JP2002116784A - Information signal processing device, information signal processing method, information signal recording and reproducing device and information signal recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To decide the attributes of input signals as to whether the signals are musical tone signals or not from a broadcasting program or the like. SOLUTION: This device consists of a frequency analyzing means for frequency analyzing the prescribed input information signals by each of prescribed sections, a peak frequency detecting means for successively detecting a prescribed plurality of the peak frequencies corresponding to the prescribed maximum value of the analysis processing values in a prescribed frequency range by each of the prescribed sections from the analysis signals of the frequency analyzing means, a signal deciding means for deciding whether the signals of the detection signal frequencies from the peak frequency detecting means in the prescribed section are within the prescribed frequency range inclusive of the detection frequencies even in another section of the front direction or rear direction continuous with this section or not, a signal detecting means for detecting the analysis signal level in the respective detection frequencies detected by the peak frequency detecting means or the signal level in the prescribed section of the input signals described above and a signal deciding means for deciding the attributes of the signals from the signals from the signal deciding means and the signal detecting means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an apparatus and a method for receiving a broadcast program and analyzing the audio signal of the broadcast program to determine the attribute of an input signal as to whether or not the signal is a tone signal.
In addition, when recording the broadcast program on a magnetic recording medium such as a semiconductor memory or a magnetic disk, an identification signal is generated according to the determination result, and reproduction control is performed at the time of reproduction using the identification signal to perform special reproduction such as digest reproduction. And a recording medium for performing the method.

[0002]

2. Description of the Related Art Recently, various data compression systems have been developed for video and audio devices for consumer use such as video and audio, and can be easily used. 2. Description of the Related Art As much information can be easily recorded on various recording media, there has been a demand for efficiently grasping recorded contents from these recording media. For example, there is a video reproducing apparatus that detects a silent section of audio and reproduces a video tape substantially in a short time without reproducing the section.

In such a device, the reproduction speed is at most twice the speed, and in order to grasp the outline of the recording in a shorter time, a part of the audio signal is detected in addition to the silent section, and that part is reproduced. A method of not doing this is also conceivable.
For example, in a news program of a broadcast program, a human voice section such as an announcer is important as information, and a music signal flowing at the beginning or in the middle of the program is considered to have a low priority as information of a news program.

When the characteristics of a human voice signal and a musical tone signal are compared, for example, the spectral characteristics of a human voice signal vary in a complicated manner depending on vowels, consonants, gender, age, etc. It is thought that the characteristics of the tone signal are easier to grasp because the tone signal is basically generated by the musical instrument except for complicated cases.

As a technique for detecting a music section from an audio signal, for example, Kenichi Minami et al .: "Video Indexing Using Sound Information and Its Application" (Transactions of the Institute of Electronics, Information and Communication Engineers) (D-
2 Vol.J81-D-2 No.3 pp.529-537 March 1998) and Kenichi Minami et al .: "Video browsing interface using sound information" (Technical Report of the Institute of Television Engineers of Japan (Vo
l.19 No.7 pp.1-6 1995).

[0006] It is described that these conventional musical sound detection techniques aim at detecting a musical sound signal in which the continuity of a spectral peak is remarkable, excluding rhythm-only music.

[0007]

However, if it is assumed that a recording / reproducing apparatus for recording / reproducing a broadcast program or a camera-integrated video camera is used for shooting, signals of various music genres are input, so that some musical sounds are input. The purpose of signal detection is not very practical, and a tone detection technique that can handle signals of a music genre as wide as possible is more effective. For this reason, in the above-mentioned conventional technology, there is a problem that it is sometimes difficult to detect a signal having a spectral continuity that is not remarkable when the musical signal is analyzed in a spectrum, depending on a music genre.

Accordingly, the present invention has a problem to be solved from the above-mentioned viewpoint to a musical sound detection technique which can cope with a range of music genres as wide as possible.

[0009]

In order to solve the above-mentioned problems, an information signal processing apparatus, an information signal processing method, and an information signal processing method according to the present invention are provided.
The information signal recording / reproducing apparatus and the information signal recording medium have the following configurations.

(1) Frequency analysis means for performing frequency analysis of a predetermined input information signal for each predetermined section, and a predetermined maximum value of an analysis processing value within a predetermined frequency range for each predetermined section from an analysis signal of the frequency analysis means. A peak frequency detecting means for sequentially detecting a plurality of peak frequencies to be detected, and a detection signal frequency from the peak frequency detecting means in the predetermined section also includes the detection frequency in another section in the forward or backward direction connected to the section. Signal determination means for determining whether a signal is detected within a predetermined frequency range, and a signal for detecting an analysis signal level at each detection frequency detected by the peak frequency detection means or a signal level in a predetermined section of the input signal Detection means, signal determination means for determining the attribute of the signal from the signal from the signal determination means and the signal from the signal detection means, Information signal processing apparatus, characterized in that was e. (2) The information signal processing device according to (1), wherein the predetermined information signal is an audio signal. (3) In the frequency analysis, after performing a predetermined averaging process or a predetermined thinning process on the information signal, FFT (fast Fourier transform process) or DCT (discrete cosine transform process)
Alternatively, the information signal processing device according to (1), wherein a frequency analysis similar to the above is performed. (4) The information signal processing device according to (1), wherein the signal detection means detects a signal-to-noise ratio from a detected signal level and a predetermined noise level within a predetermined frequency band.

