit 69 input the signal. The output level is set so that the control signal is attenuated while maintaining a predetermined gain.

  In this way, when the signal level of the input signal (control signal) is -30 dB to 0 dB, the signal levels of the output signals of the low-frequency control signal, the mid-frequency control signal, and the high-frequency control signal are converted to constant values. In the case of -30 dB or less, the value of the output signal is attenuated so as to correspond to the input signal, so that the volume fluctuation is reduced and the sense of incongruity is eliminated, and the volume is corrected so as to have a constant amplitude. Can be done.

  The contents of level conversion in the first look-up table unit 67 to the third look-up table unit 69 are not limited to the relationship between the input signal and the output signal shown in FIG. 13, but the relationship shown in FIG. May perform output level conversion based on different level conversions

  The first low-pass filter unit 71 to the third low-pass filter unit 73 respectively receive the control signals (low-frequency control signal, mid-frequency control signal, high-frequency signal received from the first lookup table unit 67 to the third lookup table unit 69. Control signal) is smoothed, and the smoothed control signals (low frequency control signal 2, middle frequency control signal 2, high frequency control signal 2) are gain set from gain calculation unit 53. The data is output to the unit 55.

  The delay unit 54 includes L channel and R channel signals (L channel and R channel low band signals, mid band) frequency-divided into three bands, a low band, a middle band, and a high band in the three band band pass filter section 51. Signal, high-frequency signal). Due to the delay processing by the delay unit 54, it is possible to perform delay correction for the attack time in the first attack release filter unit 63 to the third attack release filter unit 65 of the gain calculation unit 53.

  The gain setting unit 55 applies the gain calculation unit 53 to the L channel and R channel signals (L channel and R channel low-frequency signals, mid-frequency signals, and high-frequency signals) subjected to the delay processing in the delay unit 54. Further, the control signal (low frequency control signal 2, middle frequency control signal 2, high frequency control signal 2) output to the gain setting unit 55 is combined to correct the amplitude to be constant. is doing.

  FIG. 14 is a block diagram showing a schematic configuration of the gain setting unit 55. As shown in FIG. 14, the gain setting unit 55 includes a multiplication unit 74a that multiplies the low frequency signal by the low frequency control signal 2, and a multiplication unit 74b that multiplies the mid frequency signal by the mid frequency control signal 2. A multiplication unit 74c that multiplies the high-frequency signal by the high-frequency control signal 2, and an addition unit 74d that adds the signals multiplied by the respective control signals in the multiplication units 74a to 74c. The signal added in 74d is output to the main volume unit 5 and the audio canceller unit 3 as a music signal L2. 14 shows only the configuration on the L channel side (multiplication units 74a to 74c and addition unit 74d) for convenience, the gain setting unit 55 is on the R channel side corresponding to the configuration on the L channel side. The music signal R2 is output by multiplying and adding each control signal not only to the L channel but also to the R channel signal.

  FIG. 15 shows a maximum value hold signal (control signal) output from the maximum value detection and maximum value hold unit 52 on the basis of the low frequency signal in the sound volume correction unit 4 and a first attack release filter unit of the gain calculation unit 53. 16 shows an output signal (low frequency control signal) output from 63 and a low frequency control signal 2 output from the gain calculation unit 53 (output from the first low pass filter unit 71). The maximum value hold signal (control signal) output from the maximum value detection and maximum value hold unit 52 and the second attack release filter unit 64 of the gain calculation unit 53 based on the mid-range signal in the volume correction unit 4 FIG. 6 is a diagram showing an output signal (middle range control signal) to be output and a middle range control signal 2 output from the gain calculation unit 53 (output from the second low-pass filter unit 72). That.

  Note that as the set values of the attack time and release time in the first attack release filter unit 63 for the low frequency signal shown in FIG. 15, the attack time is set to 10 msec and the release time is set to 0.5 sec. The setting of the attack time and the release time in the third attack release filter unit 65 in the case of the high frequency signal is the attack time as in the case of the first attack release filter unit 63 (in the case of the low frequency signal). 10 msec and a release time of 0.5 sec are set.

  On the other hand, the setting of the attack time and the release time in the second attack release filter unit 64 for the mid-range signal shown in FIG. 16 is set to 2 sec as the attack time and 2 sec as the release time. In this way, by slowing down the control speed of the mid-range signal, the sense of fluctuation in the volume of the vocal (speech) region (mainly mid-range sound) of music that is audibly sensitive (easy to hear) is reduced. It is possible to correct the volume so that the feeling of discomfort is eliminated and the amplitude is constant.

  Note that the diagrams shown in FIGS. 15 and 16 show an example in which the signal level of the sound source is relatively high. On the other hand, the low-frequency signal (or high-frequency signal) when the signal level of the sound source is low is shown as in FIG. When the signal level of the sound source is low, the value of the output signal is level-converted to a value higher than the input signal in the first lookup table unit 67 to the third lookup table unit 69 as shown in FIG. Therefore, the control signal output from the gain calculation unit 53 becomes 0 dB or more, and the volume is positively increased.

