JP4841276B2 - Music signal storage device and music signal storage program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a music signal accumulating device which can extract only music signals from an input audio signal and automatically select and store only music signals matching user's preference among the extracted music signals. <P>SOLUTION: The music signal accumulating device which extracts only music signals meeting conditions which a user desires from the input audio signal and stores them in a storage device includes a music signal extracting means of extracting only music signals by analyzing the input audio signal, a music feature analyzing means of analyzing features of music based upon a music signal extracted by the music signal extracting means, a condition input means for inputting conditions for accumulating music signals extracted by the music signal extracting means in the storage device, and a determination means of determining whether the music signal extracted by the music signal extracting means is accumulated in the storage device based upon the analysis result of the music feature analyzing means and the conditions input to the condition input means. <P>COPYRIGHT: (C)2008,JPO&amp;INPIT

Description

  The present invention relates to a music signal storage device and a music signal storage program that extract only a music signal portion from a music program such as a radio broadcast or a TV broadcast and store only a music signal that matches a user's desire.

  In Japanese Patent No. 3594427, when listening to a radio broadcast, if a content other than the broadcast content desired by the user is being broadcast, the channel selection channel is automatically switched, or by another media playback device. A control device for a radio broadcast receiver capable of outputting a reproduced audio signal is disclosed.

  Japanese Patent Application Publication No. 2005-518560 discloses a system that extracts a music signal, stores the music data in a temporary memory, and stores the music signal stored in the temporary memory according to a user's storage command in the memory. .

In the invention described in Japanese Patent No. 3594427, it is not possible to store only music that suits the user's preference. In Japanese translations of PCT publication No. 2005-518560, it is possible to extract music signals and store only the music that suits the user's preference, but the user determines whether or not to store the extracted music signals in the memory. Therefore, it is not possible to automatically select and store only music that suits the user's preference.
Japanese Patent No. 3594427 JP 2005-518560 A

  The present invention extracts a music signal only from an input audio signal, and can automatically select and store only a music signal suitable for a user's preference among the extracted music signals and a music signal storage. The purpose is to provide a program.

According to a first aspect of the present invention, there is provided a music signal storage device that extracts only a music signal suitable for a condition corresponding to a user's favorite music from an input voice signal and stores the extracted music signal in a storage device, and analyzes the input voice signal. A music signal extracting means for extracting only the music signal, a music feature analyzing means for analyzing the characteristics of the music based on the music signal extracted by the music signal extracting means, and a condition corresponding to the user's favorite music Whether or not to store the music signal extracted by the music signal extracting means in the storage device based on the analysis result of the condition input means and the condition input to the condition input means Determination means for determining whether the music feature analysis result of the music feature analysis means matches the condition input to the condition input means. A first means for outputting, and a degree determined by the first means to determine whether or not to store the music signal extracted by the music signal extracting means in the storage device by comparing the threshold value with a threshold value. 2 means, and when the degree calculated by the first means is lower than the threshold, the second means must store the music signal extracted by the music signal extraction means in the storage device. Whether the data size of the music signal extracted by the music signal extraction means is smaller than the remaining capacity of the storage device when the degree calculated by the third means and the first means is higher than the threshold value A fourth means for determining the data size of the music signal extracted by the music signal extraction means by the fourth means is smaller than the remaining capacity of the storage device; The music signal extracted by the music signal extraction means is stored in the storage device, and the data size of the music signal extracted by the music signal extraction means by the fourth means is determined. When it is determined that the remaining capacity of the storage device is greater than or equal to the remaining capacity of the storage device, the music signal stored in the storage device is stored until the size of the music signal extracted by the music signal extraction unit becomes smaller than the remaining capacity of the storage device. Are determined to store the music signal extracted by the music signal extracting means in the storage device after the music feature analysis results are deleted in descending order of the degree that the music feature analysis result matches the condition input to the condition input means. The sixth means is provided.

According to a second aspect of the present invention, there is provided a music signal storage device that extracts only a music signal that satisfies a condition corresponding to a user's favorite music from the input voice signal and stores the extracted music signal in a storage device, and analyzes the input voice signal. A music signal extracting means for extracting only the music signal, a music feature analyzing means for analyzing the characteristics of the music based on the music signal extracted by the music signal extracting means, and a condition corresponding to the user's favorite music Whether or not to store the music signal extracted by the music signal extracting means in the storage device based on the analysis result of the condition input means and the condition input to the condition input means Determination means for determining whether the music feature analysis result of the music feature analysis means matches the condition input to the condition input means. A first means for outputting, and a degree determined by the first means to determine whether or not to store the music signal extracted by the music signal extracting means in the storage device by comparing the threshold value with a threshold value. 2 means, and when the degree calculated by the first means is lower than the threshold, the second means must store the music signal extracted by the music signal extraction means in the storage device. When the degree calculated by the third means for determining and the first means is higher than the threshold value, the characteristic amount analysis result of the music signal extracted by the music signal extracting means and the storage device are accumulated. Based on the feature analysis result of each music signal, a music signal that can be regarded as having the same feature analysis result as the music signal extracted by the music signal extracting means is already stored in the storage device. Fourth means for determining whether or not stored, by said fourth means, said music signal extracting music signal and feature analysis result extracted by means already been stored in the music signal that can be regarded as identical in the storage device If it is determined that the music signal extracted by the music signal extraction means is not stored in the storage device, the fifth means and the fourth means determine that the music signal extracted by the music signal extraction means is not stored in the storage device. If it is determined that a music signal whose characteristic analysis result is the same as that of the signal is not stored in the storage device, the data size of the music signal extracted by the music signal extraction means is the remaining capacity of the storage device. The sixth means for determining whether or not the music signal is smaller than the data size of the music signal extracted by the music signal extraction means by the sixth means. Is determined by the music signal extraction means by the seventh means and the sixth means for determining that the music signal extracted by the music signal extraction means is stored in the storage device. When it is determined that the data size of the music signal is greater than or equal to the remaining capacity of the storage device, the storage device until the size of the music signal extracted by the music signal extraction means becomes smaller than the remaining capacity of the storage device. The music signals stored in the music signal extraction means are deleted in order from the lowest in the degree that the music feature analysis result matches the condition input to the condition input means. Eighth means for determining to accumulate in the storage device is provided.