(5) A predetermined input information signal is frequency-analyzed for each predetermined interval, and a peak frequency corresponding to a predetermined maximum value of an analysis value in a predetermined frequency range is sequentially determined for each predetermined interval from the analysis signal obtained from the frequency analysis. A plurality of predetermined detections are performed, and it is determined whether a signal is detected within a predetermined frequency range including the detection frequency in another section in the forward or backward direction in which the detection signal frequency continues from the peak frequency detection in the predetermined section. And an attribute of the input information signal is determined by the signal determination. (6) The information signal processing method according to (5), wherein the predetermined information signal is an audio signal. (7) The frequency analysis performs a predetermined averaging process or a predetermined thinning process on the information signal, and then performs an FFT (fast Fourier transform) process, a DCT (discrete cosine transform) process, or a similar frequency analysis. (5) The information signal processing method according to (5).

(8) Frequency analysis means for frequency-analyzing an input information signal from a predetermined input means for each predetermined section, and a predetermined analysis value of an analysis value in a predetermined frequency range for each predetermined section from the analysis signal from the frequency analysis means. A peak frequency detecting means for detecting a plurality of peak frequencies corresponding to the maximum value, and a detection signal frequency from the peak frequency detecting means in the predetermined section is also detected in another section in the forward or backward direction connected to the section. Signal determination means for determining whether a signal is detected within a predetermined frequency range including a frequency; signal detection means for detecting signal levels at a plurality of detection frequencies detected by the peak frequency detection means; Signal determining means for determining the attribute of the signal from the signal from the determining means and the signal detecting means; Identification signal generation means for generating a fixed identification signal; recording means for recording the signal from the identification signal generation means and the information signal on a predetermined recording medium; a predetermined information signal recorded on the recording medium; An information signal recording / reproducing apparatus, comprising: reproducing means for reproducing the identification signal of (i), and reproduction control means for controlling reproduction from the recording medium according to the reproduced identification signal. (9) The information signal recording / reproducing apparatus according to (8), wherein the predetermined information signal is an audio signal. (10) The frequency analysis is performed by performing a predetermined averaging process or a predetermined thinning process on the information signal, and then performing FFT (fast Fourier transform process) or DCT (discrete cosine transform process) or a frequency analysis similar thereto (8). The information signal recording device according to (1). (11) The information signal recording / reproducing apparatus according to (8), wherein the signal detecting means detects a signal-to-noise ratio from the detected signal level and a predetermined noise level within a predetermined frequency band.

(12) A predetermined information signal is subjected to frequency analysis for each predetermined section, and a peak frequency corresponding to a predetermined maximum value of an analysis value in a predetermined frequency range is sequentially determined for each predetermined section from the analysis signal obtained from the frequency analysis. A plurality of detections, and from the peak frequency detection in the predetermined section, determine whether a signal is detected within a predetermined frequency range including the detection frequency in another section in the forward or backward direction in which the detection signal frequency continues to the section. An information signal characterized by recording a predetermined identification signal for identifying the predetermined section in a predetermined area of a recording medium when the predetermined section of the information signal is determined to be a predetermined attribute signal by the signal determination. recoding media. (13) The information signal recording medium according to (12), wherein the information signal is an audio signal. (14) The frequency analysis is a frequency analysis similar to FFT (fast Fourier transform) or DCT (discrete cosine transform) after subjecting the information signal to a predetermined averaging process or a predetermined thinning process (12). )
An information signal recording device according to claim 1.

In such a configuration, a predetermined input information signal such as an audio signal is frequency-analyzed, for example, for each predetermined section which is an integral multiple of one frame section of the corresponding video signal. Then, a peak frequency corresponding to a predetermined maximum value of the analysis processing value in a predetermined frequency range for each predetermined section from the analysis signal is defined as a first peak frequency, a second peak frequency,-
A predetermined peak frequency is sequentially detected in the order of ---. Then, it is determined whether a signal is detected within a predetermined frequency range including the detection frequency in another section in which the detection signal from the peak frequency detection in the predetermined section is continuous with the section. In this way, the analytic signal level at each detected peak frequency or the signal level in a predetermined section of the input signal is detected, and if the S / N of the signal is not good or a small signal equal to or lower than the predetermined level is detected. The case is determined and the musical sound is detected efficiently.

Further, an identification signal for identifying a musical tone section is recorded on a predetermined recording medium together with a voice signal or the like in accordance with the above detection result, and this is detected at the time of reproduction, so that an effective digest reproducing system or musical tone section can be obtained. An editing system such as efficient editing can be realized.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, embodiments of an information signal processing device, an information signal processing method, an information signal recording / reproducing device and an information signal recording medium according to the present invention will be described with reference to the drawings.

First, an outline of the present invention will be described. An input audio signal is subjected to predetermined frequency analysis processing such as FFT (Fast Fourier Transform) or DCT (Discrete Cosine Transform) processing for each predetermined section, and a predetermined analysis is performed from the frequency analysis signal. A plurality of predetermined peak frequencies are sequentially detected in the above frequency range. Detecting whether the peak frequency is detected in another section connected to the predetermined section, detecting the signal-to-noise ratio at the peak frequency or the input signal level in the predetermined section, and responding to the detection result with the peak detection. It is determined whether the input information signal is a predetermined tone signal.