  18A shows a signal state when the volume correction unit 4 does not perform volume correction, and FIG. 18B shows a signal state when the volume correction unit 4 performs volume correction. It is a figure. Comparing FIG. 18 (a) and FIG. 18 (b), when the volume correction is performed (FIG. 18 (b)), compared with the case where the volume correction is not performed (FIG. 18 (a)). It can be seen that the signal level is controlled to be small for a sound source having a large amplitude, and the signal level is controlled to be large for a sound source having a small amplitude.

  In this way, by performing the volume correction in the volume correction unit 4, it is possible to correct the volume so that a sense of fluctuation in the volume is reduced and there is no sense of incongruity, and a constant amplitude is obtained. Although not applied in the present embodiment, it is possible to further stabilize the amplitude by applying a combination of limiters in the volume correction in the volume correction unit 4.

[Main volume section]
The main volume unit 5 has a role of adjusting the volume of the music signal whose volume has been corrected by the volume correction unit 4 in accordance with the volume set by the passenger or the like (the amount of operation of the volume adjustment switch). .

  The volume adjustment by the main volume unit 5 is performed in conjunction with the volume set in the in-vehicle audio apparatus or based on the setting of a volume switch provided separately from the in-vehicle audio apparatus. Information relating to volume adjustment performed in the main volume unit 5 (hereinafter referred to as volume information) is output to the sound detection unit 7.

[Speech analysis section]
Next, the voice analysis unit 6 will be described. FIG. 19 is a block diagram showing a schematic configuration of the voice analysis unit 6. As shown in FIG. 19, the voice analysis unit 6 includes an effective value detection unit 75, a standard deviation detection unit 76, an average unit 77, a first moving average unit 78, a second moving average unit 79, and a division unit. 80 and a level converter 81.

  The effective value detection unit 75 has a role of detecting an effective value in a predetermined section in the output signal of the audio canceller unit 3. The standard deviation detection unit 76 has a role of detecting the standard deviation of a predetermined section with respect to the signal whose effective value is detected by the effective value detection unit 75. Further, the average unit 77 is an effective value detection unit. The unit 75 has a role of detecting the average value of the signal for which the effective value is detected.

  The first moving average unit 78 and the second moving average unit 79 have a role of taking a moving average of a predetermined section of the input signal, and the dividing unit 80 is moved and averaged by the first moving average unit 78. A voice analysis value is calculated by dividing the standard deviation by the average value obtained by moving average in the second moving average unit 79.

  The level conversion unit 81 has a role of outputting a voice analysis gain by setting a gain and an offset for the voice analysis value calculated by the division unit 80. Here, the gain and the offset are used to set a weight for voice analysis in the voice detection of the voice detection unit 7.

  FIG. 21A shows a music signal as shown in FIG. 20A (FIG. 20A shows only an L channel music signal for convenience, but an R channel music signal actually exists. ) And an audio signal as shown in FIG. 20B are input to the microphone M1 and the microphone M2, and the audio analysis value for the signal input from the audio canceller unit 3 to the audio analysis unit 6 is shown. FIG. 21B shows the level conversion of the voice analysis unit 6 when the same music signal (FIG. 20A) and voice signal (FIG. 20B) are input to the microphone M1 and the microphone M2. An output signal (speech analysis gain) output from the unit 81 is shown.

  21A and 21B, the sampling frequency in the voice analysis unit 6 is 6 kHz, the effective value detection section in the effective value detection unit 75 is 2.7 msec, and the standard deviation of the standard deviation detection unit 76 is. And the average interval of the average unit 77 is set to about 340 msec, and the moving average interval between the first moving average unit 78 and the second moving average unit 79 is set to about 2.7 sec.

  When the voice analysis value shown in FIG. 21 (a) is observed while comparing with the state of the music signal shown in FIG. 20 (a) and the state of the voice signal shown in FIG. 20 (b), the voice signal in FIG. In the section of FIG. 21 (a) corresponding to the section in which is detected (for example, the section indicated by the arrow in FIG. 21 (a)), the voice analysis value is increased to indicate the presence of voice. , Voice detection can be confirmed.

  On the other hand, the voice analysis gain shown in FIG. 21B is obtained by setting the gain to 0.5 and the offset to 1 with respect to the voice analysis value shown in FIG. The value is used as a weight for voice detection in the voice detector 7 described below.

[Audio detector]
Next, the voice detection unit 7 will be described. FIG. 22 is a block diagram illustrating a schematic configuration of the voice detection unit 7. The voice detection unit 7 includes an effective value detection unit 85, a moving average unit 86, a voice analysis gain multiplication unit 87, a voice detection threshold unit 88, an adaptive attack release filter unit 89 (adaptive attack release filter unit), a tempo A detection unit (tempo detection means) 90 and a voice holding threshold unit 91 are provided.

