JP5740914B2 - Audio output device - Google Patents

Description

  The present invention relates to an audio output device that outputs masker sounds.

  Conventionally, in offices, etc., speakers are attached to partitions, and the voice that is not related to the speaker's voice is output as a masker sound, making it difficult to hear the voice of the speaker in other nearby spaces Has been proposed (see, for example, Patent Document 1). Thereby, since it becomes difficult to understand the content of the speaker's speech, the privacy of the speaker can be maintained.

Japanese Patent Laid-Open No. 06-175666

  However, in the method of Patent Document 1, since the output position of the masker sound is fixed, the listener gets used to the masker sound, and by the so-called cocktail party effect, the speaker's voice is heard and the contents of the statement are understood. There is a fear.

  Therefore, an object of the present invention is to provide an audio output device that can appropriately suppress the cocktail party effect.

  The audio output device according to the present invention controls a masker sound generating unit that generates a masker sound, a plurality of speakers that output the masker sound, a localization position of the masker sound, and an audio signal related to the masker sound to the plurality of speakers. And a localization control unit to be supplied. The localization control unit dynamically changes the localization position of the masker sound. Specifically, the localization control unit randomly changes the localization position within a predetermined range centered on the predetermined position. The localization position can be changed by changing the delay amount of the audio signal supplied to the plurality of speakers.

  Further, the localization control unit sets the localization position at the highest probability with the predetermined position as the center, and sets the localization position with a lower probability as the distance from the center position increases. Can be dynamically changed. For example, the localization position is dynamically changed with a probability according to a Gaussian distribution. When the localization position is closer to the actual speaker position, the sound source position of the speaker and the sound source position of the masker sound are not separated from each other, and the masking effect is enhanced. However, if a third party hears the masker sound always from the same direction, they will get used to the ears, and the speaker's voice will be heard and understood by the cocktail party effect. Therefore, in order to suppress the cocktail party effect while keeping the masking effect high, the sound source position is dynamically changed, and the appearance probability of the localization position is high near the speaker position, and the appearance probability as it goes away Is preferably set to be low.

  The masker sound may be any sound, but it is desirable to provide a microphone that picks up the voice of the speaker and generate the masker sound based on the sound picked up by the microphone. For example, the voice of the speaker is held for a predetermined time, and is modified on the time axis or the frequency axis so that it does not make any meaning in the vocabulary (the conversation content cannot be understood). Or, output a general utterance voice that does not make any lexical meaning with the voice of multiple people including men and women, or approximate the frequency characteristics of this general voice formant to the voice of the speaker It is good also as what was made to do.

  In this case, the audio output device includes an echo canceller that cancels a pseudo echo signal that simulates an echo component from the speaker to the microphone, and that is supplied to the masker sound generation unit. It is desirable. As a result, the masker sound output from the speaker and wrapping around the microphone can be removed, and the masker sound can be generated based only on the voice of the speaker.

  According to this invention, since the output position of the masker sound changes dynamically, the cocktail party effect can be appropriately suppressed.

It is an arrangement drawing showing the composition of a masking system. It is a block diagram which shows the structure of a microphone, a speaker array, and an audio processing apparatus. It is a figure which shows the virtual sound source localization method by a speaker array. It is a figure explaining the dynamic change of a virtual sound source position. It is a flowchart which shows operation | movement of a speech processing unit. It is a block diagram which shows the structure of the audio | voice processing apparatus at the time of providing an echo canceller.

  FIG. 1 is a layout diagram showing the configuration of a masking system provided with an audio output device of the present invention. The masking system is installed at a dialogue counter such as a bank or dispensing pharmacy, for example, and emits a masker sound to the third party to prevent the third party from understanding the remarks of the person who is talking across the counter. To do.

  In FIG. 1, there are a speaker H1 and a listener H2 across the counter, and there are a plurality of third parties H3 at positions away from the counter. For example, the speaker H1 is a pharmacist explaining the medicine, the listener H2 is a patient listening to the medicine explanation, and the third person H3 is a patient waiting for the turn.

