JP5982671B2 - Audio signal processing method and audio signal processing system - Google Patents

Description

  The present invention relates to an audio signal processing method executed by a system that executes audio signal processing on input audio data, and more particularly to an audio signal processing method for concealing personality.

  In recent years, due to advances in voice data storage technology that stores voice data in the storage area and voice data search technology that searches for voice data, a large amount of voice data is stored in the storage area and the voice data stored in the storage area is utilized. There are many scenes to do. In particular, in medical settings, call centers, and web contents (for example, Facebook (registered trademark, the same applies hereinafter) and YOUTUBE (registered trademark, the same applies hereinafter)), personal information and personality information included in accumulated audio data Is required to be treated carefully.

  Here, the personal information is information that can identify an individual such as a name, an address, and a telephone number included in the contents of the audio data. For example, the content “A-san is illness B. All day C” included in the audio data recorded by the nurse speaks personal information. In addition, the content of the operator's voice data recorded at the call center “Person E residing in D! Change the PIN to F” corresponds to personal information. In order to protect personal information, a technique for automatically detecting a preset keyword and replacing the detected keyword with another word is known.

  On the other hand, the personality information refers to voice information that can identify an individual among information included in the voice of the voice data (sound that includes a human voice). For example, the personality information is information such as the pitch, the volume, the volume, and the vocal tract shape included in the speech waveform of the speech data. It is well known that an individual can be identified using voiceprint information which is one of personality information.

  Therefore, personality information is an individual right such as a portrait right, and personality information should be protected because there is a risk of misuse if personality information is unknowingly known to others. However, when calling a call center or uploading voice data to a server, the voice of a speaker may be heard by an unspecified number of people, so control of personality information is required.

  There are various types of personality information (for example, personality information included in utterance content, personality information included in prosody, and personality information included in phoneme).

  The personality information included in the utterance content includes, for example, utterance 癖 (filler usage frequency such as “Ano” and “Eto”), culture (grammar, selection of words, etc.), and dialect (“Dabe”). Etc.).

  Personality information included in the prosody includes gender (such as voice pitch), dialect and foreigner (such as accent), emotion (such as intonation), and physical condition (such as vocal cord noise and nasal sound).

  The personality information included in the phoneme includes body features (such as vocal tract shape), speech utterances (good tongue, how to open the mouth, how to put out the tongue), and recording environment (such as background sound and reverberation). is there.

  As a control technique for personal information, a voice changer that raises voice is known. The voice changer is used, for example, to make it difficult to specify a performer on a television or the like. In addition to the voice changer, a technique for converting a female voice into a male voice is also known as a technique for controlling personality information (see, for example, Patent Document 1).

JP-A-10-240292

  However, since the voice changer only raises the voice, personality information cannot be removed. Similarly, since the technique described in Patent Document 1 only changes the pitch of the voice, other personality information cannot be removed. For example, an individual may be identified based on personality information such as speech, voiceprint, voice quality, and speech style.

  In addition, it may be possible to remove all personality information uniformly, but the sound quality of the audio data deteriorates due to the process for removing the personality information, and the audibility of the audio data is significantly reduced. The use value of data is lost.

  In view of the above, an object of the present invention is to provide an audio signal processing method that removes personality information with strong individuality and prevents deterioration of audio data as much as possible.

  In a typical example of the present invention, in an audio signal processing method executed on input audio data, the method is executed by a system including a CPU, a storage area, and an interface. A feature amount calculation step for calculating at least one feature amount obtained by quantifying the feature amount, and a personality degree obtained by quantifying the strength of the individuality of the feature amount based on the feature amount calculated in the feature amount difference calculation step The personality degree calculating step for calculating the personality degree, and if there is a personality degree that does not satisfy the predetermined personality degree condition, the audio signal processing is performed so that the individuality degree satisfies the personality degree condition. An audio signal processing step to be performed on the data.

  The effects obtained by the representative ones of the inventions disclosed in the present application will be briefly described as follows. That is, it is possible to remove personal information having strong personality and prevent deterioration of audio data as much as possible.