The invention according to claim 3 is a music signal storage program for extracting only a music signal that meets a condition corresponding to a user's favorite music from an input voice signal and storing the extracted music signal in a storage device. A music signal extracting step for analyzing and extracting only the music signal, a music feature analyzing step for analyzing the characteristics of the music based on the music signal extracted by the music signal extracting step, and a condition corresponding to the user's favorite music The music signal extracted by the music signal extraction step is stored in the storage device based on the condition input step for inputting, and the analysis result of the music feature analysis step and the condition input in the condition input step A music signal storage program for causing a computer to execute a determination step for determining whether or not the determination step includes: A first step of calculating a degree that the music feature analysis result of the music writing feature analysis step matches the condition input in the condition input step, and comparing the degree calculated in the first step with a threshold value The second step of determining whether or not to store the music signal extracted in the music signal extraction step in the storage device, the degree calculated by the first step is the second step. If the threshold is lower than the threshold, the third step of determining that the music signal extracted by the music signal extraction step is not stored in the storage device, and the degree calculated by the first step is higher than the threshold Whether the data size of the music signal extracted by the music signal extraction step is smaller than the remaining capacity of the storage device When it is determined by the fourth step and the fourth step that the data size of the music signal extracted by the music signal extraction step is smaller than the remaining capacity of the storage device, the music signal is extracted by the music signal extraction step According to the fifth step and the fourth step in which it is determined that the music signal is stored in the storage device, it is determined that the data size of the music signal extracted by the music signal extraction step is greater than or equal to the remaining capacity of the storage device. Sometimes, the music signal stored in the storage device is input to the condition input step, and the music feature analysis result is input until the size of the music signal extracted in the music signal extraction step becomes smaller than the remaining capacity of the storage device. Are deleted in order from the one with the lowest degree of conformity to the set condition, and then the music signal extraction is performed. A sixth step of determining that the music signal extracted in the output step is stored in the storage device is provided.

According to a fourth aspect of the present invention, there is provided a music signal storage program for extracting only a music signal suitable for a condition corresponding to a user's favorite music from an input voice signal and storing the extracted music signal in a storage device. A music signal extracting step for analyzing and extracting only the music signal, a music feature analyzing step for analyzing the characteristics of the music based on the music signal extracted by the music signal extracting step, and a condition corresponding to the user's favorite music The music signal extracted by the music signal extraction step is stored in the storage device based on the condition input step for inputting, and the analysis result of the music feature analysis step and the condition input in the condition input step A music signal storage program for causing a computer to execute a determination step for determining whether or not the determination step includes: A first step of calculating a degree that the music feature analysis result of the music writing feature analysis step matches the condition input in the condition input step, and comparing the degree calculated in the first step with a threshold value The second step of determining whether or not to store the music signal extracted in the music signal extraction step in the storage device, the degree calculated by the first step is the second step. If the threshold is lower than the threshold, the third step of determining that the music signal extracted by the music signal extraction step is not stored in the storage device, and the degree calculated by the first step is higher than the threshold , A characteristic amount analysis result of the music signal extracted by the music signal extraction step, and each sound stored in the storage device Based on the characteristics analysis results of the signal, the fourth step of determining whether the music signal the music signal extracting music signal and feature analysis result extracted by the step can be regarded as the same has already been accumulated in said storage device If it is determined by the fourth step that the music signal extracted in the music signal extraction step and the music signal whose feature analysis result is considered to be the same are already stored in the storage device, the music signal A music signal that can be regarded as having the same characteristic analysis result as the music signal extracted by the music signal extraction step according to the fifth step and the fourth step is determined not to store the music signal extracted by the extraction step in the storage device. If it is determined that it is not stored in the storage device, it is extracted by the music signal extraction step. Sixth determining the data size of the music signal whether less than the remaining capacity of the storage device, the first through sixth step, the remaining data size the storage device of a music signal extracted by the music signal extraction step When it is determined that the capacity is smaller than the capacity, the music extracted by the music signal extraction step is determined by the seventh step and the sixth step which are determined to store the music signal extracted by the music signal extraction step in the storage device. When it is determined that the data size of the signal is equal to or larger than the remaining capacity of the storage device, the music signal is accumulated in the storage device until the size of the music signal extracted by the music signal extraction step becomes smaller than the remaining capacity of the storage device. To the condition that the music feature analysis result is input to the condition input step. After deleting less from those of degree to which match the sequence, characterized in that it comprises an eighth step of determining a storing music signal extracted by said music signal extracting step to said storage device.

  According to the present invention, it is possible to extract only a music signal from an input audio signal and automatically select and store only the music signal that matches the user's preference among the extracted music signals.

  Embodiments of the present invention will be described below with reference to the drawings.

[1] Description of Overall Configuration of Music Signal Storage Device FIG. 1 shows the overall configuration of a music signal storage device.

  The input signal input to the music signal storage device is an audio signal (PCM signal) of radio broadcast or TV broadcast.

  The input signal is temporarily stored in the input signal storage memory 1. The music signal extraction unit 2 specifies the music signal part based on the input signal stored in the input signal storage memory 1, and specifies the start address and the final address of the music signal part.