  The effective value detection unit 85 has a role of detecting an effective value in a predetermined section in the output signal of the audio canceller unit 3. The moving average unit 86 has a role of obtaining a moving average of a predetermined section with respect to the signal whose effective value is detected by the effective value detecting unit 85. The voice analysis gain multiplication unit 87 has a role of multiplying the voice analysis gain input from the voice analysis unit 6 and the signal whose moving average is obtained by the effective value detection unit 85.

  The voice detection threshold unit 88 has a role of detecting a voice signal based on a preset voice detection threshold (threshold). Note that the value of the voice detection threshold in the voice detection threshold unit 88 changes according to the volume information input from the main volume unit 5, and for example, when the volume of the main volume unit 5 increases by 5 dB, The detection threshold is also increased by 5 dB in conjunction with the detection threshold.

  FIG. 23 shows that the music signal as shown in FIG. 20A (including not only the L channel music signal but also the R channel music signal) and the audio signal as shown in FIG. 2 shows the input signal (speech signal × speech analysis gain) of the speech detection threshold unit 88 and the speech detection threshold set in the speech detection threshold unit 88 in the case of input from M2.

  In the situation shown in FIG. 23, the sampling frequency in the voice detector 7 is set to 6 kHz, the effective value detection section in the effective value detector 85 is set to about 21 msec, and the moving average section in the moving average section 86 is set to about 42 msec. Referring to FIG. 23, in the array microphone unit 2 and the audio canceller unit 3, the audio signal is emphasized by greatly improving the D / U, and the audio detection portion is further detected by the audio analysis gain input from the audio analysis unit 6. Since it is manifested (emphasized), it is easy to detect the voice detection threshold. Note that the sound volume correction processing in the sound volume correction unit 4 is controlled so that the signal level is reduced for a large amplitude sound source, and is controlled so that the signal level is increased for a small amplitude sound source. Therefore, it is possible to set the voice detection threshold without depending on the source and genre of the sound source.

  The tempo detection unit 90 has a role of changing a voice detection time and a holding time in the adaptive attack release filter unit 89 described below in accordance with the utterance speed (speech signal input speed) of the speaker. Thus, it is possible to perform optimum volume control by adjusting the detection accuracy according to the speaking speed of the speaker.

  FIG. 24 is a block diagram showing a schematic configuration of the tempo detection unit 90. As shown in FIG. 24, the tempo detection unit 90 includes a threshold crossing detection unit 94, a crossing gain unit 95, a moving average unit 96, a multiplication unit 97, a tempo gain unit 98, a low-pass filter unit 99, a gain And an offset portion 100.

  The threshold crossing detection unit 94 has a role of detecting the number of pulses while shifting a predetermined interval one sample at a time with respect to the voice detection signal detected based on the voice detection threshold by the voice detection threshold unit 88. The crossing gain unit 95 has a role of weighting the number of pulses detected by the threshold crossing detection unit 94. The moving average unit 96 has a role of obtaining a moving average of a predetermined section with respect to the voice detection signal detected by the voice detection threshold unit 88 based on the voice detection threshold, and the multiplication unit 97 includes a crossing gain unit. It has the role of multiplying the output of 95 and the output of the moving average unit 96.

  The tempo gain unit 98 has a role of weighting the output multiplied by the multiplication unit 97. In the low-pass filter unit 99, an output signal weighted by the multiplication unit 97 (that is, input to the low-pass filter unit 99). The signal offset) has a role of performing a process (integration process) for obtaining a change state for a predetermined time in the signal) using an integral value. Further, the gain offset unit 100 performs a gain offset process and a rounding process to detect a tempo detection value. Has a role to output. Here, the integration process of the low-pass filter unit 99 uses a first-order IIR (Infinite Impulse Response) filter. Further, when a reset signal is input from the outside, the integration obtained by the integration process is used. It is possible to clear the value. As described above, the learning function in the volume control can be realized by clearing the integration value obtained by the integration processing in accordance with the reset signal.

  An example of the operation of the tempo detection unit 90 is shown in FIG. FIG. 25A shows an audio detection signal binarized by detecting the audio detection threshold in the audio detection threshold unit 88, and FIG. 25B shows a tempo based on the binarized audio detection signal. FIG. 25C shows an integrated value (integrated output) obtained by the low-pass filter unit 99 and a tempo detection value outputted by the gain offset unit 100. FIG.

