JP7025144B2 - Electronic melody identification device, program, and electronic melody identification method - Google Patents

Description

The present invention relates to an electronic melody specifying device, a program, and an electronic melody specifying method.

The following Patent Document 1 discloses a voice recognition device that recognizes input voice by preparing various voice standard patterns in advance and performing pattern matching between the input voice and the standard pattern.

Japanese Unexamined Patent Publication No. 62-111293

Voices uttered by humans, even if they are the same word or phrase, have various spectral fluctuations and temporal fluctuations due to the individuality of the speaker such as gender, age, accent, and vocalization speed. Therefore, in a voice recognition device that recognizes the voice of an unspecified speaker, it is necessary to implement a complicated normalization circuit or the like in order to correct these fluctuations caused by the individuality of the speaker. For example, in the speech recognition device disclosed in Patent Document 1, pattern matching with a standard pattern is performed while shifting linearly expanded and contracted input speech by one frame in the entire section in order to correct temporal fluctuations.

On the other hand, electronic melodies such as station arrival / departure melody and store entry melody are often simple music created by synthesizers, and the frequency band is narrower and the frequency component is also narrower than the voice uttered by humans. Few. Moreover, unlike the voice uttered by humans, the electronic melody only reproduces the recorded sound source from the speaker, so that there is no spectral fluctuation or temporal fluctuation due to the individuality of the speaker.

Therefore, if an electronic melody specifying device for specifying an electronic melody is equipped with a complicated function equivalent to that of a voice recognition device, it will be over-engineered and the cost will increase unnecessarily.

The present invention has been made to solve such a problem, and obtains an electronic melody specifying device, a program, and an electronic melody specifying method capable of specifying an electronic melody with high accuracy by a simple configuration and processing. The purpose is to do that.

The electronic melody specifying device according to the first aspect of the present invention has an acquisition unit that acquires a time-series signal of an electronic melody to be specified, and a predetermined time-series signal of the electronic melody acquired by the acquisition unit. By frequency analysis in units of one section, a first pattern generation unit that generates a first spectrum pattern in which a plurality of first frequency spectra for each section are arranged is generated by the first pattern generation unit and the first pattern generation unit. The first spectral pattern is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and a plurality of first sections relating to the plurality of first sections included in each second section. By averaging the frequency spectra, a second pattern generator that generates a second spectrum pattern in which a plurality of second frequency spectra for each second section are arranged is generated in advance for each of a plurality of known electronic melodies. A storage unit that stores a third spectrum pattern and is divided into N second sections, a second spectrum pattern generated by the second pattern generation unit, and each stored in the storage unit. A calculation processing unit for calculating the similarity with the third spectrum pattern and a specific processing unit for specifying the electronic melody acquired by the acquisition unit based on the calculation result of the similarity with the calculation processing unit are provided. The second pattern generation unit is characterized in that the number of first sections included in each second section can be non-uniformly distributed based on the distribution of frequency components in the second spectral pattern. be.

According to the electronic melody specifying device according to the first aspect, the second pattern generation unit has a second pattern generation unit by averaging a plurality of first frequency spectra having a plurality of first sections included in each second section. A second spectrum pattern in which a plurality of second frequency spectra for each of two sections are arranged is generated. Then, the calculation processing unit calculates the degree of similarity between the second spectrum pattern generated by the second pattern generation unit and each third spectrum pattern stored in the storage unit. As a result, the amount of calculation is reduced as compared with the case where the similarity is calculated using the first spectral pattern before averaging, so that the processing load can be reduced. Further, since the third spectrum pattern can be simplified, the storage capacity required for the storage unit can be reduced.

Moreover, the second pattern generation unit can non-uniformly distribute the number of the first sections included in each second section based on the distribution of the frequency components in the second spectral pattern. Since the resolution can be adjusted according to the number of the first section included in the second section, the processing load is avoided by increasing the resolution of the important part and lowering the resolution of the unimportant part. However, it is possible to improve the identification accuracy of the electronic melody.

The electronic melody specifying device according to the second aspect of the present invention is the electronic melody specifying device according to the first aspect. The first section of the first number is allocated to the second section corresponding to the included second frequency spectrum, and when evenly distributed, the second frequency containing the second number of frequency components less than the first number. It is characterized in that the first section of the second number, which is larger than the first number , is allocated to the second section corresponding to the spectrum.

According to the electronic melody specifying device according to the second aspect, the second pattern generator is for the second section corresponding to the second frequency spectrum including a relatively large number (first number) of frequency components. A relatively small number (first number) of the first section is allocated, and a relatively large number (second number) is used for the second section corresponding to the second frequency spectrum containing a relatively small number (second number) of frequency components . Allocate the first section of (number of 2) . Since the second frequency spectrum containing a relatively large number of frequency components is an important part for identifying the electronic melody, a relatively small number of second intervals correspond to the second frequency spectrum corresponding to such a second frequency spectrum. By allocating one section, the resolution can be increased, and as a result, the accuracy of specifying the electronic melody can be improved. On the other hand, since the second frequency spectrum containing a relatively small number of frequency components is not important for identifying the electronic melody, it is relatively large for the second section corresponding to such a second frequency spectrum. By allocating the first section of the above, it is possible to avoid an increase in the processing load.

The electronic melody specifying device according to the third aspect of the present invention is the electronic melody specifying device according to the first or second aspect, and in particular, the storage unit includes each of the N second sections. The third spectrum pattern of the first group in which the number of one section is unevenly distributed, and the third of the second group in which the number of the first sections included in each of the N second sections is evenly distributed. The spectrum pattern and the like are stored, and the calculation processing unit has a non-uniform number of first sections included in each second section in the second spectrum pattern generated by the second pattern generation unit. If it is distributed, the similarity between the second spectrum pattern and the third spectrum pattern of the first group stored in the storage unit is calculated, and the second pattern generation unit generates the second spectrum pattern. When the number of the first section included in each second section is evenly distributed in the two spectrum patterns, the second spectrum pattern and the third spectrum of the second group stored in the storage unit are used. It is characterized by calculating the degree of similarity with the pattern.

According to the electronic melody specifying device according to the third aspect, the calculation processing unit is in the first group with respect to the second spectrum pattern in which the number of the first sections included in each second section is unevenly distributed. On the other hand, for the second spectrum pattern in which the number of the first sections included in each second section is evenly distributed, the third spectrum pattern of the second group is calculated. Calculate the similarity with. In this way, the electronic melody is specified by classifying the third spectrum pattern into the first group and the second group in advance and selecting the third spectrum pattern to be compared according to the type of the second spectrum pattern. It is possible to reduce the processing load and the required time while suppressing the decrease in accuracy.

The electronic melody specifying device according to the fourth aspect of the present invention is the electronic melody specifying device according to the third aspect, and in particular, the calculation processing unit is the second spectrum pattern generated by the second pattern generation unit. When the similarity with the third spectrum pattern of the first group stored in the storage unit is less than a predetermined threshold value, the second spectrum pattern and the storage unit are further stored. The similarity with the third spectrum pattern of the second group is calculated, and the second spectrum pattern generated by the second pattern generation unit and the third spectrum of the second group stored in the storage unit are stored. When the similarity with the pattern is less than a predetermined threshold value, the similarity between the second spectrum pattern and the third spectrum pattern of the first group stored in the storage unit is further calculated. It is characterized by doing.

According to the electronic melody specifying device according to the fourth aspect, the calculation processing unit is of the first group with respect to the second spectrum pattern in which the number of the first sections included in each second section is unevenly distributed. When the similarity with the third spectrum pattern is less than the threshold value, the similarity between the second spectrum pattern and the third spectrum pattern of the second group is further calculated. Further, the calculation processing unit has a similarity with the third spectrum pattern of the second group of the second spectrum pattern in which the number of the first sections included in each second section is evenly distributed, which is less than the threshold value. In the case of, the similarity between the second spectrum pattern and the third spectrum pattern of the first group is further calculated. In this way, when the similarity with the third spectrum pattern of one of the first group and the second group is less than the threshold value, the similarity with the third spectrum pattern of the other group is calculated. As a result, there is a possibility that a third spectrum pattern having a similarity equal to or higher than the threshold value may be found, so that the accuracy of specifying the electronic melody can be improved.