It is a hardware block diagram of the audio | voice signal processing system of 1st Embodiment of this invention. 1 is a functional block diagram of an audio signal processing system according to a first embodiment of the present invention. It is a functional block diagram of the voice individuality characteristic analysis part of 1st Embodiment of this invention. It is explanatory drawing of the audio | voice waveform of the audio | voice data of 1st Embodiment of this invention. It is explanatory drawing of the language feature-value of 1st Embodiment of this invention. It is explanatory drawing of the time series data of the power of the speech waveform calculated as a prosodic feature-value of 1st Embodiment of this invention. It is explanatory drawing of the time series data of the fundamental frequency of the audio | voice waveform of the audio | voice data calculated as the prosodic feature-value of 1st Embodiment of this invention. It is explanatory drawing of the time series data of the phoneme continuation length of the audio | voice waveform of the audio | voice data calculated as a prosodic feature-value of 1st Embodiment of this invention. It is explanatory drawing of the phoneme spectrum calculated as a phoneme feature-value of 1st Embodiment of this invention. It is explanatory drawing of the individuality feature-value list | wrist of 1st Embodiment of this invention. It is explanatory drawing of the individuality degree calculation model corresponding to the average F0 value of 1st Embodiment of this invention. It is a functional block diagram of the audio | voice signal processing parameter determination part of 1st Embodiment of this invention. It is explanatory drawing of the individuality degree condition of 1st Embodiment of this invention. It is explanatory drawing of the audio | voice signal processing parameter of 1st Embodiment of this invention. It is explanatory drawing of the audio | voice signal processing parameter adjustment method of 1st Embodiment of this invention. It is a functional block diagram of the audio | voice signal processing part of 1st Embodiment of this invention. It is time series data of the fundamental frequency before the average F0 value of 1st Embodiment of this invention is adjusted. It is time series data of the fundamental frequency after the average F0 value of 1st Embodiment of this invention is adjusted. It is explanatory drawing of a structure of the information processing system of 1st Embodiment of this invention. It is a functional block diagram of the server of 2nd Embodiment of this invention. It is a functional block diagram of the audio | voice signal processing parameter determination part of 2nd Embodiment of this invention. It is explanatory drawing of the specific example of the parameter adjustment completion conditions of 2nd Embodiment of this invention. It is a graph which shows the relationship between the sound quality degradation of 2nd Embodiment of this invention, and the individuality degree change of an average F0 value. It is explanatory drawing of the specific example of the individuality authority setting rule of 2nd Embodiment of this invention. It is a functional block diagram of the audio | voice signal processing parameter determination part of 3rd Embodiment of this invention. It is a graph which shows the relationship between the delay time of 3rd Embodiment of this invention, and the variation | change_quantity of the audio | voice data by audio | voice signal processing.

  First, the outline of the present invention will be described.

  In the present invention, a personality feature value obtained by quantifying the input voice data is calculated, and a personality degree obtained by quantifying the strength of the personality of the calculated personality feature value is calculated. If the individuality degree does not satisfy the individuality degree condition, audio signal processing is executed such that the individuality degree satisfies the individuality condition. As a result, personality information with strong personality can be removed, and deterioration of audio data can be prevented as much as possible.

  Analyzing the voice data of a performer imitating a celebrity, it was found that the physical features of the performer's voice data are not similar to the celebrity's target celebrity. In other words, when a person identifies a speaker from speech, utterance style, rhythm, intonation, and the like are more important than vocal tract features derived from the speaker's physical characteristics. Furthermore, when various audio data were analyzed, it was found that the physical feature amount (personality degree feature amount) of the voice individuality feature (referring to a cue for identifying an individual. The same meaning as personality information) differs depending on the speaker. found.

  In the prior art, since the same audio signal processing is performed for all speakers, the personality feature cannot be completely removed. For example, by performing speech signal processing to remove personality features for all physical parameters (prosodic features, phoneme features, linguistic features, etc.) that can be personality features, Even if it can be completely removed, with the current voice processing technology, if a large amount of signal processing is performed on the voice data, the voice deteriorates and the user cannot hear the voice, which is not practical.

  Therefore, in this patent, the individuality degree that can quantitatively evaluate the individuality strength of the individuality feature quantity is defined, and the physical parameters of the audio data are set so that the individuality degree satisfies the predetermined individuality degree condition. On the other hand, by executing the audio signal processing, it is possible to remove a personality feature having a personality strength of a predetermined value or more.

Exemplary embodiments of a speech synthesizer and a speech synthesis method according to the present invention will be explained below in detail with reference to the drawings.
(First embodiment)
A first embodiment of the present invention will be described with reference to FIGS.

  FIG. 1 is a hardware configuration diagram of an audio signal processing system 100 according to the first embodiment of this invention.

  The audio signal processing system 100 includes a CPU 103, a memory 104 as a main storage device, an auxiliary storage device 101, an audio input interface (I / F) 102, and an audio output interface (I / F) 105. The CPU 103, the main memory device 104, the auxiliary storage device 101, the audio input I / F 102, and the audio output I / F 105 are connected to each other via a bus 106.