Further, a predetermined identification signal is generated in accordance with the result of the determination, and the identification signal is recorded together with the information signal on a predetermined recording medium such as a magnetic tape, a magneto-optical disk, a hard disk or a semiconductor memory, and reproduced to reproduce a musical tone. Reproduction control is performed on a signal detection section, and a predetermined digest reproduction is performed.

Next, specific embodiments of the present invention having the above outline will be described in the following order with reference to the drawings. (1) Operation principle of the present invention (2) Tone detection system block configuration example 1 of the present invention (3) Tone detection block configuration example 2 of the present invention (4) Recording / reproducing apparatus block configuration example of the present invention (5) Operation Flowchart example 1 (6) Operation flowchart example 2

(1) Operation Principle of the Present Invention Before describing the operation principle of the present invention, a conventional technique for detecting a tone signal will be described. FIG. 10 is a conceptual diagram showing a method for detecting a tone signal according to the prior art.
As shown in FIG. 10 (1), it is assumed that the audio signal is divided into predetermined sections and FFT (fast Fourier transform) processing is performed as shown in FIG. 10 (2) to detect a spectrum peak frequency.

FIG. 10 (3) shows a case where the voice signal of FIG. 10 (1) is assumed to be a musical tone signal. For easy understanding, the spectrum peak frequency in the time direction is obtained from the result of the FFT analysis of FIG. 10 (2). FIG. 3 is a conceptual diagram in which is plotted. Here, since a music signal whose spectrum peak can be clearly detected is assumed, the spectrum peak frequency f2 (Hz) continues between the sections A1 to A4 as indicated by the black circles in FIG. It can be easily determined that the signal is a tone signal.

Here, focusing on the second peak frequency from the result of the FFT analysis of FIG. 10B, it can be seen that the peak continues at f4 which is a harmonic of the first peak frequency. Similarly, if the second peak frequency is plotted in the time direction, it can be seen that the peak frequency continues as indicated by a white circle in FIG.

However, in the case of a tone signal in an audio signal from a broadcast program or the like, various combinations of instrument sounds are used.
The spectral characteristics are complicated, and there are those which can be clearly recognized as a tone signal by a person but whose peak frequency does not show continuity as shown in FIG.

Next, the operation principle of the present invention will be described with reference to FIG. FIG. 1 (2) shows the result of FFT analysis of each voice section of FIG. 1 (1). The black circle in FIG. 1 (3) plots the first spectrum peak frequency, and the white circle plots the second spectrum peak frequency. It was done.
From FIG. 1 (3), it can be seen that the first peak frequency shows continuity in the sections A2 to A3, but does not continue in all sections A1 to A4. In many cases, an actual musical tone signal is not a signal whose spectral peak clearly continues as described above.

However, looking at the characteristic of the peak frequency in the time direction, for example, as shown in FIG. 1 (2), the first peak frequency is f in the section A1 of FIG. 1 (1).
2 (Hz), f3 (Hz) in the A2 section, f3 (Hz) in the A3 section, and f4 (Hz) in the A4 section. Regarding the second peak frequency, f3 (H
z), f2 (Hz) in section A2, f4 in section A3
(Hz) and f3 (Hz) in the A4 section. When the first peak frequency and the second peak frequency are plotted in the time direction of FIG. 1 (3), the first peak frequency or the second peak frequency itself has no continuity over a long section from A1 to A4. For example, in the case of a tone signal, the spectrum that has been present at the first peak frequency has a higher probability of being present in other signal sections.

In the above description, the first peak frequency and the second peak frequency have been described for simplicity. However, other higher peak frequencies such as the third peak frequency and the fourth peak frequency may continue.

Here, a conceptual diagram of the spectrum by the actual FFT analysis is as shown in FIG. 2, and the first peak frequency is f
6 (Hz), but the second peak frequency is f18 (H
z) and not f7 (Hz). Similarly, the third peak frequency is f28 (Hz), not f5 (Hz). Also, when the value of the power spectrum obtained by the FFT analysis is small, it may be considered as noise.
In the case of FIG. 2, a spectrum peak below the threshold value Th is not considered as a peak. In the example of FIG. 2, the fourth peak frequency is f35 (Hz), but is not more than the predetermined threshold Th, so that the fourth peak frequency is not detected.

By the way, in the case of a music signal, such as a signal having a fast tempo or a signal in which various musical instruments are mixed like an orchestra, the characteristic of the spectrum in the time direction does not always appear remarkably as described above. There is also. Therefore,
To further improve the detection accuracy of the tone signal, detection of the rhythmicity of the tone signal in addition to the spectral characteristics is considered.

FIG. 3 is a conceptual diagram for detecting rhythmicity from an audio signal. FIG. 3A shows an example of an audio signal waveform, and FIG. 3B shows an example of an envelope-detected waveform of the audio signal. In the case of a rhythmic tone signal, the envelope detection waveform has periodicity, and when the autocorrelation function is calculated, a periodic waveform with improved noise is obtained as shown in FIG. This waveform is subjected to, for example, FFT analysis to detect a peak frequency and S / N, and a waveform having a good S / N can be determined to have rhythmic properties. It is also conceivable to determine whether the signal is a signal.

As described above, by detecting the rhythm characteristic in addition to the time characteristic in the higher-order spectral peak analysis and judging the tone signal characteristics from both the detection results, the detection corresponding to a wider music genre can be relatively performed. It can be realized with a simple configuration.