The electronic melody specifying device according to the fifth aspect of the present invention is the electronic melody specifying device according to any one of the first to fourth aspects, and in particular, the calculation processing unit has a degree of similarity for each corresponding second section. It is characterized in that the representative value of N similarity degree obtained by calculating is calculated as the similarity degree between the second spectrum pattern and the third spectrum pattern.

According to the electronic melody specifying device according to the fifth aspect, the calculation processing unit calculates the similarity for each second section corresponding to the second spectrum pattern and the third spectrum pattern. Therefore, as compared with the case where the similarity is calculated for each first section, the amount of calculation is reduced, so that the processing load can be reduced.

The electronic melody specifying device according to the sixth aspect of the present invention is characterized in that, in particular, the representative value is an average value in the electronic melody specifying device according to the fifth aspect.

According to the electronic melody specifying device according to the sixth aspect, the calculation processing unit uses the average value of N similarities obtained by calculating the similarity for each corresponding second section as the second spectral pattern. It is calculated as the degree of similarity with the third spectrum pattern. This makes it possible to calculate the similarity based on the distribution of frequency components in the entire region of the spectral pattern.

The electronic melody specifying device according to the seventh aspect of the present invention is characterized in that the representative value is the maximum value in particular in the electronic melody specifying device according to the fifth aspect.

According to the electronic melody specifying device according to the seventh aspect, the calculation processing unit sets the maximum value of N similarities obtained by calculating the similarity for each corresponding second section as the second spectral pattern. It is calculated as the degree of similarity with the third spectrum pattern. This makes it possible to calculate the similarity of the electronic melody including the characteristic portion based on the distribution of the frequency components in the second frequency spectrum corresponding to the characteristic portion.

The electronic melody specifying device according to the eighth aspect of the present invention is particularly the electronic melody specifying device according to any one of the fifth to seventh aspects, and the calculation processing unit is generated based on the first spectrum pattern. (N-1) pseudo-second spectra obtained by cyclically shifting the arrangement order of the reference second spectrum pattern, which is the second spectrum pattern, and the N second sections included in the reference second spectrum pattern. It is characterized in that the similarity with the third spectrum pattern is calculated for each of the patterns.

According to the electronic melody specifying device according to the eighth aspect, the calculation processing unit obtains by cyclically shifting the arrangement order of the reference second spectrum pattern and the N second sections included in the reference second spectrum pattern. For each of the (N-1) pseudo-second spectral patterns to be performed, the similarity with the third spectral pattern is calculated. When the acquisition unit acquires the time-series signal of the electronic melody by a manual recording operation by the user, the second spectrum pattern and the third spectrum pattern are the first due to the difference between the start timing and the end timing of the recording operation. Misalignment between the two sections can occur. Therefore, the calculation processing unit calculates the similarity of not only the reference second spectrum pattern but also the pseudo second spectrum pattern with the third spectrum pattern, so that the position shift of the second section occurs. Also, the degree of similarity between the second spectrum pattern and the third spectrum pattern can be appropriately calculated.

Moreover, since the (N-1) pseudo second spectrum pattern obtained by cyclically shifting the arrangement order of the N second sections included in the reference second spectrum pattern is defined, the position of the second section is defined. Even when the deviation is large, it is possible to appropriately calculate the degree of similarity between the second spectrum pattern and the third spectrum pattern.

The electronic melody specifying device according to the ninth aspect of the present invention is particularly the electronic melody specifying device according to any one of the fifth to seventh aspects, and the calculation processing unit is generated based on the first spectrum pattern. Two pseudo-ths obtained by cyclically shifting the arrangement order of the reference second spectrum pattern, which is the second spectrum pattern, and the N second sections included in the reference second spectrum pattern, one section at a time in the front-back direction. It is characterized in that the similarity with the third spectrum pattern is calculated for each of the two spectrum patterns.

According to the electronic melody specifying device according to the ninth aspect, the calculation processing unit arranges the reference second spectrum pattern and the N second sections included in the reference second spectrum pattern one section at a time in the front-back direction. For each of the two pseudo-second spectral patterns obtained by cyclic shifting, the similarity with the third spectral pattern is calculated. This makes it possible to appropriately calculate the degree of similarity between the second spectrum pattern and the third spectrum pattern even when the position of the second section is displaced.

Moreover, since the two pseudo second spectrum patterns obtained by cyclically shifting the arrangement order of the N second sections included in the reference second spectrum pattern one section at a time in the front-back direction are defined, the entire section is defined. It is possible to appropriately calculate the degree of similarity between the second spectral pattern and the third spectral pattern while suppressing an increase in the processing load as compared with the cyclic shift.

The electronic melody specifying device according to the tenth aspect of the present invention is particularly the electronic melody specifying device according to any one of the fifth to seventh aspects, and the calculation processing unit is generated based on the first spectrum pattern. By excluding the second section of the head portion and the tail portion in chronological order from the N second sections included in the second spectrum pattern, a pseudo second section including (N-2) second sections is included. To calculate the similarity with the third spectrum pattern for each of the three pseudo second spectrum patterns obtained by defining the spectrum pattern and shifting the pseudo second spectrum pattern backward by one section in order from the beginning. It is characterized by.

According to the electronic melody specifying device according to the tenth aspect, the calculation processing unit excludes the second section of the head portion and the tail portion in the time series order from the N second sections included in the second spectrum pattern. Thereby, a pseudo second spectrum pattern including (N-2) second sections is defined, and the pseudo second spectrum pattern is shifted backward by one section in order from the beginning, and three pseudo second spectra are obtained. For each of the spectral patterns, the similarity with the third spectral pattern is calculated. This makes it possible to appropriately calculate the degree of similarity between the second spectrum pattern and the third spectrum pattern even when the position of the second section is displaced.

Moreover, since the pseudo second spectrum pattern is defined by excluding the second section of the head portion and the tail portion in the time series order from the N second sections included in the second spectrum pattern, the circulation of all sections. It is possible to suppress the increase in the processing load as compared with the shift, and to eliminate the influence of the second section of the head portion and the tail portion where the timing shift of the recording operation by the user is most likely to be reflected.

The electronic melody specifying device according to the eleventh aspect of the present invention is the electronic melody specifying device according to any one of the fifth to seventh aspects, and in particular, the calculation processing unit includes N elements included in the second spectrum pattern. Of the second section of the above, the second section of the head portion and the tail portion in chronological order is characterized in that the calculated similarity is multiplied by a weighting coefficient of less than 1.

According to the electronic melody specifying device according to the eleventh aspect, the calculation processing unit refers to the second section of the head portion and the tail portion in the time series order among the N second sections included in the second spectrum pattern. , Multiply the calculated similarity by a weighting factor of less than 1. Therefore, it is possible to reduce the influence of the second section of the beginning portion and the ending portion where the timing deviation of the recording operation by the user is most likely to be reflected.

The electronic melody specifying device according to the twelfth aspect of the present invention is the electronic melody specifying device according to any one of the first to eleventh aspects, and in particular, the first pattern generation unit uses a fast Fourier transform. The first frequency spectrum is generated by frequency analysis, and the second pattern generation unit averages a plurality of first frequency spectra to generate a second frequency spectrum, and is a complex signal before absolute value in the fast Fourier transform. It is characterized by averaging the frequencies.

According to the electronic melody specifying device according to the twelfth aspect, the second pattern generation unit averages a plurality of first frequency spectra to generate a second frequency spectrum before the absolute value in the fast Fourier transform. Perform averaging on complex signals. As a result, the influence of white noise mixed in the first frequency spectrum is reduced in the second frequency spectrum and the signal-to-noise ratio is improved, so that the accuracy of specifying the electronic melody can be improved.