  For example, the audio signal processing system 100 is incorporated as an audio processing unit in a computer such as a mobile phone and a computer. In this case, the hardware configuration shown in FIG. 1 may be realized by using hardware of a computer in which the audio signal processing system 1000 is incorporated, or provided separately from the computer in which the audio signal processing system 1000 is incorporated. May be.

  The CPU 103 governs overall control of the audio signal processing system 100. The memory 104 is used as a work area for the CPU 103. The auxiliary storage device 101 is a non-volatile storage medium, such as an HDD (Hard Disk Drive), an FD (Flexible Disk), and a flash memory. The auxiliary storage device 101 stores various programs such as a program for calculating a personality feature amount, a program for calculating a personality degree, and various data.

  The audio input I / F 102 is an interface for inputting audio data to the audio signal processing system 100. A voice input device such as a microphone may be directly connected to the voice input I / F 102, and voice data may be directly input to the voice signal processing system 100 from the voice input device. In addition, audio data may be input to the audio signal processing system 100 from a storage device such as a portable storage medium. Further, the audio data may be input to the audio signal processing system 100 by receiving the audio data via a network.

  The audio output I / F 105 is an interface for outputting audio data from the audio signal processing system 100. An audio output device such as a speaker may be directly connected to the audio output I / F 105, and audio data may be output from the audio output device. Also, the audio signal processing system 100 may output audio data to a storage device such as a portable storage medium. Further, the audio data may be output from the audio signal processing system 100 by transmitting the audio data from the audio output I / F 105 via the network.

  FIG. 2 is a functional block diagram of the audio signal processing system 100 according to the first embodiment of this invention.

  The audio signal processing system 100 includes an audio input I / F 102, an audio personality feature analysis unit 110, an audio signal processing parameter determination unit 120, an audio signal processing unit 130, and an audio output I / F 105.

  The audio input I / F 102 is an interface that receives input of audio data, and corresponds to an interface that connects a microphone, a storage medium, a network, or the like to the audio signal processing system 100, for example.

  When the audio input I / F 102 is an interface for connecting a microphone, the audio data is input to the audio signal processing system 100 as an electrical signal in which the audio is converted by the microphone. When the audio input I / F 102 is an interface for connecting a storage medium, the audio data is input to the audio signal processing system 100 by loading the audio data stored in the storage medium. When the audio input I / F 102 is an interface for connecting a network, the audio data is input to the audio signal processing system 100 as audio stream data, for example.

  Based on the voice data input to the voice input I / F 102, the voice personality feature analysis unit 110 calculates information (speech feature amount, individuality feature amount, and individuality degree) related to the personality of the voice data. . The voice individuality feature analysis unit 110 includes a voice feature amount calculation unit 111 that calculates a voice feature amount, a personality feature amount calculation unit 112 that calculates a personality feature amount, and a personality degree calculation unit 113 that calculates a personality degree. Prepare.

  The speech feature quantity calculation unit 111 analyzes speech data input to the speech input I / F 102 using a known speech feature quantity analysis model, and acquires phoneme feature quantities (such as Mel-cepstrum) and prosodic feature quantities (basic At least one or more audio feature quantities that are time-series data such as a frequency F0 pattern and power are calculated.

  The personality feature amount calculation unit 112 is calculated by the speech feature amount calculation unit 111 using a known model such as a speech recognition model, a feller identification model, an accent identification model, a vocal tract shape identification model, and a vocal cord noise extraction model. A plurality of personality feature amounts are calculated based on the voice feature amount. In other words, the personality feature amount is obtained by quantifying audio data using a plurality of preset models.

  Based on the individuality feature amount calculated by the individuality feature amount calculation unit 112, the individuality degree calculation unit 204 calculates an individuality degree that quantifies the strength of individuality of each individuality feature amount.

  The audio signal processing parameter determination unit 120, when there is an individuality degree that does not satisfy the predetermined individuality degree condition 1001 (see FIG. 11) among the individuality degrees calculated by the individuality degree calculation unit 113, An audio signal processing parameter 1003 (see FIG. 12), which is a parameter for executing audio signal processing whose sexiness satisfies the personality level condition 1001, is determined. Although details will be described later, the audio signal processing unit 130 executes audio signal processing on the audio data based on the audio signal processing parameter 1003. The audio signal processing parameter determination unit 120 includes an individuality degree condition determination unit 121, an audio signal processing parameter adjustment unit 122, and an individuality degree recalculation unit 123.