Here, the FFT analysis processing of the spectrum peak detection described above and the FF in the rhythm detection processing system are performed.
The T analysis processing may be performed by DCT (discrete cosine transform) processing, or similar processing may be performed by a frequency analysis system such as a signal frequency characteristic analysis filter in which a plurality of band-pass filters (BPFs) are combined.

(2) Example 1 of Tone Detection System Block Configuration According to the Present Invention FIG. 4 shows an example of a block configuration of a tone detection system according to the present invention. This musical tone detection system block is composed of a block for detecting the continuity of the n-th spectral peak and a block for detecting rhythmicity from the envelope waveform of the input audio signal.

[2.1] Description of Spectrum Peak Continuity Detection Block First, the spectrum peak continuity detection block will be described. The input voice signal is input to a predetermined voice section detection system 20 for processing for each predetermined voice section, that is, for each predetermined data sample, and data of a predetermined number of samples is extracted.

Here, in order to speed up the processing, a tone detection signal processing described below may be performed on the audio data that has been subjected to the signal processing by the predetermined data averaging processing corresponding to the sampling frequency. .

For example, the audio signal is converted to a sampling frequency of 4
When an A / D-converted signal is input at 8 KHz and the number of quantization bits is 16 bits, audio data for several seconds may be processed after averaging about 1/8.

When a person hears a certain sound signal, a momentary sound signal, for example, a sound of about 0.1 seconds in an extreme case, is instantaneously determined to be music or human conversation sound. This is difficult, and it is only after listening to an audio signal for a certain time period that it can be determined whether the audio is music or conversation.

Therefore, when performing signal processing, it is quite natural to make a judgment after processing data for seconds, and it is more natural to detect a predetermined number of audio data to be detected. It believed accuracy is good, because it takes an extra time until it is determined that increasing too much the number of data, the number of audio data samples to perform the process is set to the number appropriate to the prospect of a few seconds of data.

The voice data from the predetermined voice section detection system 20 is subjected to a predetermined FFT analysis process in an FFT processing system 21, and a frequency f (H
The power spectrum P (f), P (f) = (dreal × dreal + dimag × dimag) (2.1) is calculated from the real part data dreal and the imaginary part data dimag of the FFT analysis data in z).

The signal whose power spectrum has been calculated is the first signal
The spectrum frequencies from the next to the ninth order are detected and input to the respective peak frequency detection systems 401 to 409.

The S / N at the peak frequency is detected by each of the S / N detection systems 501-509.

The signals from the peak frequency detection systems 401 to 409 and the signals from the S / N detection systems 501 to 509 are input to the spectrum peak continuation length detection system 30, and the spectrum peaks as described in the above description of the operation principle are obtained. Is detected.

[2.2] Explanation of Rhythm Detection Block Next, the rhythm detection block of the input audio signal will be described. From the voice data from the predetermined voice section detection system 20, an envelope waveform is extracted by a predetermined envelope detection process in an envelope detection system 24, and a predetermined autocorrelation calculation is performed in a next autocorrelation function calculation system 25.

Here, considering that the spectrum of the correlated signal also has correlation, when performing the autocorrelation function operation, the envelope detection signal is used instead of the envelope detection signal.
As shown in another embodiment of the rhythmicity detection block (2), a method of using the signal after performing the FFT processing may be considered.

The signal subjected to the autocorrelation calculation is subjected to a predetermined FFT analysis processing in the next FFT processing system 26, and a predetermined power spectrum is detected in the power spectrum detection system 27. The peak frequency of the signal from the power spectrum detection system 27 is detected by the next peak frequency detection system 28,
Thereafter, the S / N detection system 29 detects a predetermined S / N.

Here, consider the difference between the speech sound waveform and the music (musical sound) sound waveform. FIG. 5 (1) is a conceptual diagram of a conversation voice waveform. When a normal person speaks a word, breathing is performed. Therefore, waveforms A, B, C, --- H,
There is a portion where the level of the audio waveform suddenly drops at a certain time like the portion I.

However, as shown in FIG. 5 (2), in the case of a music (musical tone) signal, generally, a musical instrument is performed, and a sudden level drop like a speech voice waveform rarely occurs. It is thought. Therefore, the signal level detection system 23
It is considered that the level of the audio signal is detected by using, and that this level reduction is used as a parameter for tone detection.

The signals from the power spectrum peak continuation length detection system 30, the S / N detection system 29, and the audio level detection system 23 are input to a general judgment system 31, and the respective detection results are used to comprehensively determine the input audio signal. Performs tone determination.

Here, each detection result may be weighted to make a comprehensive judgment. For example, the total judgment value is W, and the detection result of the spectrum peak continuation length is S.
p (0 ≦ Sp ≦ 1) and the rhythmic detection result is Rh (0 ≦ R
h ≦ 1), the sound level down detection result is Ld (0 ≦ Ld)
If ≦ 1), W = Sp + Rh + (1−Ld) (2.2) is defined as a determination parameter for convenience, W is 0 ≦ W ≦ 3, and as W becomes closer to 3, the sound signal becomes a musical tone. It can be determined that there is a high possibility that the signal is a signal.

Therefore, as a predetermined threshold value, Wth
Is set, and a voice section satisfying W> Wth (2.3) can be determined as a musical tone section.

Sp is set to Sp = 1 when the spectrum continuation length is longer than a predetermined section, and is set to Sp = 0 when continuity is not recognized.