  In the tempo detection unit 90 according to the present embodiment, the predetermined interval for detecting pulses in the crossing gain unit 95 is about 1 sec, the moving average interval in the moving average unit 96 is about 1 sec, and the crossing in the crossing gain unit 95 is The gain is set to 0.5, the tempo gain in the tempo gain unit 98 is set to 20, the normalized cutoff frequency of the filter in the low-pass filter unit 99 is set to 0.0002, and the gain offset in the gain offset unit 100 is set to 1. . 25 (a) to 25 (c), it can be understood that the tempo detection value shown in FIG. 25 (c) changes according to the voice detection signal shown in FIG. 25 (a).

  The reset signal may be input at any timing. For example, when the number of passengers or members of the car changes and the conversation state before the change is likely to change, the passenger May be reset by operating an operation switch or the like (or outputting a reset signal), or may be reset each time the vehicle is started (every time the engine is started).

  Next, the adaptive attack release filter unit 89 will be described. The adaptive attack release filter unit 89 has a role of setting an attack time at the start of the conversation and setting a release time at the end of the conversation based on the audio signals detected by the microphones M1 and M2. The adaptive attack release filter unit 89 makes the attack time and release time set in the adaptive attack release filter unit 89, that is, the voice detection time and the holding time variable according to the tempo detection value obtained by the tempo detection unit 90. There are two modes: a “mode” and a “fixed mode” in which the voice detection time and the holding time are fixed values without depending on the tempo detection value obtained by the tempo detection unit 90.

  First, the case of the fixed mode will be described, and the variable mode will be described later.

  FIG. 26A shows a voice detection signal input from the voice detection threshold unit 88 in the fixed mode, and FIG. 26B shows an application operation example of the voice holding filter in the adaptive attack release filter unit 89. FIG. In the case shown in FIG. 26 (b), the attack time and the release time are fixed (fixed mode). Specifically, the attack time is fixed at 0.2 sec and the release time is fixed at 4 sec. Has been. Therefore, when an audio signal is detected, the signal output value increases in 0.2 sec by applying the audio holding filter, and when the detection of the audio signal is completed, the signal output is reduced over 4 sec. The

  The voice holding threshold unit 91 has a role of detecting the voice interval time based on a preset voice holding threshold (threshold). FIG. 26B shows a voice holding threshold value set in the voice holding threshold unit 91, and is set to 0.3 in the present embodiment.

  In the fixed mode, as described above, the binarized voice detection signal as shown in FIG. 26A is input to the adaptive attack release filter unit 89 based on the voice detection threshold detection of the voice detection threshold unit 88. Then, the adaptive attack release filter unit 89 applies the attack time and the release time to the voice detection signal. Of the voice detection signal (voice holding filter) to which the attack time and release time are applied, the time exceeding the voice holding threshold value set in the voice holding threshold unit 91 is the voice interval time (voice detection and holding time). ) Is obtained by the voice holding threshold unit 91. FIG. 27A shows a voice interval time (voice interval detection value) obtained based on the signal states of FIGS. 26A and 26B. As described above, the case where the adaptive attack release filter unit 89 is in the variable mode will be described later.

[Volume control section]
Next, the volume control unit 8 will be described. FIG. 28 is a block diagram showing a schematic configuration of the volume control unit 8. As shown in FIG. 28, the volume control unit 8 includes an attack release filter unit 101, a level limiter unit 102, a level calculation unit 103, and multiplication units 104 and 105.

  The attack release filter unit 101 has a role of setting the fade-in time and fade-out time of the music being played based on the voice interval time detected by the voice detection unit 7.

  The level calculation unit 103 has a role of calculating a volume control amount for the level limiter unit 102 in accordance with the volume information input from the main volume unit 5. Specifically, in the level calculation unit 103 according to the present embodiment, as shown in FIG. 29, a change in volume control level that changes in response to an increase or decrease in volume information (volume adjustment amount) According to the above, it is possible to set from a plurality of types (in FIG. 29, four types are shown as an example). Thus, the change state of the volume control level with respect to the volume information (volume adjustment amount) is arbitrarily selected, and the level calculation unit 103 corresponds to the volume information from the main volume unit 5 based on the selected change state. Determine the volume control level.

  Further, the level limiter unit 102 has a role of performing adjustment processing on the output signal of the attack release filter unit 101 based on the volume control level obtained by the level calculation unit 103.

  FIG. 27B shows a change state of the output signal of the attack release filter unit 101 in the voice detection signal (voice interval time) as shown in FIG. 27A, and FIG. A change in the output signal (volume control value) is shown. In this embodiment, the attack time of the attack release filter unit 101 is set to 0.1 sec, the release time is set to 5.0 sec, and the volume control amount is set to 24 dB (0.0631 in linear). Yes. Also, the attack time set in the attack release filter unit 101 corresponds to a music fade-out time, and the release time corresponds to a music fade-in time.