  A microphone 1 is installed on the upper surface of the counter. The microphone 1 mainly collects the voice of the speaker H1 using the voice around the counter. A speaker array 2 that outputs sound toward the third party is installed in a direction where the third party of the counter exists (downward in the drawing). Note that the speaker array 2 is installed so that the listener H2 can hardly hear the sound output from the speaker array, such as under a desk.

  The microphone 1 and the speaker array 2 are connected to the sound processing device 3. The microphone 1 picks up the voice of the speaker H1 and outputs it to the voice processing device 3. The voice processing device 3 generates a masker sound for masking the voice of the speaker H1 based on the voice of the speaker H1 collected by the microphone 1 and outputs the masker sound to the speaker array 2. At this time, the sound processing device 3 dynamically controls the sound source position (virtual sound source position) of the masker sound perceived by the third party H3 by controlling the delay amount of the sound signal supplied to each speaker of the speaker array 2. Change. As a result, the third party H3 can hear that the sound source position of the masker sound is constantly moving, and the cocktail party effect due to ear habituation can be appropriately suppressed.

  Hereinafter, a specific configuration and operation for realizing the masking system will be described. FIG. 2 is a block diagram showing configurations of the microphone 1, the speaker array 2, and the sound processing device 3. The sound processing device 3 includes an A / D converter 51, a control unit 72, a masker sound generation unit 73, a delay processing unit 8, and D / A converters 61 to D / A converter 68. The speaker array 2 includes eight speakers 21 to 28. The number of speakers in the speaker array is not limited to this example.

  The A / D converter 51 inputs the sound collected by the microphone 1 and converts it into a digital sound signal. Each digital audio signal converted by the A / D converter 51 is input to the masker sound generation unit 73.

  The masker sound generation unit 73 generates a masker sound for masking the speaker voice based on the speaker voice related to the input digital voice signal. The masker sound may be any sound, but is preferably one that suppresses the discomfort of a plurality of third parties H3 existing at positions away from the counter. For example, the voice of the speaker H1 is retained for a predetermined time, and is modified on the time axis or the frequency axis so that it does not make any lexical meaning (the conversation content cannot be understood). Alternatively, whether general-purpose utterance voices that do not make any lexical meaning are stored in the built-in storage unit (not shown), and are output from the general-purpose voices including men and women Alternatively, frequency characteristics such as a general voice formant may be approximated to the voice of the speaker H1. The masker sound may be mixed with a background sound such as an air conditioning sound. By listening to such a masker sound simultaneously with the voice of the speaker H1, the third party H3 has difficulty in understanding the content of the speech of the speaker H1. The generated masker sound is output to each of the delays 81 to 88 of the delay processing unit 8.

  The delays 81 to 88 of the delay processing unit 8 are provided corresponding to the speakers 21 to 28 of the speaker array 2, respectively, and individually change the delay amount of the audio signal supplied to each speaker. The delay amount of the delays 81 to 88 is controlled by the control unit 72.

  The control unit 72 can set the virtual sound source at a predetermined position by controlling the delay amounts of the delays 81 to 88. FIG. 3 is a diagram showing a virtual sound source localization method using a speaker array.

  As shown in the figure, the control unit 72 sets the virtual sound source V at a predetermined position (for example, the position of the speaker H1). The distance from the virtual sound source V to each speaker of the speaker array 2 is different, but the masker sound is output in order from the speaker closest to the virtual sound source V (speaker 21 in the figure), and the speaker 28 is sequentially from the speaker 22 over time. A plurality of third parties H3 that are located away from the counter by outputting the sound up to the counter have a speaker at the same distance from the virtual sound source position that is the focal point (the position of the speaker indicated by the dotted line in the figure) In addition, it is possible to perceive that masker sounds are emitted simultaneously from the positions of these virtual speakers. Therefore, the third party H3 virtually perceives that a masker sound was emitted from the position of the speaker H1.