Rh is considered in the same manner. When the rhythmicity is remarkably detected, Rh = 1 is set, and when the rhythmicity is not recognized, Rh = 0.

Ld is set to Ld = 0 when there is no level drop of the audio waveform over a predetermined section, and is set to Ld = 1 when the level is frequently detected.

Here, in order to perform weighting as described above, α (0 <α <1), β
(0 <β <1) and γ (0 <γ <1), W = α · Sp + β · Rh + γ · (1-Ld) (2.4) As W is closer to 3, the audio signal is a music signal. May be determined to be high, and an identification signal in the recording / reproducing apparatus described later may be generated according to the determination result.

Further, the level reduction of the audio signal can be easily detected, and it can be determined that the silent signal section is not a music signal without performing a series of predetermined signal processing in the present invention. Then, it is conceivable to multiply the whole by the parameter of the level down to obtain W = γ · (1−Ld) · (α · Sp + β · Rh) (2.5) In this case, the closer W is to 2, the higher the probability of being a musical tone.

(3) Example 2 of Tone Detection Block Configuration According to the Present Invention FIG. 6 shows another embodiment of the tone detection block. Considering that a recording / reproducing apparatus to which the present invention is applied uses an input image / audio signal as a tuner system to record / reproduce a broadcast program or the like, depending on the performance of the tuner and the reception state of the broadcast, a buzz sound is generated in the audio signal. And other noise.

In an extreme case, the buzz sound is a sound signal having continuity such as "boo" and its spectrum peak continuation length also continues, and it may be difficult to distinguish the buzz sound from the spectrum continuation characteristic of a music signal. There is.

The embodiment shown in FIG. 6 is an example of a block configuration in which a buzz sound possibly existing in a low frequency band is detected to reduce the possibility of erroneous detection.

The audio signal is input to the predetermined audio section detection system 20 as described above, and the audio data of a predetermined number of sampled data subjected to the averaging process is converted into the spectrum peak continuity detection system block systems 200 and 600, and the rhythm. It is input to the gender detection block system 300 and the audio level detection system 23, respectively.

Here, the spectrum peak continuity detection block systems 200 and 600 and the rhythmic detection block system 3
00 is the same as the block described with reference to FIG.

Spectral peak continuity detection block 60
The input audio data of 0 is data that has been subjected to the thinning process or the averaging process and has a lower frequency band in terms of frequency characteristics than the input audio data of the spectrum peak continuity detection block 200.

For example, as described above, assuming that the audio signal has a sampling frequency of 48 KHz and the number of quantization bits is 16, the input audio data of the spectral peak continuity detection block system 200 is data obtained by averaging about 1/8, It is assumed that the input speech data of the spectrum peak continuity detection block system 600 is data obtained by further thinning out the data input to the spectrum peak continuity detection block system 200 by about か or averaging the data.

That is, since the sampling frequency is assumed to be 48 KHz, the frequency band of the input audio data of the spectrum peak continuity detection block system 200 is about (48 KHz / 2) / 8 = 3 KHz.

The frequency band of the audio data input to the spectrum peak continuity detection system 600 is about 3 KHz / 4 = about 750 Hz because the frequency band is thinned out or averaged to about 1/4.

Here, the input audio data band of the spectrum peak continuity detection block system 200 is about 3 KHz, but when a general music signal is considered, its fundamental frequency falls within that frequency band. Therefore, it is considered that there is no problem even if the number of audio data is reduced by averaging to speed up the signal processing as described above.

The frequency band of the input voice data of the spectrum peak continuity detection block system 600 is 750 Hz.
It can be seen that it can be detected from the fact that the frequency spectrum of the buzz sound is in the low band.

Spectral peak continuity detection block system 2
The output signals from 00 and 600, the output signal from the rhythmic detection block 300, and the output signal from the audio level detection system 23 are input to the overall determination system 31, and it is determined whether the input audio signal is a tone signal. .

Here, as in the case of the block configuration of FIG. 4 described above, it is considered that the determination result is comprehensively performed by weighting the detection results from the respective detection systems.

As in the case of the above (2.4), the overall judgment result is W, the result of the spectrum peak continuity detection block system 200 is Sp1, the weighting factor is α1,
The detection result of the spectrum peak continuity detection block system 600 is Sp2, its weighting factor is α2, the detection result of the rhythmic detection block system 300 is Ld, its weighting factor is β, and the detection result of the audio level detection system 23 is L.
d, assuming that the weighting coefficient is γ, W = α1 · Sp1 + α2 (1-Sp2) + β · Rh + γ (1-Ld) (3.1) In this case, it is determined that the closer W is to 4, the more likely the audio signal in that section is a tone signal.

Or, in the same way as in (2.5), W = γ (1-Ld) (α1 · Sp1 + α2 (1-Sp2) + β · Rh) (3.2) In this case, it is determined that the closer the W is to 3, the more likely the audio signal in that section is a tone signal.

(4) Example of Block Configuration of Recording / Reproducing Apparatus According to the Present Invention FIG. 7 is an example of a block configuration diagram of an embodiment in which the present invention is applied to an information signal recording / reproducing apparatus. First, the recording system will be described.

The video signal is supplied to the video signal A / D conversion processing system 1
Then, the audio signal is input to the audio signal A / D conversion processing system 11 and is subjected to a predetermined A / D conversion process. After this video signal is subjected to a predetermined A / D conversion process, the video signal is input to the next video signal encoding processing system 2, where a predetermined band compression process such as MPEG is performed, and the video signal is input to the multiplexing processing system 3. The audio signal is input to an audio signal encoding processing system 10 and a musical sound detection processing system 16 after predetermined A / D conversion.