The electronic melody specifying device according to the thirteenth aspect of the present invention is the electronic melody specifying device according to any one of the first to twelfth aspects, and in particular, the second pattern generation unit has a plurality of first frequency spectra. It is characterized in that a second frequency spectrum is generated by performing quantization using a predetermined threshold value on the frequency spectrum obtained by averaging.

According to the electronic melody specifying device according to the thirteenth aspect, the second pattern generator quantizes the frequency spectrum obtained by averaging the plurality of first frequency spectra using a predetermined threshold value. To generate a second frequency spectrum. Since the second spectrum pattern can be simplified by quantizing the second frequency spectrum, the amount of calculation can be reduced. Moreover, since the specific target is an electronic melody, even if the second frequency spectrum is quantized, the effect of reducing the specific accuracy is small. Therefore, it is possible to reduce the amount of calculation while suppressing the deterioration of the specific accuracy of the electronic melody.

The electronic melody specifying device according to the fourteenth aspect of the present invention is the electronic melody specifying device according to the thirteenth aspect, and in particular, the storage unit is averaged and quantized in the same manner as in the second frequency spectrum. It is characterized in that the third spectrum pattern is stored.

According to the electronic melody specifying device according to the fourteenth aspect, the storage unit stores a third spectrum pattern that has been averaged and quantized in the same manner as the second frequency spectrum. Since the third spectral pattern can be further simplified by quantization, the storage capacity required for the storage unit can be further reduced.

In the program according to the fifteenth aspect of the present invention, the computer mounted on the electronic melody specifying device has an acquisition means for acquiring a time-series signal of the electronic melody to be specified, and an electronic melody acquired by the acquisition means. A first pattern generation means for generating a first spectrum pattern in which a plurality of first frequency spectra for each first section are arranged by frequency analysis of a time-series signal in a predetermined first section unit, and the first. The first spectrum pattern generated by the pattern generation means of 1 is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and a plurality of pieces included in each second section. A second pattern generation means for generating a second spectrum pattern in which a plurality of second frequency spectra for each second section are arranged by averaging a plurality of first frequency spectra for the first section, and a plurality of known patterns. A storage means for storing a third spectrum pattern, which is pre-generated for each of the electronic melody of the above and divided into N second sections, a second spectrum pattern generated by the second pattern generation means, and the above-mentioned. The electronic melody acquired by the acquisition means is specified based on the calculation processing means for calculating the similarity with each third spectrum pattern stored in the storage means and the calculation result of the similarity with the calculation processing means. A program for functioning as a specific processing means, wherein the second pattern generation means is the number of first sections included in each second section based on the distribution of frequency components in the second spectral pattern. Is characterized by being able to be distributed unevenly.

According to the program according to the fifteenth aspect, the second pattern generation means is performed for each second section by averaging a plurality of first frequency spectra for the plurality of first sections included in each second section. A second spectrum pattern in which a plurality of second frequency spectra of the above are arranged is generated. Then, the calculation processing means calculates the degree of similarity between the second spectrum pattern generated by the second pattern generation means and each third spectrum pattern stored in the storage means. As a result, the amount of calculation is reduced as compared with the case where the similarity is calculated using the first spectral pattern before averaging, so that the processing load can be reduced. Further, since the third spectrum pattern can be simplified, the storage capacity required for the storage means can be reduced.

Moreover, the second pattern generation means can non-uniformly distribute the number of the first sections included in each second section based on the distribution of the frequency components in the second spectral pattern. Since the resolution can be adjusted according to the number of the first section included in the second section, the processing load is avoided by increasing the resolution of the important part and lowering the resolution of the unimportant part. However, it is possible to improve the identification accuracy of the electronic melody.

The method for specifying an electronic melody according to a sixteenth aspect of the present invention includes (A) a step of acquiring a time-series signal of the electronic melody to be specified, and (B) a time of the electronic melody acquired by the step (A). By frequency-analyzing the series signal in a predetermined first section unit, a step of generating a first spectrum pattern in which a plurality of first frequency spectra for each first section are arranged, and (C) the step (B). The generated first spectrum pattern is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and a plurality of first sections are included in each second section. By averaging the first frequency spectra, a step of generating a second spectrum pattern in which a plurality of second frequency spectra are arranged for each second section, and (D) pre-generated for each of a plurality of known electronic melodies. , The step of storing the third spectrum pattern divided into N second sections, (E) the second spectrum pattern generated by the step (C), and each stored by the step (D). A step of calculating the similarity with the third spectrum pattern, and (F) a step of specifying the electronic melody acquired by the step (A) based on the calculation result of the similarity by the step (E). The step (C) is characterized in that the number of the first sections included in each second section can be distributed non-uniformly based on the distribution of the frequency components in the second spectral pattern. be.

According to the electronic melody identification method according to the sixteenth aspect, in step (C), the second section is obtained by averaging the plurality of first frequency spectra of the plurality of first sections included in each second section. A second spectrum pattern in which a plurality of second frequency spectra are arranged for each is generated. Then, in step (E), the similarity between the second spectrum pattern generated by step (C) and each third spectrum pattern stored by step (D) is calculated. As a result, the amount of calculation is reduced as compared with the case where the similarity is calculated using the first spectral pattern before averaging, so that the processing load can be reduced. Further, since the third spectrum pattern can be simplified, the storage capacity required for storing the third spectrum pattern in the step (D) can be reduced.

Moreover, in step (C), the number of first sections included in each second section can be non-uniformly distributed based on the distribution of frequency components in the second spectral pattern. Since the resolution can be adjusted according to the number of the first section included in the second section, the processing load is avoided by increasing the resolution of the important part and lowering the resolution of the unimportant part. However, it is possible to improve the identification accuracy of the electronic melody.

According to the present invention, it is possible to specify an electronic melody with high accuracy by a simple configuration and processing.

It is a block diagram which shows the structure of the electronic melody specifying apparatus which concerns on embodiment of this invention. It is a figure which shows the relationship between the time-series signal of an electronic melody, and a frame. It is a figure which shows an example of the spectrum pattern generated by the frequency analysis part schematically. It is a figure which shows the relationship between a frame and a segment. It is a figure which shows an example of the spectrum pattern generated by the pattern generation part schematically. It is a figure which shows an example of the spectrum pattern generated by the pattern generation part schematically. It is a figure for demonstrating the calculation method of the similarity degree by the similarity degree calculation part. It is a figure which shows the structure for realizing the electronic melody identification apparatus by software. It is a figure which shows an example of the spectrum pattern generated by the even distribution. It is a figure which shows the relationship between a frame and a segment at the time of unequal distribution. It is a figure which shows the other example of the spectral pattern generated by the even distribution. It is a figure which shows the spectrum pattern which is stored in the storage part. It is a flowchart which shows 1st example about the processing flow of the similarity calculation unit and the specific processing unit. It is a flowchart which shows the 2nd example about the processing flow of the similarity calculation unit and the specific processing unit. It is a figure for demonstrating the 1st example about the position shift measure of the corresponding segment. It is a figure for demonstrating the 2nd example about the position shift measure of the corresponding segment. It is a figure for demonstrating the 3rd example about the position shift measure of the corresponding segment. It is a figure for demonstrating the 4th example about the position shift measure of the corresponding segment. It is a figure for demonstrating the 5th example about the position shift measure of the corresponding segment. It is a figure for demonstrating the sixth example about the position shift measure of the corresponding segment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In addition, the elements with the same reference numerals in different drawings indicate the same or corresponding elements.

<Basic configuration>
FIG. 1 is a block diagram showing a configuration of an electronic melody specifying device 1 according to an embodiment of the present invention. As an example of the application, the electronic melody specifying device 1 is mounted on a mobile terminal such as a smartphone or a game machine, and a user who visits a specific station or store can use the arrival / departure melody peculiar to the station or the like for proof of visit. It is used to record store-specific entry melody. As will be described later, the electronic melody specifying device 1 stores in advance reference spectrum patterns relating to various known reference electronic melody, and obtains the acquired electronic melody by comparing the spectral pattern with the reference spectral pattern. It is specified which reference electronic melody corresponds to the electronic melody.