  The music signal feature analysis unit 3 analyzes the music signal portion specified by the music signal extraction unit 2 and calculates a music sensitivity space coordinate value corresponding to the music of the music signal portion.

  The condition input unit 4 calculates music sensitivity space coordinate values corresponding to the user's favorite music. The desired music selection unit 5 is based on the spatial distance between the music sensitivity space coordinate value calculated by the music signal feature analysis unit 3 and the music sensitivity space coordinate value calculated by the condition input unit 4. It is determined whether or not the music of the music signal part specified by the user corresponds to the user's favorite music, and if it is determined that the music corresponds to the user's favorite music, the music signal part is stored in the storage memory 6. .

[2] Description of Music Signal Extractor 2 FIG. 2 shows the configuration of the music signal extractor 2.

  The music signal extraction unit 2 extracts and outputs only the music signal from the input signal. Basically, whether or not there is a sense of rhythm in the voice is determined by the autocorrelation evaluation value for evaluating the autocorrelation of the voiceband signal, and whether or not the voice rhythm matches the accompaniment rhythm. Judgment is made based on a cross-correlation evaluation value for evaluating the correlation between the signal and the out-of-speech signal, and music is determined when each evaluation value is high.

  The input signal stored in the input signal storage memory 1 (see FIG. 1) is sent to the voice band pass filter 12 and to the voice band cut filter 13. As the audio band pass filter 12, a band pass filter that passes a signal in a band of 300 Hz to 3 kHz is used. The audio band cutoff filter 13 includes a low-pass filter that passes a signal of 300 Hz or less and a band-pass filter that passes a signal of a band of 3 kHz to 10 kHz, and a signal of 300 Hz or less and 3 kHz to 10 kHz of the input signal. Pass the band signal. The reason why the signal to be passed through the audio band cut-off filter 13 is limited to 10 kHz or less is that the sampling frequency band varies depending on the input content and the upper limit frequency differs, so that the upper limit thereof is unified.

The signal (voice band signal) X _{n that} has passed through the voice band pass filter 12 is sent to the autocorrelation evaluation value calculation unit 14 and to the cross correlation evaluation value calculation unit 15. The signal (voice out-of-band signal) Y _{n that} has passed through the voice band cut-off filter 13 is sent to the cross-correlation evaluation value calculation unit 15.

As shown in FIG. 3, the autocorrelation evaluation value calculation unit 14 is based on a voice band signal X _n that has passed through the voice band pass filter 13, and has an interval (evaluation value) consisting of a predetermined number of samples (512 * 1024 in this example) For each calculation interval D), an autocorrelation evaluation value (total sum of autocorrelation of power values) for evaluating the autocorrelation of the voice band signal is calculated. The evaluation value calculation section D corresponds to about 12 seconds if the sampling frequency of the input signal is 44.1 kHz. When calculating an autocorrelation evaluation value for each evaluation value calculation section D, a sample in the section D and a predetermined number of samples (256 * 1024 in this example) immediately before the section are used.

A method for calculating an autocorrelation evaluation value in an arbitrary evaluation value calculation section will be described. Each sample used for calculating an autocorrelation evaluation value in a certain evaluation value calculation section is represented by X _n (n = −M, − (M−1),..., −1, 0, 1, 2,..., N). It will be expressed as X ₀ represents the head sample of the evaluation value calculation section.

First, an autocorrelation evaluation value HS_AV _j (j = 0 to 255) is calculated based on the following equation (1).

The autocorrelation evaluation value HS_AV _j is a large value if the audio band signal X _n has periodicity. That is, if the voice has a rhythmic sense, the autocorrelation evaluation value HS_AV _j is a large value.

Next, the range of j> 10, and HS_AVmax values HS_AV _j is maximized. Let HS_AVmax be the autocorrelation evaluation value for the section L. The reason why j> 10 is set is that HS_AV _j may increase in the range of j ≦ 10 even if there is no periodicity. The autocorrelation evaluation value HS_AVmax calculated by the autocorrelation evaluation value calculation unit 14 is sent to the music determination evaluation value calculation unit 16.

The cross-correlation evaluation value calculation unit 15 determines a predetermined number of samples (in this example, 512 *) based on the voice band signal X _n that has passed through the voice band pass filter 12 and the voice out-of-band signal Y _n that has passed through the voice band cutoff filter 13. For each evaluation value calculation section consisting of 1024), a cross-correlation evaluation value (total sum of cross correlations of power values) for evaluating the correlation between the audio band signal and the out-of-speech signal is calculated. Specifically, the cross-correlation evaluation value HM_AV is calculated based on the following equation (2).

The cross-correlation evaluation value HM_AV is a large value if there is a correlation between the voice band signal X _n and the voice band out-of-band signal Y _n . That is, if the rhythm of the sound matches the rhythm of the accompaniment, the cross correlation evaluation value HM_AV becomes a large value. The cross-correlation evaluation value HM_AV calculated by the cross-correlation evaluation value calculation unit 15 is sent to the music determination evaluation value calculation unit 16.

The music determination evaluation value calculation unit 16 includes an autocorrelation evaluation value HS_AVmax calculated by the autocorrelation evaluation value calculation unit 14, a crosscorrelation evaluation value HM_AV calculated by the crosscorrelation evaluation value calculation unit 15, and a corresponding evaluation value. based on the mean power HS_AV0 voiceband signal X _n in the calculation interval, calculates an evaluation value evaluation value U _t music determination. Specifically, the music determination evaluation value U _t is calculated based on the following equation (3).

U _t = {HS_AVmax / HS_AV0} * {K1 + (HM_AV / HS_AV0})
... (3)

  In the above formula (3), K1 is a constant, and here K1 = 2.