  The individuality degree condition determination unit 121 determines whether the individuality degree satisfies a predetermined individuality degree condition 1001. When it is determined that the individuality degree satisfies the individuality degree condition 1001, the individuality degree condition determination unit 121 outputs the audio signal processing parameter 1003 to the audio signal processing unit 130. On the other hand, when it is determined that the individuality degree does not satisfy the individuality degree condition 1001, the individuality degree condition determination unit 121 shifts the processing to the audio signal processing parameter adjustment unit 122.

  The audio signal processing parameter adjustment unit 122 determines that the individuality degree does not satisfy the individuality degree condition 1001 by the individuality degree condition determination unit 121, so that the audio signal processing parameter 1003 satisfies the individuality degree condition 1001. Adjust. Based on the audio signal processing parameter 1003 adjusted by the audio signal processing parameter adjustment unit 122, the individuality degree recalculation unit 123 recalculates and calculates the individuality degree of the audio data when the audio signal processing is executed. The individuality degree is output to the individuality degree condition determination unit 121. The individuality degree condition determination unit 121 determines whether the adjusted individuality degree calculated again by the individuality degree recalculation unit 123 satisfies the individuality degree condition 1001.

  In the audio signal processing parameter determination unit 120, the individuality degree condition determination unit 121, the audio signal processing parameter adjustment unit until the individuality degree condition determination unit 121 determines that the individuality degree satisfies the individuality degree condition 1001. 122 and the individuality degree recalculation unit 123 repeatedly execute the adjustment processing of the audio signal processing parameter 1003.

  The audio signal processing unit 130 executes audio signal processing based on the audio signal processing parameter 1003 input from the audio signal processing parameter determination unit 120. The voice signal processing unit 130 includes a personality feature removal unit 131 and a voice synthesis unit 132.

  The personality feature removing unit 131 performs audio signal processing on the audio data based on the audio signal processing parameter 1003 so as to remove the individuality that does not satisfy the individuality condition. The voice synthesizer 213 synthesizes voice data that has been subjected to voice signal processing by the personality feature remover 131.

  The audio output I / F 105 is an interface for outputting audio data, and corresponds to an interface for connecting a speaker, a storage medium, a network, or the like to the audio signal processing system 100, for example.

  When the audio output I / F 105 is an interface for connecting a speaker, the audio data is converted into an electric signal and output from the speaker. Further, when the audio output I / F 105 is an interface for connecting a storage medium, the audio data is output by writing the audio data subjected to the audio signal processing to the storage medium. Further, when the audio output I / F 105 is an interface for connecting a network, the audio data is output to the network as audio stream data, for example.

  Each component described above is realized by the CPU 103 executing a program corresponding to each component. Further, a part or all of each of the above-described components may be realized by hardware, for example, by designing with an integrated circuit. Moreover, the case where it implement | achieves with a software by interpreting and executing the program which implement | achieves each component mentioned above was demonstrated, but information, such as a program, a table, and a file which implement | achieve the function of each component, is only memory. In addition, it can be stored in a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD, and can be downloaded and installed via a network or the like as necessary. It goes without saying that it is possible.

  Next, details of the operation of each component of the audio signal processing system will be described below.

(Operation of voice personality feature analysis unit 110)
First, the operation of the voice personality feature analysis unit 110 will be described with reference to FIG. FIG. 3 is a functional block diagram of the voice personality feature analysis unit 110 according to the first embodiment of this invention.

  The voice personality feature analysis unit 110 receives voice data from the voice input I / F 102.

  Next, the audio feature amount calculation unit 111 will be described. The audio feature amount calculation unit 111 includes at least one module for calculating a predefined audio feature amount. Then, the module included in the sound feature amount calculation unit 111 calculates the sound feature amount 310 corresponding to itself, and inputs the calculated sound feature amount 310 to the personality feature amount calculation unit 112.

  The speech feature value 310 is roughly classified into a language feature value 311 related to text information of speech data, a prosodic feature value 312 related to the prosody of speech data, and a phoneme feature value 313 related to the phoneme of speech data. 3 includes a language feature quantity calculation module 301, a prosodic feature quantity calculation module 302, and a phoneme feature quantity calculation module 303.

  The language feature amount calculation module 301 calculates a language feature amount 311 based on the speech waveform of the input speech data. The language feature quantity 311 is, for example, an utterance text and filler information, and will be described in detail with reference to FIG.

  The prosodic feature value calculation module 302 calculates the prosodic feature value 312 based on the speech waveform of the input speech data. The prosodic feature value 312 includes the power of the speech waveform of the speech data, the speech fundamental frequency (F0) of the speech waveform of the speech data, the rhythm (phoneme duration), and the details will be described with reference to FIGS. 6A to 6C.