As shown in the connection relationship of FIG. 1, the electronic melody specifying device 1 includes a microphone 11, an AD converter (ADC) 12, buffers 13, 15, frequency analysis unit 14, pattern generation unit 16, similarity calculation unit 17, and identification. It is configured to include a processing unit 18 and a storage unit 19. The microphone 11, the ADC 12, and the buffer 13 function as an acquisition unit 10 for acquiring an electronic melody by recording. Each element constituting the electronic melody specifying device 1 may be configured by hardware or software. Alternatively, the electronic melody specifying device 1 may be configured by using both hardware and software, such as performing simple calculation by hardware and performing determination processing by software.

When recording is started by the user operating the mobile terminal, the time-series signal of the electronic melody to be specified is input from the microphone 11 to the AD converter 12, and after being converted into digital data by the AD converter 12. , Temporarily stored in the buffer 13.

The frequency analysis unit 14 generates a spectrum pattern SP01 in which a plurality of frequency spectra for each frame are arranged by performing a fast Fourier transform (FFT) on the time series signal of the electronic melody input from the buffer 13 in a predetermined frame unit. do.

FIG. 2 is a diagram showing the relationship between the time-series signals I0 of the electronic melody and the frames F01 to F78. Assuming that the assumed electronic melody length is "Tm", the sampling frequency is "Fs", and the FFT length is "Nf", the number of frames is determined by the relational expression of Tm × Fs / Nf. As an example, in the present embodiment, Tm = 4.992 (sec), Fs = 8 (kHz), and Nf = 512 are set, and as a result, the number of frames is "78". The frame length is 4.992 / 78 = 64 (msec). The frames are arranged in chronological order, and as shown in FIG. 2, the head is the frame F01 and the end is the frame F78.

FIG. 3 is a diagram schematically showing an example of the spectrum pattern SP01 generated by the frequency analysis unit 14. By frequency-analyzing the time-series signal I0 of the electronic melody in frame units, frequency spectra X01 to X78 corresponding to each frame F01 to F78 are generated. The spectrum pattern SP01 is obtained by arranging the frequency spectra X01 to X78 in chronological order. With reference to FIG. 1, the spectrum pattern SP01 generated by the frequency analysis unit 14 is temporarily stored in the buffer 15.

The pattern generation unit 16 divides the spectrum pattern SP01 input from the buffer 15 into N segments (N is plural) each containing a plurality of frames, and a plurality of frames relating to the plurality of frames included in each segment. The spectrum pattern SP02 is generated by averaging the frequency spectra. The number N of the segments is set to an optimum value according to the required specific accuracy, the required time, and the like, and in the example of this embodiment, N = 6.

FIG. 4 is a diagram showing the relationship between the frame (first section) F01 to F78 and the segment (second section) S01 to S06. The segments are arranged in chronological order, and as shown in FIG. 4, the head is the segment S01 and the end is the segment S06. In the example of this embodiment, since the number of frames is "78" and the number of segments is "6", 78/6 = 13 frames are included in one segment. For example, the segment S01 includes frames F01 to F13, and the segment S02 includes frames F14 to F26.

FIG. 5 is a diagram schematically showing an example of the spectrum pattern SP02 generated by the pattern generation unit 16. By averaging the frequency spectra X01 to X78 included in the spectrum pattern SP01 in segment units, frequency spectra Y01 to Y06 corresponding to each segment S01 to S06 are generated, respectively. For example, the frequency spectrum Y01 is generated by averaging the frequency spectra X01 to X13 with respect to the frames F01 to F13 included in the segment S01. The spectrum pattern SP02 is obtained by arranging the frequency spectra Y01 to Y06 in chronological order. The pattern generation unit 16 averages the complex signal immediately after the FFT by the frequency analysis unit 14. That is, instead of averaging the complex signal after making it an absolute value as in a general averaging method, averaging is performed in the state of the complex signal before the absolute value, and then the absolute value is performed.

Next, the pattern generation unit 16 generates the spectrum pattern SP02A by quantizing the spectrum pattern SP02. This process is optional, and may be omitted if the storage capacity and calculation load of the electronic melody specifying device 1 are sufficient.

FIG. 6 is a diagram schematically showing an example of the spectrum pattern SP02A generated by the pattern generation unit 16. The frequency spectra Y01A to Y06A are generated by performing 1-bit quantization using a predetermined threshold value V1 on the frequency spectra Y01 to Y06 shown in FIG. That is, in the frequency spectra Y01 to Y06, the frequency components less than the threshold value V1 are all set to "0" in the frequency spectra Y01A to Y06A, and the frequency components having the threshold value V1 or more in the frequency spectra Y01 to Y06 are frequencies. All of the spectra Y01A to Y06A are set to "1". The spectrum pattern SP02A is obtained by arranging the frequency spectra Y01A to Y06A in chronological order.

With reference to FIG. 1, the storage unit 19 stores a plurality of spectral patterns SP03 relating to various known electronic melodies. The spectrum pattern SP03 is generated in advance by the same processing as the above-mentioned spectrum pattern SP02A (frequency analysis in frame units, averaging in segment units, 1-bit quantization using the threshold value V1).

The similarity calculation unit 17 calculates the similarity between the spectrum pattern SP02A generated by the pattern generation unit 16 and each spectrum pattern SP03 stored in the storage unit 19.

FIG. 7 is a diagram for explaining a similarity calculation method by the similarity calculation unit 17. In the example of this embodiment, the spectral patterns SP03 and SP02A are each divided into six segments S01 to S06. First, the similarity calculation unit 17 calculates the similarity K01 to K06 for each of the corresponding segments S01 to S06 between the spectral patterns SP03 and SP02A by using an arbitrary calculation formula for obtaining the similarity. Next, the representative values of the six similarity degrees K01 to K06 obtained for each of the segments S01 to S06 are calculated as the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A. As the representative value, for example, an average value can be used. In that case, the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A is (K01 + K02 + K03 + K04 + K05 + K06) / 6. In addition to the average value, the maximum value can be used as the representative value, and in that case, the maximum value among the similarity degrees K01 to K06 becomes the similarity degree between the spectrum pattern SP03 and the spectrum pattern SP02A. The similarity calculation unit 17 sequentially reads out all the spectrum patterns SP03 stored in the storage unit 19 to sequentially calculate the similarity between each spectrum pattern SP03 and the spectrum pattern SP02A.

With reference to FIG. 1, the specific processing unit 18 identifies the electronic melody acquired by the acquisition unit 10 based on the calculation result of the similarity by the similarity calculation unit 17. That is, among all the spectrum patterns SP03 stored in the storage unit 19, one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A is specified, and the electronic melody acquired by the acquisition unit 10 is the spectrum. It is specified that it is an electronic melody corresponding to the pattern SP03.