In the audio signal encoding processing system 10, a predetermined band compression processing such as MPEG audio is performed, and is input to the next multiplex processing system 3.

The tone detection processing system 16 performs predetermined tone detection determination processing as described in the above tone detection block. The output of the tone detection processing system 16 is input to the system controller system 17, and the identification signal is controlled by the system controller system 17. A predetermined identification signal is generated in the generation system 12 and input to the multiplex processing system 3.

In the multiplexed signal processing system 3, predetermined signals inputted such as video data, audio data, identification data and the like are processed by predetermined multiplexed signal processing, and the next recording processing system 4 adds error correction codes and interleaves. After a predetermined signal processing such as is performed, it is recorded on a predetermined recording medium 5.

Next, the reproducing system will be described. The signal reproduced from the recording medium 5 is subjected to predetermined error correction processing and deinterleaving processing in a reproduction signal processing system 6, and the next reproduction data separation processing system 7 converts the video data, audio data, and identification data into predetermined signals. Each is separated by the processing.

The separated video data signal is input to a video data decoding processing system 8 and subjected to a predetermined decoding process, and then subjected to a predetermined D / A conversion process in a video signal D / A conversion processing system 9. Output video signal.

The audio data signal similarly separated is
After being input to the audio data decoding processing system 14 and subjected to predetermined decoding processing, the audio signal D / A conversion processing system 15 performs predetermined D / A conversion processing and outputs an audio signal.

The identification data signal separated by the reproduction data separation processing system 7 is subjected to predetermined detection processing by the identification signal detection system 18, and the output signal is output to the system controller system 17.
To enter.

The system controller system 17 controls the reproduction operation of the recording medium 5 via the recording medium control system 13 in accordance with the detected identification signal, and skips the tone signal detection section or controls only the tone detection section. A predetermined special reproduction operation such as reproduction is performed.

(5) Example 1 of Operation Flowchart FIG. 8 is an operation flowchart of the present invention, and the operation will be described with reference to this flowchart.

Starting from S0, the initial number of loops m = 0 is set in S1, and an audio signal is input in S2.

In S3, data of a predetermined section or a predetermined number of sample data of the input audio signal is detected, and in S4, an initial value setting n = 0 of the number of peak detection loops is set.

In step S5, a predetermined FFT analysis process is performed on the audio data detected in each predetermined section. In step S6, a power spectrum is calculated. In step S7, an n-th peak frequency is detected.

At S8, the detected n-th peak frequency is stored in a predetermined memory, and at S9, n is incremented by 1 to detect the next peak.

In S10, it is determined whether a predetermined number of n-order peak frequencies have been detected. If the predetermined peak frequency has not been detected yet, the process returns to S7 to detect the peak frequency.

If it is determined in S10 that all of the predetermined peak frequencies have been detected, the process proceeds to S11, where the number of times of the signal processing in the predetermined section is incremented by one, and in S12, it is determined whether the processing for the predetermined time ma has been completed. You.

If the processing for the predetermined time has not yet been completed in S12, the flow returns to S3, and the same peak frequency detection in the next predetermined section is performed.

If it is determined in S12 that the detection process for the predetermined time has been completed, the flow shifts to S13, where the initial setting n = 0 of the number of times of the spectrum peak continuation length routine is set.

In S14, the continuation length initial value kn of the n-th peak is obtained.
= 0, the spectrum peak continuation length over the ma section is detected in S15, and if the predetermined continuity is recognized in S16, the continuation length counter is incremented by 1 in S17, and if not, the process proceeds to the next S18. I do.

In S18, the number of times of the n-th peak continuity detection is set to 1
It is determined in the next S19 whether the detection of the continuation length of the predetermined peak has been completed.
Returning to step 5, the continuation length is detected.

If it is determined in S19 that the continuation length detection has been completed, the S / N at the peak frequency detected in S20 is detected, and the S / N level detected in S21 is determined.

Here, the S / N detection is performed after the peak detection and the continuation length detection, but may be performed when the power spectrum is detected in S6.

If the S / N is equal to or more than the predetermined level in S21, the continuation length of the spectrum peak is determined in S22 in S22. 1>
If it is determined in S23 that this is not the case, <0 in S25
Set>.

If it is determined in S21 that the S / N is not higher than the predetermined level, <0> is set in S25.

As described above, in addition to performing the tone determination using only the spectrum continuation length detection method according to the present invention, it is conceivable to detect rhythmicity as another determination parameter.
This will be described below.

(6) Example 2 of Operation Flowchart FIG. 9 is an example of an operation flowchart according to the present invention, which is different from the above-described operation flowchart. In this operation flowchart, in addition to the spectrum continuity determination, rhythmicity is added to the determination parameters.

Starting from P0, the process proceeds to the spectrum continuity detection routine of P1. Note that this spectrum continuity detection routine is the same as that of the operation flowchart 1 described above, and is therefore omitted here.

After detecting the spectrum continuity, the audio signal in the predetermined section is envelope-detected at P2, and the autocorrelation function is calculated at P3.

The signal for which the autocorrelation function is calculated in P3 is P
The FFT analysis processing is performed in step 4, the power spectrum is calculated in step P5, and the S / N of the first peak frequency is detected in step P6.