  As shown in FIGS. 27B and 27C, the level limiter unit 102 inverts the increase / decrease change of the output signal of the attack release filter unit 101 so that the volume control value becomes 1 in a section where no sound is detected. Conversion is performed so that the volume control value is reduced to about 0.05 in a section in which sound is detected. As shown in FIG. 27C, the level limiter unit 102 changes / sets the volume control value to a value lower than 1 (for example, a value close to 0) at the time of detecting the sound, and sets the volume control value to 1 at the time of detecting the sound. By changing and setting to the value of, the signal level of the music signal can be reduced in conjunction with the timing of voice detection.

  The multipliers 104 and 105 have a role of performing multiplication processing of the signal output (volume control value) of the level limiter unit 102 and the two-channel music signals (music signal L3 and music signal R3) in the main volume unit 5. ing. In the multipliers 104 and 105, the volume control signal shown in FIG. 27C is multiplied by the music signal L3 and the music signal R3, respectively, and the volume-controlled music signal L4 and music signal R4 are the power amplifier section. 9 is output. The music signals L4 and R4 subjected to volume control in the multipliers 104 and 105 are output from the speakers 10a and 10b via the power amplifier unit 9.

  FIG. 20C shows a music signal as shown in FIG. 20A (FIG. 20A shows only an L channel music signal for convenience, but an R channel music signal actually exists. ) Is reproduced by the in-vehicle audio apparatus, and audio signals (volume control) are output from the speakers 10a and 10b when an audio signal as shown in FIG. 20B is input to the microphone M1 and the microphone M2. The music signal for which the volume control is performed in the unit 8) is shown. As can be seen by comparing FIG. 20A to FIG. 20C, when a conversation is performed and the voice of the speaker is collected in the microphone M1 and the microphone M2, the collected sound is collected. The music signal is output in a state where the signal level of the music signal is reduced in accordance with the timing of the sound. For this reason, the party who is having a conversation can enjoy a conversation without feeling uncomfortable in the vehicle without being hindered by the music that is controlled to be reduced according to the conversation.

  Further, as described above, the voice interval time (voice detection time and holding time) detected by the voice detection unit 7 is a fixed value, and the music fade-out time (attack release filter) is based on the voice interval time. The attack time of the section 101 is set to 0.1 sec, and the fade-in time of the music (release time of the attack release filter section 101) is set to 5.0 sec. Therefore, as shown in FIG. When the audio signal is acquired by the microphones M1 and M2, the signal level of the music information is quickly reduced. On the other hand, when the acquisition of the audio signal is completed, the signal of the music signal is slowly maintained for a while (5.0 sec). Returned level.

  Next, the case where the setting in the adaptive attack release filter unit 89 of the voice detection unit 7 is “variable mode” will be described.

  In the case of the variable mode, the attack time and release time set in the adaptive attack release filter unit 89, that is, the voice detection time and the holding time change according to the tempo detection value of the tempo detection unit 90.

  FIG. 30 is a diagram showing the relationship between the attack time (sound detection time) and the release time (holding time) determined according to the tempo detection value obtained by the tempo detection unit 90. In FIG. 30, when the tempo detection time is less than 3, that is, when an instantaneous road change sound during vehicle travel is acquired by the microphones M1 and M2, or when a short utterance such as a monologue is acquired by the microphones M1 and M2. When it can be determined that the voice speed of the voice signal is high and the voice interval is long (as in “(A) road surface change sound, short utterance” shown in FIG. 30), the voice signal is detected. Even if it is detected even if it is not performed, it is possible to prevent the signal level of the music signal from being inadvertently reduced when a road change sound or a short utterance is performed by shortening the holding time. Furthermore, even if the signal level is reduced, the music information can be restored to the original signal level in a short time.

  Further, when the tempo detection time required by the tempo detection unit 90 is 3 or more and 7 or less, that is, when it can be determined that the utterance speed is slow and the utterance interval is slightly long ("(B shown in FIG. 30" ) When the voice speed is slow and the voice interval is slightly long), the attack time and the release time are set so that the attack time for voice detection is short and the release time for holding the voice is long. In this way, by setting the attack time and the release time, the signal level of the music signal can be reduced according to the start of the conversation, and it is determined that the voice interval is long. The level can be restored slowly so that there is no sense of incongruity.

  Further, when the tempo detection time required by the tempo detection unit 90 is 7 or more, that is, when it can be determined that the voice speed is high and the voice interval is short (“(C) voice speed: fast shown in FIG. 30”. In the case of “sound interval: short”), the sound holding time is shortened by shortening the release time compared to the case where the tempo detection time is 3 or more and 7 or less. It is possible to improve the responsiveness in returning the level.

  FIG. 31A shows a case where the sound speed is short (detection time is short) and the sound interval is very long (the detection interval is very long) in the detection state of the sound signal (“(A) shown in FIG. 30”). FIG. 31B shows an application operation example of the speech holding filter of the adaptive attack release filter unit 89 in the fixed mode. FIG. 31 (c) shows an example of the adaptive operation of the speech holding filter of the adaptive attack release filter unit 89 in the same variable mode as FIG. 31 (a).