As described above, according to the electronic melody specifying device 1 according to the present embodiment, the frequency analysis unit 14 (first pattern generation unit) sets the time series signal I0 of the electronic melody acquired by the acquisition unit 10 into a frame (1st pattern generation unit). By frequency analysis in units of (first section), a spectrum pattern SP01 (first spectrum pattern) in which a plurality of frequency spectra X (first frequency spectra) are arranged for each frame is generated. Further, the pattern generation unit 16 (second pattern generation unit) averages a plurality of frequency spectra X relating to a plurality of frames included in each segment (second section) to obtain a frequency spectrum Y (for each segment). A spectrum pattern Y02A (second spectrum pattern) in which a plurality of second frequency spectra) are arranged is generated. Then, the similarity calculation unit 17 (calculation processing unit) determines the similarity between the spectrum pattern Y02A generated by the pattern generation unit 16 and each spectrum pattern SP03 (third spectrum pattern) stored in the storage unit 19. calculate. As a result, the amount of calculation is reduced as compared with the case where the similarity is calculated using the spectrum pattern SP01 before averaging, so that the processing load can be reduced. Further, since the spectrum pattern SP03 can be simplified, the storage capacity required for the storage unit 19 can be reduced.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the similarity calculation unit 17 obtains N pieces by calculating the similarity for each segment corresponding to the spectrum pattern SP02A and the spectrum pattern SP03. The representative value of the similarity is calculated as the similarity between the spectrum pattern SP02A and the spectrum pattern SP03. Therefore, as compared with the case where the similarity is calculated for each frame, the amount of calculation is reduced, so that the processing load can be reduced.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the similarity calculation unit 17 sets the average value of N similarities obtained by calculating the similarity for each corresponding segment as a spectrum pattern. It is calculated as the degree of similarity between SP02A and the spectrum pattern SP03. This makes it possible to calculate the similarity based on the distribution of frequency components in the entire region of the spectral pattern.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the similarity calculation unit 17 sets the maximum value of N similarities obtained by calculating the similarity for each corresponding segment as a spectrum pattern. It is calculated as the degree of similarity between SP02 and the spectrum pattern SP03. This makes it possible to calculate the similarity of the electronic melody including the characteristic portion based on the distribution of the frequency components in the frequency spectrum SP02 corresponding to the characteristic portion.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the pattern generation unit 16 averages the plurality of frequency spectra X to generate the frequency spectrum Y, and the complex before the absolute value in the fast Fourier transform. Perform averaging on the signal. As a result, the influence of white noise mixed in the frequency spectrum X is reduced in the frequency spectrum Y and the signal-to-noise ratio is improved, so that the accuracy of specifying the electronic melody can be improved.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the pattern generation unit 16 performs quantization using the threshold value V1 on the frequency spectra Y01 to Y06, thereby performing frequency spectra Y01A to Y06A. To generate. Since the spectrum pattern SP02 can be simplified by quantizing the frequency spectra Y01 to Y06, the amount of calculation can be reduced. Moreover, since the specific target is an electronic melody, even if the frequency spectra Y01 to Y06 are quantized, the effect of reducing the specific accuracy is small. Therefore, it is possible to reduce the amount of calculation while suppressing the deterioration of the specific accuracy of the electronic melody.

Further, according to the electronic melody specifying device 1 according to the present embodiment, the storage unit 19 stores the spectrum pattern SP03 that has been averaged and quantized in the same manner as the frequency spectra Y01A to Y06A. Since the spectral pattern SP03 can be further simplified by quantization, the storage capacity required for the storage unit 19 can be further reduced.

FIG. 8 is a diagram showing a configuration for realizing the electronic melody specifying device 1 by software. When the CPU 20 executes the program 50 read out to the RAM 21, each function of the electronic melody specifying device 1 described above is realized by software processing. In other words, the program 50 uses the computer mounted on the electronic melody specifying device 1 as an acquisition means for acquiring the time-series signal I0 of the electronic melody to be specified, and the frequency of the acquired time-series signal I0 in frame units. The first pattern generation means for generating the spectrum pattern SP01 by analysis, and the spectrum pattern SP02A by dividing the spectrum pattern SP01 into N segments and averaging a plurality of frequency spectra X included in each segment. A second pattern generation means for generating the spectrum pattern SP03, a storage means for storing the spectrum pattern SP03, a similarity calculation means for calculating the similarity between the spectrum pattern SP02A and the spectrum pattern SP03, and a similarity calculation means for calculating the similarity. Based on the result, it is a program for functioning as a specific processing means for specifying the electronic melody acquired by the acquisition means.

Hereinafter, various modifications of the electronic melody specifying device 1 according to the above embodiment will be described. The modifications described below can be applied in combination as appropriate.

<Modification example 1 (measures for improving specific accuracy)>
In the above embodiment, the frames F01 to F78 are evenly distributed in each of the segments S01 to S06 as shown in FIG. 4, but the number of frames included in each segment is based on the distribution of the frequency components in the spectrum pattern SP02A. May be unevenly distributed.

FIG. 9 is a diagram showing an example of the spectrum pattern SP02A generated by even distribution. According to the above example, the total number of frames is "78" and the number of segments is "6", so that the number of frames per segment when evenly distributed is "13".

The pattern generation unit 16 counts the number of frequency components having a threshold value V1 or more in each frequency spectrum X01 to X78 by referring to the spectrum pattern SP01 stored in the buffer 15. Then, assuming that the frequencies are evenly distributed, the number of frequency components included in each of the segments S01 to S06 is totaled for each of the segments S01 to S06. In the example shown in FIG. 9, the segments S01 to S06 are "2", "6", "6", "2", "4", and "6" in this order. The total number is "26".

Next, the pattern generation unit 16 calculates the ratio of the frequency components for each segment by dividing the number of frequency components for each segment by the total number of frequency components in all the segments. In the example shown in FIG. 9, the segments S01 to S06 are "1/13", "3/13", "3/13", "1/13", "2/13", and "3/13" in this order.

Next, the pattern generation unit 16 subtracts a value obtained by multiplying the total number of frames by the above-mentioned "ratio of frequency components for each segment" from a value twice the "number of frames per segment when evenly distributed". Calculates the number of frames for each segment in the case of non-equal distribution. In the example shown in FIG. 9, the segments S01 to S06 are "20", "8", "8", "20", "14", and "8" in this order.

FIG. 10 is a diagram showing the relationship between the frames F01 to F78 and the segments S01 to S06 in the case of non-uniform distribution. As shown in FIG. 10, the segment S01 includes 20 frames F01 to F20, the segment S02 contains 8 frames F21 to F28, and the segment S03 contains 8 frames F29 to F36. S04 includes 20 frames F37 to F56, segment S05 includes 14 frames F57 to F70, and segment S06 contains 8 frames F71 to F78.

FIG. 11 is a diagram showing another example of the spectral pattern SP02A generated by even distribution. The frequency spectra Y01A to Y06A include one frequency component having a threshold value V1 or higher. Even when the pattern generation unit 16 has a non-uniform distribution function, if the electronic melody having an even frequency component distribution as shown in FIG. 11 is a specific target, the pattern generation unit 16 may be used. It will be evenly distributed.

Similarly to the above, when the spectrum pattern SP03 is generated, the number of frames included in each segment is unevenly distributed based on the frequency component distribution in the spectrum pattern SP02A when the spectrum pattern SP02A is evenly distributed.

FIG. 12 is a diagram showing a spectrum pattern SP03 stored in the storage unit 19. The spectrum pattern SP03 is classified into a non-uniform group G1 that aggregates the spectrum pattern SP03 generated by the non-uniform distribution and an equal group G2 that aggregates the spectrum pattern SP03 generated by the uniform distribution, and is stored in the storage unit 19. Will be done. When the reference electronic melody has a non-uniform frequency component distribution as shown in FIG. 9, since the non-uniform distribution is applied, the spectrum pattern SP03 related to the electronic melody is classified into the non-uniform group G1. .. On the other hand, when the reference electronic melody has an even frequency component distribution as shown in FIG. 11, since the equal distribution is applied, the spectrum pattern SP03 related to the electronic melody is classified into the equal group G2.

FIG. 13 is a flowchart showing a first example regarding the processing flow of the similarity calculation unit 17 and the specific processing unit 18.

First, in step R101, the similarity calculation unit 17 determines whether or not the spectrum pattern SP02A input from the pattern generation unit 16 has a non-uniform distribution.

In the case of non-uniform distribution, the similarity calculation unit 17 then searches for the non-uniform group G1 in step R102 to sequentially determine the similarity between each spectrum pattern SP03 and the spectrum pattern SP02A belonging to the non-uniform group G1. calculate. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the non-uniform group G1. Here, it is assumed that the spectrum pattern E1 is selected.

Next, in step R103, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E1.

If the result of the determination in step R101 is even distribution, then in step R104, the similarity calculation unit 17 searches for the equality group G2 to obtain the spectral patterns SP03 and the spectral patterns SP02A belonging to the equality group G2. Calculate the degree of similarity in order. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the equality group G2. Here, it is assumed that the spectrum pattern E2 is selected.

Next, in step R105, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E2.

FIG. 14 is a flowchart showing a second example regarding the processing flow of the similarity calculation unit 17 and the specific processing unit 18.