  Here, the controller 72 dynamically changes the position of the virtual sound source V by dynamically changing the delay amount of the masker sound signal supplied to each speaker. FIG. 4 is a diagram for explaining the dynamic change of the virtual sound source position. The figure shows an example in which the position of the left virtual sound source V2 is changed toward the speaker H1 from the state where the position of the virtual sound source V1 is set on the right side toward the speaker H1 when viewed from the third party H3. .

  The control unit 72 changes the delay amounts of the delays 81 to 88 every elapse of a predetermined time (for example, every elapse of 1 second). For example, as shown in FIG. 4, when the virtual sound source V1 existing on the right side from the third party H3 toward the speaker H1 is set, the delay amount of the audio signal supplied to the right speaker 21 is reduced. The delay amount of the audio signal supplied to the left speaker 28 is set large, but when the virtual sound source V2 existing on the left side is set, the delay amount of the audio signal supplied to the speaker 21 is increased. The delay amount of the audio signal supplied to the speaker 28 is set small. Then, the third person H3 perceives that the output position of the masker sound has moved from the position of the virtual sound source V1 to the position of the virtual sound source V2. For this reason, even if the same masker sound is output, the sound source position changes, and the synthesized sound (sound heard simultaneously) with the speaker H1 changes and can be heard. For this reason, it is possible to prevent the ear habituation of a plurality of third parties H3 existing at positions away from the counter and appropriately suppress the cocktail party effect.

  Moreover, in the example of the figure, the control part 72 sets the movement area Z inside the circle centering on the center position S (it corresponds to the position of the microphone 1 in the example of the figure), and this movement area Z The position of the virtual sound source is changed at random. Of course, a virtual sound source may be set outside this moving area Z, but as the listener moves away from the position of the speaker H1, the listener can easily perceive the localization position of the masker sound and the speaker H1 as another position. Since the masking effect becomes low, it is desirable to change the position within a certain range from the position close to the speaker H1 to suppress the cocktail party effect.

  Furthermore, the control unit 72 dynamically changes the virtual sound source position so that the probability of setting the virtual sound source position at the center position S is the highest, and the probability is set with a lower probability as the distance from the center position S increases. Is also possible. For example, the virtual sound source position is dynamically changed with a probability according to a Gaussian distribution. In the example of FIG. 4, in the movement area Z, the virtual sound source position appears with a higher probability as the black position, and the virtual sound source position appears with a lower probability as the white position. Since the masking effect can be increased closer to the position of the speaker H1, the appearance probability of the virtual sound source is set higher near the position of the speaker H1, and the appearance probability is set lower as the distance from the speaker H1 increases.

  The center position S may be set in advance assuming the position of the microphone and the position of the speaker. However, an arbitrary position behind the speaker array (for example, about 0.5 m behind the center of the speaker array). The moving area Z may be set to an arbitrary value such as a radius of 1 m, or an operation unit (not shown) for operation by the user may be provided to accept manual input from the user. Good. Alternatively, it may be automatically set according to the width of the speaker array. For example, a straight line connecting the end speaker 21 and the speaker 28 of the speaker array is set, this straight line is the long side, and a right triangle or equilateral triangle connecting the speaker 21, the speaker 28, and the center position S is set. Then, the radius of the circle of the moving area Z is set to the distance between the speaker 21 (or speaker 28) and the center position S.

  Next, FIG. 5 is a flowchart showing the operation of the voice processing device 3. The voice processing device 3 starts this operation at the first activation (when the power is turned on), and thereafter performs this operation every predetermined time (for example, every one second). First, the voice processing device 3 stands by until a speaker voice is collected (s11). For example, when a sound of a predetermined level or higher that can be determined to be sound is picked up, it is determined that a speaker voice is picked up. When the speaker voice is not picked up and the conversation is not performed, the masker sound is unnecessary, so the masker sound is generated and the localization process is awaited. However, this process may be omitted, and a masker sound generation and localization process may always be performed.