The music determination evaluation value U _t calculated by the music determination evaluation value calculation unit 16 is given to the extraction unit 17. Based on the music determination evaluation value U _t given from the music determination evaluation value calculation unit 16, the extraction unit 17 determines whether or not the corresponding evaluation value calculation section is music.

Whether or not the corresponding evaluation value calculation section is music is performed by comparing the music determination evaluation value U _t with a threshold value α. If U _t > α, it is determined that the evaluation value calculation section is music, and if U _t ≦ α, it is determined that the evaluation value calculation section is non-music. Then, the part where the music evaluation value calculation section continues is specified as the music signal part. The start position and end position of the specified music signal part are detected.

It should be noted that in a portion where music and non-music are switched, detailed analysis may be performed to detect an exact switching location (start position and end position of the music signal portion). For example, as shown in FIG. 4, when U _t−1 > α is determined in a certain evaluation value calculation section D _t−1 and U _t ≦ α is determined in the next section D _t , both these sections Input signals corresponding to D _t-1 and D _t are fetched from the input signal storage memory 1 (see FIG. 1). Then, a power value is calculated for every 256 samples for the captured input signal, and the presence or absence of a silent section is determined. If there is a silent section, the section is detected as an accurate switching point.

  When there is no silent section, frequency analysis by FFT is performed for every 256 samples, and the portion with the largest amount of change (sum of differences) is detected as an accurate switching location. In this way, the start position and end position of the audio signal part can be detected more accurately.

The voice band cutoff filter 13 in FIG. 2 is omitted, and an input signal (a signal including the voice band signal X _n and the signal Y _n other than the voice band signal) is input as the signal Y _n input to the cross-correlation evaluation value calculation unit 15. You may make it use. That is, instead of calculating the cross-correlation evaluation value for evaluating the correlation between the voice band signal and the out-of-band signal, the cross-correlation for evaluating the correlation between the voice band signal and the input signal. An evaluation value may be calculated.

Assuming that the input signal (X _n + Y _n ) is Y _n and the obtained cross-correlation evaluation value is HM_AV_type2, the music determination evaluation value U _t is expressed by the following equation (4).

  In the above formula (4), K2 is a constant. If K2 = 1, this is equivalent to the case where the audio band cut-off filter 13 is provided.

  FIG. 5 shows a modification of the music signal extraction unit 2.

  The difference from the music signal extraction unit 2 in FIG. 2 is that instead of the audio band pass filter 12 and the audio band cut filter 13 in FIG. 2, an FFT analysis unit 21, an audio band signal power calculation unit 22, and an out-of-band signal power calculation. This is in that the portion 23 is provided.

The FFT analysis unit 21 performs FFT analysis for every 1024 samples (every small section) of the input signal. The subsection number is n, the frequency axis argument is m (0 to 511), and the obtained power spectrum value is represented by PS _{(n, m)} . Note that m = 7 to m = 70 corresponds to the voice band.

Voice band signal power calculator 22, based on the following equation (5), to calculate the cumulative value Ps_x _n of the power spectrum in the speech band.

The voice band out-of-band signal power calculation unit 23 calculates the cumulative value PS_Y _n of the power spectrum in the band outside the voice band based on the following equation (6).

Autocorrelation evaluation value calculation unit 24, based on the power spectrum accumulated value Ps_x _n calculated by the audio band signal power calculation unit 22, a predetermined number of samples (in this example 512 * 1024 samples) consisting of sections (evaluation value calculating section ), The autocorrelation evaluation value (the sum of the autocorrelation of the power values) is calculated every time. When calculating the autocorrelation evaluation value for each evaluation value calculation section, the shorter the Ps_x _n for each small section of the evaluation value in the calculation interval, a predetermined number of immediately before the evaluation value calculation section (256 in this example) and Ps_x _n intervals is used.

A method for calculating an autocorrelation evaluation value in an arbitrary evaluation value calculation section will be described. When calculating the autocorrelation evaluation value of a certain evaluation value calculation section is to represent the power spectrum accumulated value of the small section of the top of the evaluation value calculation interval as Ps_x _0.

First, an autocorrelation evaluation value HS_AV _j (j = 0 to 255) is calculated based on the following equation (7).

The autocorrelation evaluation value HS_AV _j is a large value if the audio band signal has periodicity. That is, if the voice has a rhythmic sense, the autocorrelation evaluation value HS_AV _j is a large value.

Next, HS_AV _MAX is a value where HS_AV _j is maximized in the range of j> 10. Let HS_AV _{MAX be the} autocorrelation evaluation value of the evaluation value calculation section. The reason why j> 10 is set is that HS_AV _j may increase in the range of j ≦ 10 even if there is no periodicity. The autocorrelation evaluation value HS_AV _MAX calculated by the autocorrelation evaluation value calculation unit 24 is sent to the music determination evaluation value calculation unit 16.

Cross-correlation evaluation value calculation unit 25, based on the power spectrum accumulated value Ps_y _n calculated by the audio band signal power calculation unit 22 a power spectrum accumulated value Ps_x _n and voice-band signal power calculation unit 23 calculated by a predetermined For each section (evaluation value calculation section) consisting of the number of samples (512 * 1024 samples in this example), a cross-correlation evaluation value (total sum of cross correlations of power values) is calculated. Specifically, the cross-correlation evaluation value HM_AV is calculated based on the following formula (8).

  The cross-correlation evaluation value HM_AV is a large value if there is a correlation between the voice band signal and the out-of-band signal. That is, if the rhythm of the sound matches the rhythm of the accompaniment, the cross correlation evaluation value HM_AV becomes a large value. The cross correlation evaluation value HM_AV calculated by the cross correlation evaluation value calculation unit 25 is sent to the music determination evaluation value calculation unit 16.