At P7, the level of the detected S / N is determined, and when it is determined that the level is equal to or higher than the predetermined level, the determination R at P8 that rhythm is recognized is set to <1>.

If the level is not equal to or higher than the predetermined level in P7, it is considered that rhythmicity is not recognized so much, and the judgment R is set to <0.
>

After the rhythmicity is determined as described above, P
The process proceeds to the musical tone comprehensive determination process of FIG. 10, and it is determined whether the input audio signal is a music signal based on the results of the spectrum continuation length determination and the rhythmicity determination described above.

In P11, in the above determination of the spectrum continuation length, in the case of the judgment <1> that is recognized as a tone signal, the process proceeds to P12, and in the case of the judgment <1> in which rhythmicity is also recognized, it is determined that the signal is most likely to be a music signal. Determination G is set to <11>.

If the spectrum continuation length is recognized but the rhythmicity is not recognized in P12, <10> is set as the overall judgment G to be the next music signal.

If the spectrum continuity is determined to be <0> in P11, the process proceeds to P14, and if only rhythmicity is recognized, the priority is low in music determination in P16, but the possibility of a music signal is high. <01> is set as the overall judgment G that may exist.

At P14, no spectral continuity was observed.
If the rhythmicity is not recognized, the general judgment G is set to <00> indicating that the signal is not a music signal.

As described above, in the operation flowchart here, four possibilities are set as the judgment of the music signal likeness of the audio signal from the judgment of the spectrum continuation length and the judgment of the rhythmicity. Alternatively, the determination result may be music only when the spectrum continuation length determination and the rhythmicity determination are recognized.

When the four kinds of determinations are performed as described above, for example, when an audio signal and an identification signal are recorded on a recording medium and the identification signal is reproduced at the time of reproduction to perform special reproduction, the special reproduction time is not required. Indicates that the identification signal is 4 depending on the playback speed.
It is also conceivable that the reproduction operation is controlled in accordance with the four types of identification signals.

For example, when a special reproduction operation is to be performed in the shortest time in consideration of an operation of skipping reproduction of a music section, an identification signal corresponding to <11>, <10>, and <01> is determined in the general judgment G. May be skipped in all the sections in which reproduction is detected.

If it is desired to perform special reproduction in the next short time, the overall judgment G may skip the identification signal section corresponding to <11> or <10>, and the section considered to be the most music signal section. When it is desired to perform only the skip operation, the skip operation may be performed only in the identification signal section corresponding to the comprehensive judgment G according to <11>.

For example, in the case of recording and reproducing a music / song program in a broadcast program, etc.,
May be reproduced according to the special reproduction time.

[0112]

According to the present invention, a musical tone section can be easily detected from a broadcast program or the like, whereby, for example, when a music program or a song program is recorded and reproduced, an index is automatically entered into the music section. By playing back that part in a digest for playback, music and song programs can be enjoyed efficiently and effectively.

Further, it is possible to efficiently detect a musical tone even in a musical tone signal in which rhythmicity is not remarkable in a music, and thereby a musical tone detecting system independent of a music genre can be realized with a relatively simple configuration.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing the operation principle of information signal processing according to the present invention.

FIG. 2 is a conceptual diagram showing an example of a power spectrum by the FET analysis.

FIG. 3 is a conceptual diagram showing an example of a rhythm detection method for the musical sound signal.

FIG. 4 is a block diagram showing the configuration of the musical sound detection system.

FIG. 5 is a conceptual diagram showing a difference between the conversation voice waveform and a music (musical sound) waveform.

FIG. 6 is a conceptual diagram showing a configuration example of another embodiment of the musical tone detection system block.

FIG. 7 is a block diagram showing an example of the recording / reproducing apparatus of the present invention.

FIG. 8 is a flowchart showing an example of the operation of the present invention.

FIG. 9 is a flowchart showing an example of the operation of the present invention.

FIG. 10 is a conceptual diagram showing a tone detection in the related art.

[Explanation of symbols]

1: video signal A / D conversion system, 2: video signal encoding system, 3: multiplex processing system, 4: recording signal processing system, 5: recording medium, 6: reproduction signal processing system, 7: reproduction data separation processing System, 8: video data encoding system, 9: video signal D
/ A conversion system, 10: audio data encoding system, 1
1: audio signal A / D conversion system, 12: identification signal generation system, 1
3: recording medium control system, 14: audio data decode processing system, 15: audio data D / A processing system, 16: musical sound detection system, 17: system controller system, 18: identification signal detection system, 20: predetermined audio section detection System, 21: FFT (Fast Fourier Transform) processing system, 22: Power spectrum detection system, 23: Signal level detection system, 24: Envelope detection system, 25: Autocorrelation function operation system, 26: FFT (Fast Fourier Transform) processing System, 27: power spectrum detection system,
28: peak frequency detection system, 29: S / N detection system, 3
0: spectrum peak duration detection system, 31: comprehensive judgment system, 32: FFT (fast Fourier transform) processing system, 20
0: Spectrum peak continuity detection block system, 300:
Rhythm detection block system, 401: first spectrum peak frequency detection system, 402: second spectrum peak frequency detection system, 403: third spectrum peak frequency detection system, 409: ninth spectrum peak frequency detection system,
501: S / N detection system, 502: S / N detection system, 50
3: S / N detection system, 509: S / N detection system, 600: Spectral peak continuity detection block system