  As shown in FIG. 31 (b), in the fixed mode, the value of the voice holding filter exceeds the voice holding threshold, so that voice holding is performed and volume control for a predetermined time is executed. On the other hand, as shown in FIG. 31 (c), in the variable mode, the value of the voice holding filter does not exceed the voice holding threshold, so that the volume control is not performed. For this reason, since the volume control by the automatic volume control device 1 is not performed for a short utterance such as an instantaneous road surface change sound or monologue when the vehicle is running, the volume correction corresponding to the unexpected volume change can be suppressed. It is possible to prevent the output of a music signal that makes the passenger in the vehicle feel uncomfortable.

  On the other hand, FIG. 32A shows the detection state of the sound detection value when the sound speed is slightly short (the detection time is slightly short) and the sound interval is slightly long (the detection interval is slightly long) in the sound signal detection state. FIG. 32B shows an application operation example of the speech holding filter of the adaptive attack release filter unit 89 in the case of the fixed mode, and FIG. 32C shows the same variable mode as that in FIG. 10 shows an adaptive operation example of the speech holding filter of the adaptive attack release filter unit 89.

  As shown in FIG. 32 (b), in the fixed mode, since the value of the sound holding filter may be lower than the sound holding threshold, the volume control is likely to be released from time to time, but on the other hand, FIG. 32 (c) As shown in FIG. 4, in the variable mode, the volume control is performed in the first stage, but based on the integration result of the detection value in the tempo detection unit 90, that is, based on the learning function, it can be gradually determined that the speech is being performed, Since it will not fall below the threshold, the volume control can be maintained, and it is possible to prevent the volume from changing unexpectedly and causing the listener to feel uncomfortable.

  As described above, in the automatic volume control device 1 according to the present embodiment, when sound is detected in the microphone M1 and the microphone M2, the output volume of music output from the in-vehicle audio device is automatically reduced. Therefore, it is possible to perform a smooth conversation without operating the volume control switch and the mute switch each time a conversation is performed.

  In particular, the automatic volume control device 1 according to the present embodiment employs a configuration in which the directivity is emphasized by using the omnidirectional microphone M1 and the unidirectional microphone M2, so that the speaker's It is possible to acquire sound with high accuracy.

  Further, by applying the NLMS adaptive algorithm in the adaptive filter unit 24 of the array microphone unit 2 and further applying the LMS adaptive algorithm in the first adaptive filter unit 35 and the second adaptive filter unit 36 of the audio canceller unit 3, It is possible to reduce signal components (noise components) other than the audio signal component in FIG. 2 while ensuring effective and high convergence, and it is possible to improve the detection accuracy of the audio signal in the audio band.

  In particular, the adaptive filter unit 24 of the array microphone unit 2 applies the NLMS adaptive algorithm to reduce noise components, and further, according to the number of channels of the sound source, the first adaptive filter unit 35 of the audio canceller unit 3. In the second adaptive filter unit 36, a configuration in which the adaptive filter units are cascade-connected is adopted, and the adaptive speed is increased for each portion to which the filter processing is applied earlier in the filter processing in each adaptive filter unit. It is possible to quickly converge the signal to which the processing is applied.

  Further, the band limiting width set in the first band pass filter unit 31 and the second band pass filter unit 32 of the audio canceller unit 3 is the band limiting width set in the first band pass filter unit 21 of the array microphone unit 2. Since a wider bandwidth is set, it is possible to improve the audio cancellation performance near the cutoff frequency of the band limit.

  A signal excellent in D (desired signal: audio signal) / U (undesired signal: music signal) can be obtained by applying a plurality of adaptive algorithms as described above and setting the adaptive limit width of the bandpass filter.

  Further, in the gain calculation unit 53 of the volume correction unit 4, the attack time and release time set for the low-frequency control signal and the high-frequency control signal in the first attack release filter unit 63 and the third attack release filter unit 65 are set. In comparison, by setting the attack time and release time set for the mid-range control signal in the second attack release filter unit 64 as long, the control processing for the mid-range signal can be delayed. For this reason, it is possible to reduce the volume fluctuation in the vocal sound area of music that is audibly and audibly audible (easy to hear), eliminate the sense of incongruity, and perform volume correction so that the amplitude is constant.

  Further, in the gain calculation unit 53 of the volume correction unit 4, the output signal levels of the low-frequency control signal, the mid-range control signal, and the high-frequency control signal are input to the first lookup table unit 67 to the third lookup table unit 69. Is equal to or higher than a predetermined value (in this embodiment, −30 dB or higher), the signal level of the output signal is set to a constant value, and the signal level of the input signal is set to a predetermined value (this embodiment). In the case of −20 dB to −16 dB), the signal level of the output signal is converted so that the signal level is higher than that of the input signal. Volume correction can be performed.