  The operations of the music determination evaluation value calculation unit 16 and the extraction unit 17 are the same as those of the music signal extraction unit 2 of FIG.

  In the above modification, the cross-correlation evaluation value calculation unit 25 calculates the cross-correlation evaluation value for evaluating the correlation between the power spectrum cumulative value in the voice band and the power spectrum cumulative value in the band outside the voice band. Although calculated, a cross-correlation evaluation value for evaluating the correlation between the power spectrum cumulative value in the voice band and the power spectrum cumulative value in the entire frequency band of the input signal may be calculated. .

  If the input signal is a compressed audio signal, the same result can be obtained even if MDCT coefficients generated in the process of decoding are used instead of FFT analysis.

[3] Description of Music Signal Feature Analysis Unit 3 and Condition Input Unit 4 FIG. 6 shows a detailed configuration of the music signal feature analysis unit 3 and the condition input unit 4.

[3-1] Description of Music Signal Feature Analysis Unit 3 The music signal feature analysis unit 3 includes an initial music sensitivity space generation unit 101 that generates a music sensitivity space through listening experiments, and a music signal extraction unit 2 (see FIG. 1). And an automatic indexing unit 102 for automatically calculating music sensitivity space coordinate values of the music of the music signal portion specified by (hereinafter referred to as extracted music).

[3-1-1] Description of the Initial Music Sensitivity Space Generation Unit 101 The initial music sensitivity space generation unit 101 is calculated by a song impression evaluation unit 111 that calculates a song impression value by a listening experiment, and a song impression evaluation unit 111. And a factor analysis unit 112 that performs factor analysis using the music impression value.

  The song impression evaluation unit 111 will be described. For 10 subjects, 100 sample pop songs were presented, and each sensitivity word pair was evaluated on a 7-point scale.

The following four types of sensitivity word pairs were used.
(1) Bright-dark (2) clear-crammed (3) smooth-crisp (4) mild-intense

  From this experimental result, the music impression value for each sensitivity word pair was calculated for each sample music piece. The music impression value for each sensitivity word pair is represented by an average value of the evaluation values of the sensitivity word pairs.

  The factor analysis unit 112 performs a factor analysis with the music impression value for each sensitivity word pair of each sample music as a multivariate, and for each sample music, the psychological correlation of the music impression with each factor as an axis. The coordinate data (music sensitivity space coordinate value) of the space to express (music sensitivity space) is obtained. Thus, the music sensitivity space is a space having a plurality of factor axes. In this example, it is assumed that there are three types of factors and three types of factor axes.

  Each sample music, the music impression value and the music sensitivity space coordinate value obtained for each sample music are used as a teacher signal of the automatic indexing unit 102.

[3-1-2] Description of Automatic Indexing Unit 102 The automatic indexing unit 102 analyzes the extracted music and extracts a physical feature value extraction unit 121 that extracts physical feature values such as tempo and rhythm. A physical feature value / musical impression value conversion unit 122 that converts the physical feature value extracted by the feature value extraction unit 121 into a music impression value for each sensitivity word pair, and a physical feature value / musical impression value conversion unit 122. Is provided with a music impression value / music sensitivity space coordinate conversion unit 123 that converts the music impression value for each sensitivity word pair obtained into music sensitivity space coordinate values. The music sensitivity space coordinate value obtained by the music impression value / music sensitivity space coordinate conversion unit 123 is given to the desired music selection unit 5 (see FIG. 1).

FIG. 7 shows the configuration of the automatic indexing unit 102 more specifically.
The physical feature quantity / musical impression value conversion unit 122 includes regression analysis units 122a to 122d corresponding to four types of sensitivity word pairs. The music impression value / musical sensitivity space coordinate conversion unit 123 includes a neural network.

  A plurality of physical feature quantities related to the corresponding sensitivity word pairs are input to the regression analysis units 122 a to 122 d in the physical feature quantity / musical impression value conversion unit 122. The types of physical feature values input to the regression analysis units 122a to 122d are determined in advance.

  Each regression analysis part 122a-122d converts a physical feature-value into the music impression value of the corresponding sensitivity word pair based on following Formula (9).

  In the above equation (9), OUT is an output value of the regression analysis unit, and represents an estimated value of the music impression value regarding the sensitivity word pair corresponding to the regression analysis unit. In [k] represents the k-th physical feature amount input to the regression analysis unit. W [k] is a weighting coefficient for In [k]. W [0] is a weighting coefficient.

  The weighting factors W [k] and W [0] in the above equation (9) are obtained in advance based on the above-mentioned listening experiment results and the physical feature values of the sample music pieces used for the above. The weighting factors W [k] and W [0] are determined as follows. The result of the listening experiment, that is, the song impression value for each sensitivity word pair obtained for each sample song is used as a teacher signal (output value OUT). Further, the physical feature amount extracted by the physical feature amount extraction unit 121 for each sample music piece is set as the input value In. Then, the weighting factors W [k] and W [0] are calculated so that the regression analysis results for all the sample music pieces have values closest to the teacher signal for the sample music piece.

  Learning of the neural network constituting the music impression value / musical sensitivity space coordinate conversion unit 123 will be described. The input signal at the time of learning is the result of the above-described listening experiment, that is, the song impression value for each sensitivity word pair obtained for each sample song. The teacher signal at the time of learning is a music sensitivity space coordinate value (coordinate value of each factor axis) of each song obtained by the factor analysis unit 112.

[3-2] Description of Condition Input Unit 4 The condition input unit 4 includes a user input unit 201 and a music impression value / musical sensitivity space coordinate conversion unit 202 as shown in FIG.