First, in step R201, the similarity calculation unit 17 determines whether or not the spectrum pattern SP02A input from the pattern generation unit 16 has a non-uniform distribution.

In the case of non-uniform distribution, the similarity calculation unit 17 then searches for the non-uniform group G1 in step R202 to sequentially determine the similarity between each spectrum pattern SP03 and the spectrum pattern SP02A belonging to the non-uniform group G1. calculate. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the non-uniform group G1. Here, it is assumed that the spectrum pattern E1 is selected.

Next, in step R203, the specific processing unit 18 determines whether or not the similarity e1 between the spectrum pattern E1 and the spectrum pattern SP02A is equal to or higher than the predetermined threshold value V2.

When the similarity e1 is equal to or higher than the threshold value V2, then in step R204, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E1.

If the similarity e1 is less than the threshold value V2 as a result of the determination in step R203, then in step R205, the similarity calculation unit 17 searches for the equality group G2 to search for each spectral pattern belonging to the equality group G2. The degree of similarity between SP03 and the spectral pattern SP02A is calculated in order. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the equality group G2. Here, it is assumed that the spectrum pattern E2 is selected.

Next, in step R206, the specific processing unit 18 determines whether or not the similarity e2 between the spectrum pattern E2 and the spectrum pattern SP02A exceeds the similarity e1.

If the similarity e2 exceeds the similarity e1, then in step R207, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E2. On the other hand, when the similarity e2 is equal to or less than the similarity e1, then in step R208, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E1.

If the result of the determination in step R201 is even distribution, then in step R209, the similarity calculation unit 17 searches for the equality group G2 to obtain the spectral patterns SP03 and the spectral patterns SP02A belonging to the equality group G2. Calculate the degree of similarity in order. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the equality group G2. Here, it is assumed that the spectrum pattern E3 is selected.

Next, in step R210, the specific processing unit 18 determines whether or not the similarity e3 between the spectrum pattern E3 and the spectrum pattern SP02A is equal to or higher than the predetermined threshold value V2.

When the similarity e3 is equal to or higher than the threshold value V2, the specific processing unit 18 next in step R211 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E3.

If the similarity e3 is less than the threshold value V2 as a result of the determination in step R210, then in step R212, the similarity calculation unit 17 searches for the non-uniform group G1 to belong to each non-uniform group G1. The degree of similarity between the spectrum pattern SP03 and the spectrum pattern SP02A is calculated in order. Then, the specific processing unit 18 selects one spectrum pattern SP03 having the maximum similarity with the spectrum pattern SP02A from all the spectrum patterns SP03 belonging to the non-uniform group G1. Here, it is assumed that the spectrum pattern E4 is selected.

Next, in step R213, the specific processing unit 18 determines whether or not the similarity e4 between the spectrum pattern E4 and the spectrum pattern SP02A exceeds the similarity e3.

If the similarity e4 exceeds the similarity e3, then in step R214, the specific processing unit 18 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E4. On the other hand, when the similarity e4 is equal to or less than the similarity e3, the specific processing unit 18 next in step R215 identifies that the electronic melody acquired by the acquisition unit 10 is an electronic melody corresponding to the spectrum pattern E3.

According to the electronic melody specifying device 1 according to the present modification, the pattern generation unit 16 can non-uniformly distribute the number of frames included in each segment based on the distribution of the frequency components in the spectrum pattern SP02A. Since the resolution can be adjusted according to the number of frames included in the segment, the electronic melody is distributed so as to increase the resolution of important parts and decrease the resolution of non-important parts while avoiding an increase in processing load. It is possible to improve the specific accuracy of.

Further, according to the electronic melody specifying device 1 according to the present modification, the pattern generation unit 16 allocates a relatively small number of frames to the segment corresponding to the frequency spectrum Y including a relatively large number of frequency components. A relatively large number of frames are allocated to the segment corresponding to the frequency spectrum Y containing a relatively small number of frequency components. Since the frequency spectrum Y containing a relatively large number of frequency components is an important part for identifying an electronic melody, a relatively small number of frames should be allocated to the segments corresponding to such a frequency spectrum Y. As a result, the resolution can be increased, and as a result, the accuracy of specifying the electronic melody can be improved. On the other hand, since the frequency spectrum Y containing a relatively small number of frequency components is not important for identifying the electronic melody, a relatively large number of frames are allocated to the segments corresponding to such a frequency spectrum Y. By doing so, it becomes possible to avoid an increase in the processing load.

Further, according to the electronic melody specifying device 1 according to the present modification, the similarity calculation unit 17 has a non-uniform group G1 with respect to the spectrum pattern SP02A in which the number of frames included in each segment is non-uniformly distributed. The similarity with the spectrum pattern SP03 of (first group) is calculated, while the spectrum pattern SP02A in which the number of frames included in each segment is evenly distributed is the same group G2 (second group). The degree of similarity with the spectrum pattern SP03 is calculated. In this way, the spectrum pattern SP03 is classified into the non-uniform group G1 and the uniform group G2 in advance, and the spectrum pattern SP03 to be compared is selected according to the type of the spectrum pattern SP02A. It is possible to reduce the processing load and the required time while suppressing the decrease.

Further, according to the electronic melody specifying device 1 according to the present modification, the similarity calculation unit 17 is in the non-uniform group G1 with respect to the spectrum pattern SP02A in which the number of frames included in each segment is non-uniformly distributed. When the similarity with the spectrum pattern SP03 is less than the threshold value V2, the similarity between the spectrum pattern SP02A and the spectrum pattern SP03 of the equal group G2 is further calculated. Further, the similarity calculation unit 17 determines that the similarity with the spectrum pattern SP03 of the equality group G2 is less than the threshold value V2 with respect to the spectrum pattern SP02A in which the number of frames included in each segment is evenly distributed. Further, the similarity between the spectrum pattern SP02A and the spectrum pattern SP03 of the non-uniform group G1 is calculated. As described above, when the similarity with the spectrum pattern SP03 of one group of the non-uniform group G1 and the equal group G2 is less than the threshold value V2, the similarity with the spectrum pattern SP03 of the other group is calculated. As a result, the spectrum pattern SP03 whose similarity is equal to or higher than the threshold value V2 may be discovered, so that the accuracy of specifying the electronic melody can be improved.

In the above description, an example of allocating the number of frames included in each segment is described on the premise that the number of segments is a fixed value in order to avoid an increase in the amount of calculation, but an increase in the amount of calculation is acceptable. In that case, the accuracy of specifying the electronic melody can be improved by dividing the segment corresponding to the important part into a plurality of segments (the number of segments increases).

<Transformation example 2 (positional deviation countermeasure 1)>
When the acquisition unit 10 acquires the time-series signal I0 of the electronic melody by a manual recording operation by the user, the spectrum pattern SP02A and the spectrum pattern SP03 correspond to each other due to the difference between the start timing and the end timing of the recording operation. Segment misalignment can occur. In the following modification examples 2 to 7, even when the misalignment of the segment occurs, measures for reducing the influence caused by the misalignment will be described.

FIG. 15 is a diagram for explaining a first example of measures against misalignment of corresponding segments. First, the similarity calculation unit 17 calculates the similarity between the spectrum pattern SP03 and the reference spectrum pattern SP02A, using the spectrum pattern SP02A generated based on the spectrum pattern SP01 as the reference spectrum pattern, as in the above embodiment.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02B1 by cyclically shifting the reference spectrum pattern SP02A forward by one segment (the segment overflowing forward is circulated at the end). Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02B1 is calculated.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02B2 by cyclically shifting the pseudo-spectral pattern SP02B1 forward by one segment. Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02B2 is calculated.

Hereinafter, similarly, the similarity calculation unit 17 defines the pseudo-spectral patterns SP02B3 to SP02B5 in order by cyclically shifting forward one segment at a time, and calculates the similarity between the spectrum pattern SP03 and the pseudo-spectral patterns SP02B3 to SP02B5 in order. ..

Then, the maximum similarity among the total N (6 in this example) similarity calculated in this way is determined as the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A.