  In addition, when the voice detection threshold unit 88 of the voice detection unit 7 detects the voice interval time, an output signal having an excellent D / U by the processing of the array microphone unit 2 and the audio canceller unit 3 described above. Since the voice analysis gain with weight applied to the voice detection part is applied, the voice detection part in the output signal can be revealed (emphasized), and the presence / absence of voice can be prepared and reliably determined It becomes. In addition, since the voice detection threshold is set for the signal in which the voice detection part is made obvious and the voice interval time is detected when the voice is heard, it depends on the source and genre of the sound source. Therefore, the voice detection threshold can be easily set, and the detection accuracy of the voice interval time can be improved.

  Furthermore, since the voice detection threshold set in the voice detection threshold unit 88 of the voice detection unit 7 is determined based on the volume information input from the main volume unit 5, the voice detection threshold is set according to the volume adjustment switch operation. It becomes possible to optimize in conjunction with the volume adjustment amount (volume information). Therefore, the control amount of the audio signal can be controlled according to the volume, and the control amount can be arbitrarily set.

  Also, since the voice tempo (speaker's speech speed) can be detected and the voice detection time (fade-in time) and holding time (fade-out time) in volume control can be changed according to the tempo, there is no sense of incongruity. Volume control can be performed.

  Furthermore, since the determination contents of the voice detection time (fade-in time) and the holding time (fade-out time) can be calculated (learned) again according to the presence or absence of the reset input, the voice detection and holding time is appropriately changed. Therefore, it is possible to reduce a sense of incongruity with unexpected volume changes.

  Furthermore, since the volume control in the level calculation unit 103 of the volume control unit 8 can be performed according to the volume information of the main volume unit 5, the signal level (volume of the music signal) is appropriately adjusted according to the volume of the main volume unit 5. ) Can be changed.

  In the automatic volume control device 1, the attack time corresponding to the music fade-out time and the music fade-in time according to the sound detection signal acquired from the sound detection unit 7 in the attack release filter unit 101 of the volume control unit 8 are set. Since the corresponding release time is set, it is possible to arbitrarily set the music fade-out and fade-in time by changing the time setting.

  The automatic volume control device according to the present invention described above in detail with reference to the drawings for the automatic volume control device according to the present invention is not limited to the above-described embodiment. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present invention. Understood.

DESCRIPTION OF SYMBOLS 1 ... Automatic volume control apparatus 2 ... Array microphone part (array microphone means, audio | voice signal extraction means)
3 ... Audio canceller (audio canceller means, audio signal extraction means)
4 ... Volume correction section (volume correction means)
5 ... main volume part 6 ... voice analysis part (voice analysis means)
7: Voice detection unit (voice detection means)
8: Volume control unit (volume control means)
9... Power amplifier units 10a and 10b... Speaker 21... First band pass filter unit 22 (of array microphone unit) 2. Second band pass filter unit 23 (of array microphone unit)... Delay unit 24 of (array microphone unit). Adaptive filter unit 25 (of the array microphone unit) FIR unit 26 of the adaptive filter unit NLMS unit 27 of the adaptive filter unit Adder unit 28 of the adaptive filter unit Vehicle 28a Driver's seat 28b Passenger seat 31 ... First band pass filter unit 32 (of the audio canceller unit) ... Second band pass filter unit 33 (of the audio canceller unit) ... First delay unit 34 (of the audio canceller unit) ... Second (of the audio canceller unit) Delay unit 35 ... first adaptive filter unit 36 (of audio canceller unit) ... second adaptive filter (of audio canceller unit) Filter unit 37 ... First FIR unit 38 (of the first adaptive filter unit) ... Second FIR unit 39 (of the second adaptive filter unit) ... First LMS unit 40 (of the first adaptive filter unit) ... (of the second adaptive filter unit) ) Second LMS unit 41... First addition unit 42 (of the first adaptive filter unit)... Second addition unit 51 (of the second adaptive filter unit) 3 band bandpass filter unit 52 (of the volume correction unit). Maximum value detection and maximum value hold unit 53 (of correction unit) Gain calculation unit 54 (of volume correction unit) Delay unit 55 (of volume correction unit) Gain setting unit 56 (of volume correction unit) (3-band band) First low-pass filter unit 57 (of the pass filter unit) Second low-pass filter unit 58 (of the 3-band band-pass filter unit) Delay units 59, 60 (of the 3-band band-pass filter unit) (3-band band-pass) Adder unit 63 (of the filter unit) First attack release filter unit 64 (of the gain calculation unit) Second attack release filter unit 65 (of the gain calculation unit) Third attack release filter unit 67 (of the gain calculation unit) First look-up table unit 68 (for gain calculation unit) Second look-up table unit 69 (for gain calculation unit) Third look-up table unit 71 (for gain calculation unit) First (for gain calculation unit) Low-pass filter unit 72 ... second low-pass filter unit 73 (for gain calculation unit) third low-pass filter unit 74a, 74b, 74c (for gain calculation unit) multiplication unit 74d (for gain setting unit) ) Adder 75 ... RMS value detector 76 (of speech analyzer) ... Standard deviation detector 77 (of speech analyzer) ... Average unit 78 (of speech analyzer) ... (sound First moving average unit 79 (of the voice analysis unit) ... Second moving average unit 80 (of the voice analysis unit) ... Dividing unit 81 (of the voice analysis unit) ... Level conversion unit 85 (of the voice analysis unit) ... (Speech detection unit) ) Effective value detection unit 86 ... Moving average unit 87 (of voice detection unit) ... Voice analysis gain multiplication unit 88 (of voice detection unit) ... Voice detection threshold unit 89 (of voice detection unit) ... (of voice detection unit) Adaptive attack release filter (adaptive attack release filter means)
90 ... Tempo detection section (of the voice detection section) (tempo detection means)
91 ... Voice holding threshold part 94 (of the voice detection part) ... Threshold crossing detection part 95 (of the tempo detection part) ... Crossing gain part 96 (of the tempo detection part) ... Moving average part 97 (of the tempo detection part) ... (Tempo Multiplying unit 98 (of detection unit) Tempo gain unit 99 (of tempo detection unit) Low pass filter unit 100 (of tempo detection unit) Gain offset unit 101 (of tempo detection unit) Attack release filter (of volume control unit) Unit 102 ... level limiter unit 103 (for volume control unit) level calculation unit 104, 105 (for volume control unit) multiplication unit M1, M2 (for volume control unit) microphone