  The user inputs a song impression value (seven levels) for each sensitivity word pair related to the song to be stored in the storage memory 6 to the user input unit 201. The music impression value for each sensitivity word pair input to the user input unit 201 is given to the music impression value / music sensitivity space coordinate conversion unit 202. As the music impression value / musical sensitivity space coordinate conversion unit 202, a neural network constituting the music impression value / musical sensitivity space coordinate conversion unit 123 in the automatic indexing unit 102 is used as it is.

  The music impression value / musical sensitivity space coordinate conversion unit 202 inputs the music impression value for each sensitivity word pair given from the user input unit 201 to the neural network, and acquires the music sensitivity space coordinate value corresponding thereto. . The song impression value / musical sensitivity space coordinate conversion unit 202 gives the acquired music sensitivity space coordinate value to the desired music selection unit 5.

[4] Description of Desired Music Selection Unit 5 The operation of the desired music selection unit 5 will be described. The music sensitivity space coordinate values for the extracted music obtained by the automatic indexing unit 102 in the music signal feature analysis unit 3 are represented by (Z1 [i], Z2 [i], Z3 [i]). i is a song number. The music impression space coordinate value obtained by the music impression value / music sensitivity space coordinate conversion unit 202 in the condition input unit 4 is defined as (a1, a2, a3).

  The desired music selection unit 5 calculates the spatial distance L of the extracted music based on the following equation (10). This spatial distance L indicates the degree to which the extracted music matches the storage condition desired by the user. However, the greater the spatial distance L, the lower the degree of matching, and the smaller the spatial distance L, the higher the degree of matching.

L = (Z1 [i] −a1) ² + (Z2 [i] −a2) ² + (Z3 [i] −a3) ² (10)

  Then, it is determined whether or not the spatial distance L is smaller than a predetermined threshold A. When the distance L is smaller than the predetermined threshold A, it is determined that the extracted music is stored in the storage memory 6. If the distance L is greater than or equal to the predetermined threshold A, it is determined that the extracted music is not stored in the storage memory 6. If it is determined to be stored in the storage memory 6, the extracted music is read from the input signal storage memory 1 and stored in the storage memory 6.

  The desired music selection unit 5 may determine whether or not to store the music data in consideration of the music sensitivity space coordinate value and the spatial distance L of the music data already stored in the storage memory 6.

  FIG. 8 shows a determination processing procedure by the desired music selection unit 5.

  Here, the music sensitivity space coordinate value for each music data already stored in the storage memory 6 is represented by (Z1 [i], Z2 [i], Z3 [i]). Here, i is assumed to be 0 to (n−1). The music sensitivity space coordinate values for the extracted music obtained by the automatic indexing unit 102 in the music signal feature analysis unit 3 are represented by (Z1 [n], Z2 [n], Z3 [n]).

  First, the spatial distance L of the extracted music is calculated (step S1). Next, it is determined whether or not the distance L is smaller than the threshold value S (step S2). However, the threshold value S is obtained by the following equation (11) with A as a fixed value.

  S = A × (1 + remaining memory capacity / total memory capacity) (11)

  In other words, the threshold value S increases as the remaining memory capacity increases, and therefore, the possibility that it is determined that the threshold value S is accumulated increases as the remaining memory capacity increases.

  If it is determined in step S2 that L ≧ S, it is determined that the extracted music is not stored (step S7).

  If it is determined in step S2 that L <S, it is determined whether or not the same music as the extracted music is already stored in the storage memory (step S3). This determination is performed by determining whether or not there is music data that satisfies the condition of the following expression (12), where β is a threshold value.

Z1 [n] -Z1 [i] <β and Z2 [n] -Z2 [i] <β and Z2 [n] -Z3 [i] <β (12)

  If there is music data that satisfies the condition of equation (12), it is determined that the same music as the extracted music is already stored in the storage memory, and it is determined that the extracted music is not stored (step S7). .

  If there is no music data that satisfies the condition of Expression (12), it is determined that the same song as the extracted music is not already stored in the storage memory, and whether or not the data size of the extracted music is smaller than the remaining memory capacity. Is determined (step S4). If the data size of the extracted music is smaller than the remaining memory capacity, it is determined that the extracted music is stored (step S6).

  If it is determined in step S4 that the data size of the extracted music is greater than or equal to the remaining memory capacity, the stored music data having the largest spatial distance L is deleted (step S5). Then, the process returns to step S4. Steps S4 and S5 are repeated until it is determined in step S4 that the data size of the extracted music is greater than or equal to the remaining memory capacity. If it is determined in step S4 that the data size of the extracted music is greater than or equal to the remaining memory capacity, it is determined that the extracted music is to be stored (step S6).

  According to the above embodiment, if the user designates a song impression value for each sensitivity word pair such as “bright-dark”, only the music signal corresponding to the music desired by the user is automatically stored in the storage memory. It can be accumulated.

  In addition, when the remaining capacity of the storage memory is large, the threshold value S is controlled to increase the possibility that it is determined to be stored, so that music data is easily stored in the storage memory. When the remaining capacity of the storage memory is smaller than the data size of the extracted music, the music data stored in the storage memory having the largest spatial distance is deleted, so that the storage memory contains the desired music. Near music data will be accumulated. It is also possible to prevent the same music from being accumulated repeatedly.

  9, 10 and 11 show modifications of the music signal storage device.

  FIG. 9 shows a configuration example of the music signal storage device when the input signal is uncompressed data, as in the above embodiment. This music signal storage device is provided with an encode 7 for encoding music data when storing the music data in the storage memory 6. Therefore, encoded data obtained by the encode 7 is stored in the storage memory 6.