According to the electronic melody specifying device 1 according to this modification, the similarity calculation unit 17 cyclically shifts the arrangement order of the reference spectrum pattern SP02A and the N segments S01 to S06 included in the reference spectrum pattern SP02A. For each of the obtained (N-1) pseudo-spectral patterns SP02B1 to SP02B5, the degree of similarity with the spectrum pattern SP03 is calculated. In addition to the reference spectrum pattern SP02A, (N-1) pseudo-spectral patterns SP02B1 to SP02B5 that are cyclically shifted by one segment over the entire section are defined, so that even if the segment misalignment is large. , The degree of similarity between the spectrum pattern SP02A and the spectrum pattern SP03 can be appropriately calculated.

<Transformation example 3 (positional deviation countermeasure 2)>
FIG. 16 is a diagram for explaining a second example of measures against misalignment of the corresponding segment. First, the similarity calculation unit 17 calculates the similarity between the spectrum pattern SP03 and the reference spectrum pattern SP02A, using the spectrum pattern SP02A generated based on the spectrum pattern SP01 as the reference spectrum pattern, as in the above embodiment.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02C1 by cyclically shifting the reference spectrum pattern SP02A forward by one segment. Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02C1 is calculated.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02C2 by cyclically shifting the reference spectrum pattern SP02A backward by one segment. Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02C2 is calculated.

Then, the maximum similarity among the total of three similarity calculated in this way is determined as the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A.

According to the electronic melody specifying device 1 according to this modification, the similarity calculation unit 17 arranges the reference spectrum pattern SP02A and the N segments S01 to S06 included in the reference spectrum pattern SP02A in the front-back direction by one segment. For each of the two pseudo-spectral patterns SP02C1 and SP02C2 obtained by cyclically shifting, the degree of similarity with the spectral pattern SP03 is calculated. In addition to the reference spectrum pattern SP02A, two pseudo-spectral patterns SP02C1 and SP02C2 that are cyclically shifted one segment at a time in the front-back direction are defined, so that the processing load is increased as compared with the circulation shift of all sections (FIG. 15). It is possible to appropriately calculate the degree of similarity between the spectrum pattern SP02A and the spectrum pattern SP03 while suppressing the above.

<Transformation example 4 (positional deviation countermeasure 3)>
FIG. 17 is a diagram for explaining a third example of measures against misalignment of the corresponding segment. First, the similarity calculation unit 17 excludes the first and last segments S01 and S06 from the N segments S01 to S06 included in the spectrum pattern SP02A generated based on the spectrum pattern SP01 in chronological order. Thereby, a pseudo-spectral pattern including (N-2) segments S02 to S05 is defined.

Next, the similarity calculation unit 17 is similar to the spectrum pattern SP03 with respect to each of the three pseudo-spectral patterns SP02D1 to SP02D3 obtained by shifting the pseudo-spectral pattern backward by one segment in order from the beginning of the spectrum pattern SP03. Calculate the degrees in order.

Then, the maximum similarity among the total of three similarity calculated in this way is determined as the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A.

According to the electronic melody specifying device 1 according to this modification, the similarity calculation unit 17 has the first and last segments S01 and S06 of the N segments S01 to S06 included in the spectrum pattern SP02A in chronological order. By excluding (N-2), a pseudo-spectral pattern including (N-2) segments S02 to S05 is defined. Then, the similarity with the spectrum pattern SP03 is calculated for each of the three pseudo spectrum patterns SP02D1 to SP02D3 obtained by shifting the pseudo-spectral pattern backward by one segment in order from the beginning of the spectrum pattern SP03. Since the pseudo-spectral pattern is defined by excluding the first and last segments S01 and S06 in chronological order from the N segments S01 to S06 included in the spectrum pattern SP02A, the cyclic shift of the entire section (Fig.) Compared with 15), it is possible to suppress the increase in the processing load and eliminate the influence of the segments S01 and S06 at the beginning and the end where the timing deviation of the recording operation by the user is most likely to be reflected.

<Deformation example 5 (positional deviation countermeasure 4)>
FIG. 18 is a diagram for explaining a fourth example of measures against misalignment of the corresponding segment. First, the similarity calculation unit 17 calculates the similarity between the spectrum pattern SP03 and the reference spectrum pattern SP02A, using the spectrum pattern SP02A generated based on the spectrum pattern SP01 as the reference spectrum pattern, as in the above embodiment.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02E1 by shifting out the reference spectrum pattern SP02A forward by one segment (the segment overflowing forward is deleted). Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02E1 is calculated.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02E2 by shifting out the pseudo-spectral pattern SP02E1 forward by one segment. Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02E2 is calculated.

Hereinafter, similarly, the similarity calculation unit 17 defines the pseudo-spectral patterns SP02E3 to SP02E5 in order by shifting out one segment at a time, and calculates the similarity between the spectrum pattern SP03 and the pseudo-spectral patterns SP02E3 to SP02E5 in order. ..

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02E6 by shifting out the reference spectrum pattern SP02A by one segment to the rear (the segment overflowing to the rear is deleted). Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02E6 is calculated.

Next, the similarity calculation unit 17 defines the pseudo-spectral pattern SP02E7 by shifting out the pseudo-spectral pattern SP02E6 by one segment backward. Then, the degree of similarity between the spectrum pattern SP03 and the pseudo-spectral pattern SP02E7 is calculated.

Hereinafter, similarly, the similarity calculation unit 17 defines the pseudo-spectral patterns SP02E8 to SP02E10 in order by shifting out one segment at a time, and calculates the similarity between the spectrum pattern SP03 and the pseudo-spectral patterns SP02E8 to SP02E10 in order. ..

Then, the maximum similarity among the total 2N-1 similarity calculated in this way (11 in this example) is determined as the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A.

<Deformation example 6 (positional deviation countermeasure 5)>
FIG. 19 is a diagram for explaining a fifth example of measures against misalignment of the corresponding segment. As described in the above embodiment, in order to calculate the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A, the similarity calculation unit 17 first, first, for each of the corresponding segments S01 to S06 between the spectrum patterns SP03 and SP02A. The similarity degrees K01 to K06 are calculated. Next, the representative values of the N similarity degrees K01 to K06 obtained for each of the segments S01 to S06 are calculated as the degree of similarity between the spectrum pattern SP03 and the spectrum pattern SP02A.

At that time, the similarity calculation unit 17 according to this modification sets the calculated similarity K01 and K06 for the first and last segments S01 and S06 in chronological order among the N segments S01 to S06. Multiply each weight factor W (eg 0.5) less than 1. Therefore, when the average value is used as the representative value, the similarity between the spectrum pattern SP03 and the spectrum pattern SP02A is (W × K01 + K02 + K03 + K04 + K05 + W × K06) / 6.

According to the electronic melody specifying device 1 according to this modification, the similarity calculation unit 17 has the first and last segments S01 in chronological order among the N segments S01 to S06 included in the spectrum pattern SP02A. For S06, the calculated similarity K01 and K06 are multiplied by a weighting coefficient W less than 1. Therefore, it is possible to reduce the influence of the segments S01 and S06 at the beginning and the end where the timing deviation of the recording operation by the user is most likely to be reflected.

<Deformation example 7 (positional deviation countermeasure 6)>
FIG. 20 is a diagram for explaining a sixth example of measures against misalignment of the corresponding segment. In the above embodiment, the frames F01 to F78 are evenly distributed to the segments S01 to S06 as shown in FIG. 4, but the pattern generation unit 16 according to this modification is the N segments included in the spectrum pattern SP02A. Of S01 to S06, with respect to the head portion and end portion segments S01 and S06 in chronological order, a smaller number of frames are allocated than the number of frames allocated to the central portion segments S02 to S05. In the example shown in FIG. 20, 9 frames are allocated to each of the head portion and the tail portion segments S01 and S06, and 15 frames are allocated to each of the central portion segments S02 to S05.