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (reference) // G10L 101: 027 F term (reference) 5C053 GB11 GB22 JA12 JA21 LA07 LA14 5D015 AA06 CC01 CC03 DD03 5D044 AB05 AB07 BC01 CC05 DE19 DE40 DE49 EF05 FG18 FG23 GK12 HL11

Claims (14)

[Claims] A frequency analysis means for frequency-analyzing a predetermined input information signal for each predetermined section, and a predetermined maximum value of an analysis processing value in a predetermined frequency range for each predetermined section from an analysis signal of the frequency analysis means. A peak frequency detecting means for sequentially detecting a plurality of peak frequencies, a predetermined signal including the detected frequency in another section in a forward or backward direction in which the detection signal frequency from the peak frequency detecting means in the predetermined section is continuous with the section. Signal determination means for determining whether a signal is detected within a frequency range; and signal detection for detecting an analysis signal level at each detection frequency detected by the peak frequency detection means or a signal level in a predetermined section of the input signal. Means, and signal determination means for determining an attribute of the signal from a signal from the signal determination means and a signal from the signal detection means. An information signal processing device, characterized in that: 2. The information signal processing apparatus according to claim 1, wherein said predetermined information signal is an audio signal. 3. The method of claim 2, wherein the frequency analysis is performed by subjecting the information signal to a predetermined averaging process or a predetermined thinning process, and then performing an FFT.
2. The information signal processing apparatus according to claim 1, wherein (fast Fourier transform processing) or DCT (discrete cosine transform processing) or a frequency analysis similar thereto is performed. 4. The information signal processing apparatus according to claim 1, wherein said signal detecting means detects a signal-to-noise ratio from a detected signal level and a predetermined noise level within a predetermined frequency band. 5. A predetermined input information signal is frequency-analyzed for each predetermined section, and a peak frequency corresponding to a predetermined maximum value of an analysis value in a predetermined frequency range is predetermined for each predetermined section from the analysis signal from the frequency analysis. A plurality of detections are performed, and it is determined whether a signal is detected within a predetermined frequency range including the detection frequency in another section in the forward or backward direction in which the detection signal frequency is continuous with the section from the peak frequency detection in the predetermined section. An information signal processing method, wherein an attribute of the input information signal is determined by the signal determination. 6. The information signal processing method according to claim 5, wherein said predetermined information signal is an audio signal. 7. The method according to claim 1, wherein the frequency analysis is performed by subjecting the information signal to a predetermined averaging process or a predetermined thinning process, and then performing an FFT.
6. The information signal processing method according to claim 5, wherein (fast Fourier transform) processing, DCT (discrete cosine transform) processing, or frequency analysis similar thereto is performed. 8. A frequency analysis means for frequency-analyzing an input information signal from a predetermined input means for each predetermined section, and a predetermined maximum value of an analysis value in a predetermined frequency range for each predetermined section from the analysis signal from said frequency analysis means. A peak frequency detecting means for detecting a predetermined plurality of peak frequencies corresponding to the value, and a detection signal frequency from the peak frequency detecting means in the above-mentioned predetermined section is also detected in another section in the forward or backward direction connected to the section. Signal determination means for determining whether a signal is detected within a predetermined frequency range including: signal detection means for detecting signal levels at a plurality of detection frequencies detected by the peak frequency detection means; Means and signal determination means for determining the attribute of the signal from the signal from the signal detection means; Identification signal generating means for generating an identification signal, a recording means for recording the signal from the identification signal generating means and the information signal on a predetermined recording medium, a predetermined information signal recorded on the recording medium and a predetermined An information signal recording / reproducing apparatus, comprising: reproducing means for reproducing an identification signal; and reproduction control means for controlling reproduction from the recording medium according to the reproduced identification signal. 9. An information signal recording / reproducing apparatus according to claim 8, wherein said predetermined information signal is an audio signal. 10. The frequency analysis according to claim 1, wherein said information signal is subjected to a predetermined averaging process or a predetermined decimation process and then subjected to an FFT.
9. The information signal recording apparatus according to claim 8, wherein (fast Fourier transform processing) or DCT (discrete cosine transform processing) or a frequency analysis similar thereto is performed. 11. An information signal recording / reproducing apparatus according to claim 8, wherein said signal detecting means detects a signal-to-noise ratio from a detected signal level and a predetermined noise level within a predetermined frequency band. 12. A predetermined information signal is subjected to frequency analysis for each predetermined section, and a peak frequency corresponding to a predetermined maximum value of an analysis value in a predetermined frequency range is sequentially determined for each predetermined section from the analysis signal from the frequency analysis. A plurality of detections, it is determined whether a signal is detected within a predetermined frequency range including the detection frequency also in another section in the forward or backward direction where the detection signal frequency is continuous from the peak frequency detection in the predetermined section to the section, When a predetermined section of the information signal is determined to be a predetermined attribute signal by the signal determination, a predetermined identification signal for identifying the predetermined section is recorded in a predetermined area of a recording medium. Medium. 13. The information signal recording medium according to claim 12, wherein said information signal is an audio signal. 14. The method according to claim 1, wherein the frequency analysis is performed by subjecting the information signal to a predetermined averaging process or a predetermined thinning process and then performing an FFT.
13. The information signal recording apparatus according to claim 12, wherein (fast Fourier transform processing), DCT (discrete cosine transform processing), or frequency analysis similar thereto is performed.