  FIG. 10 shows a configuration example of a music signal storage device when the input signal is encoded data. In this music signal storage device, a decoder 8 for decoding an input signal is provided at a subsequent stage of the input signal storage memory 1. The input signal (encoded data) is stored in the input signal storage memory 1. The input signal stored in the input signal storage memory 1 is decoded by the decoder 8 and then given to the music signal extraction unit 2 and the music signal feature analysis unit 3. When the music data is stored in the storage memory 6, the encoded data in the input signal storage memory 1 is stored in the storage memory 6.

  FIG. 11 shows a configuration example of the music signal storage device when the input signal is uncompressed data. In this music signal storage device, an input signal is sent to the input signal storage memory 1 and to the encoder 9. The data encoded by the encoder 9 is stored in the code data temporary storage memory 10. When the music data is stored in the storage memory 6, the encoded data in the code data temporary storage memory 10 is stored in the storage memory 6.

FIG. 12 shows a further modification of the music signal storage device.
In this music signal storage device, in order to efficiently extract and store desired music from a plurality of input sources (broadcast 1 to broadcast n), the input signal storage memory 1, the music signal extraction unit 2 and the music signal in the first embodiment are stored. A plurality of combinations of feature analysis units 3 are provided. Each system is connected to the desired music selection unit 5. Input signals of different channels are input to each system. A playback device 30 for playing back music data stored in the storage memory 6 is connected to the subsequent stage of the storage memory 6.

It is a block diagram which shows the structure of a music signal storage apparatus. It is a block diagram which shows the structure of a music signal extraction part. Autocorrelation evaluation value calculating section 4 is a schematic diagram showing that for calculating the autocorrelation evaluation value for each section D _t of a predetermined number of samples. It is explanatory drawing for demonstrating the method of performing a detailed analysis in the part which the evaluation value calculation area of music and the non-music evaluation value calculation area switch, and detecting an exact switching location. It is a block diagram which shows the modification of the music signal extraction part. 3 is a block diagram showing a detailed configuration of a music signal feature analysis unit 3 and a condition input unit 4. FIG. 2 is a block diagram showing more specifically the configuration of an automatic indexing unit 102. FIG. It is a flowchart which shows the determination processing procedure by the condition input part. It is a block diagram which shows the modification of a music signal storage device. It is a block diagram which shows the other modification of a music signal storage apparatus. It is a block diagram which shows the other modification of a music signal storage device. It is a block diagram which shows the other modification of a music signal storage device.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input signal storage memory 2 Music signal extraction part 3 Music signal characteristic analysis part 4 Condition input part 5 Desired music selection part 6 Accumulation memory

Claims (4)