According to the electronic melody specifying device 1 according to this modification, the pattern generation unit 16 has the first and last segments S01 and S06 of the N segments S01 to S06 included in the spectrum pattern SP02A in chronological order. With respect to, a smaller number of frames are allocated than the number of frames allocated to the segments S02 to S05 in the central portion. Therefore, it is possible to reduce the influence of the segment at the beginning and the end where the timing deviation of the recording operation by the user is most likely to be reflected.

1 Electronic melody identification device 10 Acquisition unit 14 Frequency analysis unit 16 Pattern generation unit 17 Similarity calculation unit 18 Specific processing unit 19 Storage unit 50 Program

Claims (16)

An acquisition unit that acquires the time-series signal of the electronic melody that is the specific target,
By frequency-analyzing the time-series signal of the electronic melody acquired by the acquisition unit in a predetermined first section unit, a first spectrum pattern in which a plurality of first frequency spectra for each first section are arranged is generated. The first pattern generator and
The first spectrum pattern generated by the first pattern generation unit is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and is included in each second section. A second pattern generation unit that generates a second spectrum pattern in which a plurality of second frequency spectra for each second section are arranged by averaging a plurality of first frequency spectra related to a plurality of first sections.
A storage unit that stores a third spectral pattern, which is generated in advance for each of a plurality of known electronic melodies and is divided into N second sections.
A calculation processing unit that calculates the degree of similarity between the second spectrum pattern generated by the second pattern generation unit and each third spectrum pattern stored in the storage unit.
Based on the calculation result of the similarity by the calculation processing unit, the specific processing unit that specifies the electronic melody acquired by the acquisition unit and the specific processing unit.
Equipped with
The second pattern generation unit is an electronic melody specifying device capable of unequally distributing the number of first sections included in each second section based on the distribution of frequency components in the second spectral pattern. The second pattern generation unit allocates the first section of the first number evenly to the second section corresponding to the second frequency spectrum including the first number of frequency components when evenly distributed . Allocate the first section of the second number more than the first number to the second section corresponding to the second frequency spectrum containing the second number of frequency components less than the first number when allocated. , The electronic melody specifying device according to claim 1. The storage unit includes the third spectrum pattern of the first group in which the number of the first sections included in each of the N second sections is unevenly distributed, and each of the N second sections. The third spectrum pattern of the second group in which the number of the first sections to be formed is evenly distributed is stored.
The calculation processing unit
In the second spectrum pattern generated by the second pattern generation unit, when the number of the first sections included in each second section is unevenly distributed, the second spectrum pattern and the storage Calculate the similarity with the third spectral pattern of the first group stored in the unit,
In the second spectrum pattern generated by the second pattern generation unit, when the number of the first sections included in each second section is evenly distributed, the second spectrum pattern and the storage unit are used. The electronic melody specifying device according to claim 1 or 2, which calculates the degree of similarity with the third spectrum pattern of the second group stored in. The calculation processing unit
When the similarity between the second spectrum pattern generated by the second pattern generation unit and the third spectrum pattern of the first group stored in the storage unit is less than a predetermined threshold value. Further, the degree of similarity between the second spectrum pattern and the third spectrum pattern of the second group stored in the storage unit is calculated.
When the similarity between the second spectrum pattern generated by the second pattern generation unit and the third spectrum pattern of the second group stored in the storage unit is less than a predetermined threshold value. The electronic melody specifying device according to claim 3, further comprising calculating the degree of similarity between the second spectrum pattern and the third spectrum pattern of the first group stored in the storage unit. The calculation processing unit calculates the representative values of N similarities obtained by calculating the similarity for each corresponding second section as the similarity between the second spectrum pattern and the third spectrum pattern. Item 4. The electronic melody specifying device according to any one of Items 1 to 4. The electronic melody specifying device according to claim 5, wherein the representative value is an average value. The electronic melody specifying device according to claim 5, wherein the representative value is the maximum value. The calculation processing unit
A reference second spectrum pattern, which is a second spectrum pattern generated based on the first spectrum pattern, and
(N-1) pseudo-second spectral patterns obtained by cyclically shifting the arrangement order of N second intervals included in the reference second spectral pattern, and
The electronic melody specifying device according to any one of claims 5 to 7, which calculates the similarity with the third spectrum pattern for each of the above. The calculation processing unit
A reference second spectrum pattern, which is a second spectrum pattern generated based on the first spectrum pattern, and
Two pseudo second spectral patterns obtained by cyclically shifting the arrangement order of N second segments included in the reference second spectral pattern by one interval in the front-back direction, and
The electronic melody specifying device according to any one of claims 5 to 7, which calculates the similarity with the third spectrum pattern for each of the above. The calculation processing unit
By excluding the second section of the first part and the last part in chronological order from the N second sections included in the second spectrum pattern generated based on the first spectrum pattern, (N-2) pieces. A pseudo second spectral pattern including the second interval of
Claims 5 to 7 calculate the similarity with the third spectrum pattern for each of the three pseudo second spectrum patterns obtained by shifting the pseudo second spectrum pattern backward by one section in order from the beginning. The electronic melody specifying device described in any one of them. The calculation processing unit multiplies the calculated similarity by a weighting coefficient of less than 1 for the second section of the head portion and the tail portion in chronological order among the N second sections included in the second spectrum pattern. The electronic melody specifying device according to any one of claims 5 to 7. The first pattern generation unit generates a first frequency spectrum by frequency analysis using a fast Fourier transform.
The second pattern generation unit averages a plurality of first frequency spectra to generate a second frequency spectrum, and averages the complex signal before the absolute value in the fast Fourier transform. The electronic melody specifying device according to any one of 11 to 11. The second pattern generation unit generates a second frequency spectrum by performing quantization using a predetermined threshold value on the frequency spectrum obtained by averaging a plurality of first frequency spectra. The electronic melody specifying device according to any one of claims 1 to 12. The electronic melody specifying device according to claim 13, wherein the storage unit stores a third spectrum pattern that has been averaged and quantized in the same manner as the second frequency spectrum. The computer installed in the electronic melody identification device,
An acquisition method for acquiring a time-series signal of an electronic melody that is a specific target,
By frequency-analyzing the time-series signal of the electronic melody acquired by the acquisition means in a predetermined first section unit, a first spectrum pattern in which a plurality of first frequency spectra for each first section are arranged is generated. The first pattern generation means and
The first spectrum pattern generated by the first pattern generation means is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and is included in each second section. A second pattern generation means for generating a second spectrum pattern in which a plurality of second frequency spectra for each second section are arranged by averaging a plurality of first frequency spectra for a plurality of first sections.
A storage means for storing a third spectral pattern, which is pregenerated for each of a plurality of known electronic melodies and divided into N second sections.
A calculation processing means for calculating the similarity between the second spectrum pattern generated by the second pattern generation means and each third spectrum pattern stored in the storage means.
Specific processing means for specifying the electronic melody acquired by the acquisition means based on the calculation result of the similarity by the calculation processing means, and
It is a program to function as
The second pattern generation means is a program capable of unequally distributing the number of first sections included in each second section based on the distribution of frequency components in the second spectral pattern. (A) A step of acquiring a time-series signal of an electronic melody to be specified, and
(B) The first spectrum in which a plurality of first frequency spectra for each first section are arranged by frequency-analyzing the time-series signal of the electronic melody acquired in the step (A) in a predetermined first section unit. The steps to generate the pattern and
(C) The first spectral pattern generated by the step (B) is divided into N (multiple N) second sections, each of which contains a plurality of first sections, and is included in each second section. A step of generating a second spectrum pattern in which a plurality of second frequency spectra for each second section are arranged by averaging a plurality of first frequency spectra related to the plurality of first sections.
(D) A step of storing a third spectral pattern, which is pregenerated for each of a plurality of known electronic melodies and divided into N second sections.
(E) A step of calculating the similarity between the second spectrum pattern generated by the step (C) and each third spectrum pattern stored by the step (D).
(F) A step of specifying the electronic melody acquired in the step (A) based on the calculation result of the similarity in the step (E), and
Equipped with
In step (C), an electronic melody specifying method capable of unequally distributing the number of first sections included in each second section based on the distribution of frequency components in the second spectral pattern.

Images