Claims (7)

An audio signal extracting means for extracting an audio signal related to the audio band as an acoustic signal by applying an adaptive algorithm and applying a band limiting process of the audio band component to the audio signal collected by the microphone;
Volume for performing volume correction of the music signal by dividing the music signal from the sound source for each band and maintaining the signal level in each divided band at a constant value when the signal level of the music signal is equal to or higher than a certain level Correction means;
From the sound signal extracted by the sound signal extraction means, a sound analysis means for obtaining a sound analysis gain in which the weight of the sound detection part is performed;
Applying the audio analysis gain obtained by the audio analysis unit to the audio signal extracted by the audio signal extraction unit, thereby revealing the audio detection part in the audio signal and indicating the presence or absence of audio detection Voice detection means for obtaining a detection signal;
A volume control unit that reduces the output level of the music signal that has undergone the volume correction by the volume correction unit based on the voice detection signal obtained by the voice detection unit when detecting the voice; Volume control device. The voice detection means further includes
Based on the voice detection signal, the voice detection value at every predetermined time in the voice detection portion is integrated to obtain a change in the detection state of the voice at the predetermined time, and the utterance of the speaker based on the obtained integration value Tempo detection means for calculating a tempo detection value for judging speed;
Based on the tempo detection value obtained by the tempo detection means, an attack time corresponding to the sound detection time and a release time corresponding to the sound detection holding time are determined, and the determined attack time and release time are determined. The automatic sound volume control apparatus according to claim 1, further comprising: an adaptive attack release filter unit that sets the sound detection signal. The tempo detection means recalculates the tempo detection value for judging the utterance speed of the speaker by clearing the integration value obtained by the integration processing based on the input of the reset signal. The automatic volume control device according to claim 2. The adaptive attack release filter means includes:
A variable mode for setting the attack time and the release time based on the tempo detection value; and a fixed mode for setting the attack time and the release time to a predetermined value regardless of the tempo detection value. The automatic volume control device according to claim 2 or 3. The audio signal extraction means includes
Array microphone means for extracting an audio signal associated with an audio band by applying an NLMS adaptive algorithm after performing a first band limiting process corresponding to an audio band component on the audio signal collected by the microphone; ,
Adaptation that is cascade-connected in accordance with the number of channels of the music signal after performing a second band limiting process corresponding to the audio band component on the audio signal related to the audio band extracted by the array microphone means 5. An audio canceller unit that applies a multistage LMS adaptive algorithm to the audio signal that has been subjected to the second band limiting process using a filter. The automatic volume control device according to item 1. The bandwidth limit width of the second bandwidth limit process in the audio canceller means includes an upper limit value and a lower limit value of the bandwidth limit width of the first bandwidth limit process in the array microphone means, and the bandwidth limit of the first bandwidth limit process. The automatic volume control device according to claim 5, wherein the automatic volume control device is set to have a bandwidth slightly wider than the width. The automatic volume control according to any one of claims 1 to 6, wherein the volume control means changes an output level reduction amount in the music signal in accordance with a volume state of the sound source. apparatus.

Images