In a music signal storage device that extracts only a music signal that matches a condition corresponding to a user's favorite music from an input voice signal and stores it in a storage device,
Music signal extraction means for analyzing the input audio signal and extracting only the music signal;
Music feature analysis means for analyzing the characteristics of the music based on the music signal extracted by the music signal extraction means;
Condition input means for inputting conditions corresponding to the user's favorite music, and extracted by the music signal extraction means based on the analysis result of the music feature analysis means and the conditions input to the condition input means Determination means for determining whether or not to store the stored music signal in the storage device,
The determination means includes
First means for calculating a degree that the music feature analysis result of the music feature analysis means matches a condition input to the condition input means; and comparing the degree calculated by the first means with a threshold value. the includes a second hand stage determines whether to accumulate the music signal extracted by said music signal extraction means in the storage device,
The second means includes
A third means for determining that the music signal extracted by the music signal extraction means is not stored in the storage device when the degree calculated by the first means is lower than the threshold;
Fourth means for determining whether the data size of the music signal extracted by the music signal extraction means is smaller than the remaining capacity of the storage device when the degree calculated by the first means is higher than the threshold value. ,
When the fourth means determines that the data size of the music signal extracted by the music signal extraction means is smaller than the remaining capacity of the storage device, the music signal extracted by the music signal extraction means is stored in the storage device. A fifth means for determining to accumulate in
When the fourth means determines that the data size of the music signal extracted by the music signal extraction means is greater than or equal to the remaining capacity of the storage device, the size of the music signal extracted by the music signal extraction means is The music signals stored in the storage device are deleted in ascending order of the degree that the music feature analysis result matches the condition input to the condition input means until the remaining capacity of the storage device becomes smaller. A sixth means for determining that the music signal extracted by the music signal extracting means is stored in the storage device ;
A music signal storage device comprising: In a music signal storage device that extracts only a music signal that matches a condition corresponding to a user's favorite music from an input voice signal and stores it in a storage device,
Music signal extraction means for analyzing the input audio signal and extracting only the music signal;
Music feature analysis means for analyzing the characteristics of the music based on the music signal extracted by the music signal extraction means;
Condition input means for inputting conditions corresponding to the user's favorite music; and
Determining means for determining whether or not to store the music signal extracted by the music signal extracting means in the storage device based on the analysis result of the music feature analyzing means and the condition input to the condition input means; Has
The determination means includes
First means for calculating a degree to which a music feature analysis result of the music feature analysis means matches a condition input to the condition input means; and
A second means for determining whether or not to store the music signal extracted by the music signal extraction means in the storage device by comparing the degree calculated by the first means with a threshold;
The second means includes
A third means for determining that the music signal extracted by the music signal extraction means is not stored in the storage device when the degree calculated by the first means is lower than the threshold;
When the degree calculated by the first means is higher than the threshold, the characteristic amount analysis result of the music signal extracted by the music signal extraction means and the characteristic analysis of each music signal stored in the storage device A fourth means for determining, based on the result, whether or not a music signal whose characteristic analysis result is the same as the music signal extracted by the music signal extracting means is already stored in the storage device;
If it is determined by the fourth means that a music signal whose characteristic analysis result is the same as the music signal extracted by the music signal extracting means is already stored in the storage device, the music signal extraction A fifth means for determining that the music signal extracted by the means is not stored in the storage device;
If it is determined by the fourth means that the music signal extracted by the music signal extracting means is not stored in the storage device with a music signal whose characteristic analysis result is the same, the music signal extracting means Sixth means for determining whether or not the data size of the music signal extracted by the step is smaller than the remaining capacity of the storage device;
When the sixth means determines that the data size of the music signal extracted by the music signal extraction means is smaller than the remaining capacity of the storage device, the music signal extracted by the music signal extraction means is stored in the storage device. A seventh means for determining that it is accumulated in
When the sixth means determines that the data size of the music signal extracted by the music signal extraction means is equal to or larger than the remaining capacity of the storage device, the size of the music signal extracted by the music signal extraction means is The music signals stored in the storage device are deleted in ascending order of the degree that the music feature analysis result matches the condition input to the condition input means until the remaining capacity of the storage device becomes smaller. And an eighth means for determining that the music signal extracted by the music signal extracting means is stored in the storage device;
Features and to Ruoto music signal storage device, further comprising a. A music signal storage program for extracting only a music signal suitable for a condition corresponding to a user's favorite music from an input voice signal and storing it in a storage device,
A music signal extraction step of analyzing the input audio signal and extracting only the music signal;
A song feature analysis step for analyzing the feature of the song based on the music signal extracted by the music signal extraction step;
A condition input step for inputting conditions corresponding to the user's favorite music; and
A determination step of determining whether or not to store the music signal extracted in the music signal extraction step in the storage device based on the analysis result of the music feature analysis step and the condition input in the condition input step; A music signal storage program to be executed by a computer,
The determination step includes
A first step of calculating a degree to which the music feature analysis result of the music feature analysis step matches the condition input in the condition input step; and
A second step of determining whether to store the music signal extracted by the music signal extraction step in the storage device by comparing the degree calculated by the first step with a threshold;
The second step includes
A third step of determining that the music signal extracted by the music signal extraction step is not stored in the storage device when the degree calculated by the first step is lower than the threshold;
When the degree calculated by the first step is higher than the threshold value, a fourth step determines whether or not the data size of the music signal extracted by the music signal extraction step is smaller than the remaining capacity of the storage device ,
When it is determined by the fourth step that the data size of the music signal extracted by the music signal extraction step is smaller than the remaining capacity of the storage device, the music signal extracted by the music signal extraction step is stored in the storage device. A fifth step for determining to accumulate, and
When it is determined by the fourth step that the data size of the music signal extracted by the music signal extraction step is greater than or equal to the remaining capacity of the storage device, the size of the music signal extracted by the music signal extraction step is The music signals stored in the storage device are deleted in ascending order of the degree that the music feature analysis result matches the condition input in the condition input step until the remaining capacity of the storage device becomes smaller. A sixth step of determining that the music signal extracted by the music signal extraction step is to be stored in the storage device;
Features and to Ruoto music signal stored program in that it comprises. A music signal storage program for extracting only a music signal suitable for a condition corresponding to a user's favorite music from an input voice signal and storing it in a storage device,
A music signal extraction step of analyzing the input audio signal and extracting only the music signal;
A song feature analysis step for analyzing the feature of the song based on the music signal extracted by the music signal extraction step;
A condition input step for inputting conditions corresponding to the user's favorite music; and
A determination step of determining whether or not to store the music signal extracted in the music signal extraction step in the storage device based on the analysis result of the music feature analysis step and the condition input in the condition input step; A music signal storage program to be executed by a computer,
The determination step includes
A first step of calculating a degree to which the music feature analysis result of the music feature analysis step matches the condition input in the condition input step; and
A second step of determining whether to store the music signal extracted by the music signal extraction step in the storage device by comparing the degree calculated by the first step with a threshold;
The second step includes
A third step of determining that the music signal extracted by the music signal extraction step is not stored in the storage device when the degree calculated by the first step is lower than the threshold;
When the degree calculated by the first step is higher than the threshold, the feature amount analysis result of the music signal extracted by the music signal extraction step and the feature analysis of each music signal stored in the storage device A fourth step for determining whether or not a music signal whose characteristic analysis result is the same as the music signal extracted in the music signal extraction step is already stored in the storage device based on the result ;
If it is determined by the fourth step that the music signal extracted in the music signal extraction step and the music signal whose characteristic analysis result is the same are already stored in the storage device, the music signal extraction is performed. A fifth step of determining that the music signal extracted in the step is not stored in the storage device;
If it is determined by the fourth step that a music signal whose feature analysis result is the same as the music signal extracted by the music signal extraction step is not stored in the storage device, the music signal extraction step A sixth step of determining whether or not the data size of the music signal extracted by the step is smaller than the remaining capacity of the storage device;
When the sixth step determines that the data size of the music signal extracted by the music signal extraction step is smaller than the remaining capacity of the storage device, the music signal extracted by the music signal extraction step is stored in the storage device. A seventh step for determining to accumulate, and
When it is determined by the sixth step that the data size of the music signal extracted by the music signal extraction step is greater than or equal to the remaining capacity of the storage device, the size of the music signal extracted by the music signal extraction step is The music signals stored in the storage device are deleted in ascending order of the degree that the music feature analysis result matches the condition input in the condition input step until the remaining capacity of the storage device becomes smaller. And an eighth step of determining that the music signal extracted by the music signal extraction step is stored in the storage device.
Features and to Ruoto music signal stored program in that it comprises.

Images