  When the speaker voice is collected, the voice processing device 3 generates a masker sound by the masker sound generation unit 73 (s12). Note that it is desirable that the masker sound has an aspect in which the volume changes according to the level of the collected speaker voice. When the level of the collected speaker voice is low, the voice of the speaker reaches the third party H3 at a low level and it is difficult to grasp the content of the conversation, so the masker sound level can also be lowered. On the other hand, when the level of the collected speaker voice is high, the speaker voice reaches the third party H3 at a high level and it is easy to grasp the content of the conversation. In addition, the level of the masker sound is changed at the moment when the virtual sound source position dynamically changes, and the third person H3 may perceive the position of the virtual sound source to change little by little to reduce discomfort. .

  Then, the sound processing device 3 sets the delay amount by the control unit 72 so that the localization position of the masker sound changes randomly (s13). For example, as shown in FIG. 4, the virtual sound source position has a higher probability as it is closer to the center within a predetermined range (within the movement area Z) from the center position S (position closer to the speaker H1), and with a lower probability as the distance from the center increases. So that the delay amount of the audio signal supplied to each speaker is dynamically changed.

  As described above, the voice processing device 3 dynamically changes the virtual sound source position of the masker sound, so that the third party H3 can hear that the masker sound always moves, and the cocktail. The party effect can be appropriately suppressed.

  As shown in FIG. 6, the voice processing device 3 may include an echo canceller. FIG. 6 is a block diagram showing the configuration of the audio processing device 3 provided with an echo canceller. Components that are the same as those in FIG. 1 are given the same reference numerals, and descriptions thereof are omitted.

  The audio processing device 3 in this example includes an echo canceller 75 that inputs the audio signal output from the A / D converter 51. The echo canceller 75 receives the sound signal related to the masker sound from the masker sound generation unit 73, and filters the sound signal related to the masker sound using an adaptive filter that simulates the transfer characteristic of the sound transfer system from the speaker to the microphone. The echo component is reduced by subtracting the signal input from the A / D converter 51. Further, the echo canceller 75 may be provided as many as the number of speaker units in the speaker array. Since there are as many acoustic transmission systems (echo paths) from the speakers to the microphones as there are speakers, an adaptive filter that ideally estimates the echo path for each speaker is provided, and the audio signal supplied to each speaker is It is desirable to estimate and subtract echo components by filtering.

  Note that the voice processing device 3 can be realized by using hardware and software of an information processing device such as a general personal computer, instead of a device dedicated to the masking system shown in the present embodiment.

  In the present embodiment, an example is shown in which one microphone that collects the voice of the speaker H1 is provided, but a plurality of microphones may be provided. Moreover, the aspect which provides the microphone array which arranged the some microphone may be sufficient. In this case, the position of the speaker H1 can be detected by detecting the phase difference of the sound collected by each microphone of the microphone array, and the above-described center position S and moving region Z are detected by the detected speaker H1. (Or a position close to the speaker H1).

  Further, as the means for specifying the position, a technique based on information other than sound, such as using image recognition or a sensor, may be used.

H1 ... speaker H2 ... listener H3 ... third party 1 ... microphone 2 ... speaker array 3 ... speech processing device

Claims (4)

A masker sound generator for generating masker sounds;
A plurality of speakers for outputting the masker sound;
A localization control unit for controlling a localization position of the masker sound and supplying an audio signal related to the masker sound to the plurality of speakers;
With
The sound output device, wherein the localization control unit randomly changes the localization position of the masker sound within a predetermined range centered on a predetermined position . The localization control unit sets the localization position so that the center position is set to the localization position with the highest probability around the predetermined position, and the localization position is set with a lower probability as the distance from the center position increases. The audio output device according to claim 1, wherein the audio output device is dynamically changed. A microphone that picks up the voice of the speaker
The masking sound generating unit, an audio output device according to claim 1 or 2 to generate the masking sound based on the sound picked up by the microphone. Cancel the pseudo echo signal sound according to the masking sound is pseudo echo component reaching the microphone from the loudspeaker from the audio picked up by the microphone, claim 3 comprising an echo canceller to be supplied to the masking sound generating unit The audio output device according to 1.