  The phoneme feature quantity calculation module 303 calculates a phoneme feature quantity 313 from the speech waveform of the input speech data. The phoneme feature quantity 313 is, for example, a phoneme spectrum of a voice waveform of voice data, and details will be described with reference to FIG.

  The language feature value 311, prosodic feature value 312, and phoneme feature value 313 when FIG. 4 is input to the speech individuality feature analysis unit 110 as speech data will be described with reference to FIGS. 5 to 7.

  FIG. 4 is an explanatory diagram of an audio waveform of audio data according to the first embodiment of this invention.

  The vertical axis shown in FIG. 4 represents the sound pressure of the speech waveform of the speech data input to the speech individuality feature analysis unit 110, and the horizontal axis represents time.

  FIG. 5 is an explanatory diagram of the language feature quantity 311 according to the first embodiment of this invention.

  FIG. 5 shows the language feature quantity 311 calculated by the language feature quantity calculation module 301 based on the voice data shown in FIG. In Figure 5, "(er etc.) Hello, (er etc.) I (er etc.) Tanaka and another to you" is calculated as the spoken text and language feature value filler information by language feature amount calculation module 301 311 . In addition, the part in the parenthesis shown in FIG. 5 is filler information.

  As described above, the prosody feature quantity calculation module 302 uses the time series data (see FIG. 6A) of the power of the speech waveform of the speech data and the speech fundamental frequency (F0) of the speech waveform of the speech data as the prosody feature amount 312. Series data (see FIG. 6B) and rhythm (phoneme duration) (see FIG. 6C) are calculated.

  FIG. 6A is an explanatory diagram of power waveform time-series data calculated as the prosodic feature value 312 according to the first embodiment of this invention.

  The vertical axis in FIG. 6A indicates the power of the speech waveform, and the horizontal axis in FIG. 6A indicates time. The power of the speech waveform is calculated by multiplying the common logarithm of the value obtained by dividing the square value of the sound pressure effective value by the square value of the reference sound pressure by 10.

  FIG. 6B is an explanatory diagram of time-series data of the fundamental frequency of the speech waveform of speech data calculated as the prosodic feature value 312 according to the first embodiment of this invention.

  The vertical axis in FIG. 6B indicates the fundamental frequency, and the horizontal axis in FIG. 6B indicates time.

  FIG. 6C is an explanatory diagram of time-series data of phoneme durations of the speech waveform of speech data calculated as the prosodic feature value 312 according to the first embodiment of this invention.

  In FIG. 6C, the speech waveform shown in FIG. 4 is divided for each phoneme, and the time during which each phoneme is uttered is shown as the time from the dotted line to the next dotted line as the phoneme duration.

  FIG. 7 is an explanatory diagram of the phoneme spectrum calculated as the phoneme feature quantity 313 according to the first embodiment of this invention.

  The vertical axis in FIG. 7 indicates frequency, and the horizontal axis in FIG. 7 indicates time. In FIG. 7, the darker the frequency, the greater the distribution.

  Next, the personality feature amount calculation unit 112 will be described. The personality feature quantity calculation unit 112 includes at least one module for calculating a predefined personality feature quantity. The module included in the personality feature amount calculation unit 112 calculates a personality feature amount 320 corresponding to itself, and inputs the calculated personality feature amount 320 to the personality degree calculation unit 113.

  A list of personality feature quantities 320 is shown in FIG. FIG. 8 is an explanatory diagram of the personality feature quantity list 800 according to the first embodiment of this invention.

  As shown in FIG. 8, the individuality feature amount calculation unit 112, for example, filler appearance frequency, word usage frequency, grammatical error frequency, dialect appearance frequency, dialect accent appearance frequency, average F0 value, F0 pattern shape, phonological feature amount. , And voice quality feature quantities are calculated. The personality feature quantity calculation unit 112 may calculate a personality feature quantity 320 other than the personality feature quantity 320 in the personality feature quantity list 800 shown in FIG. 8, or the personality feature quantity shown in FIG. It is not necessary to calculate all the personality features 320 of the list 800.

  The filler appearance frequency is a probability that filler information appears in one accent phrase. For example, the language feature value 311 shown in FIG. 5, "(er etc.) Hello,", "(er etc.) I", "(er etc.) Tanaka and", and "you anymore." Four of There are accent phrases, and among these accent phrases, there are three accent phrases in which filler information (edo) appears. Therefore, the filler appearance frequency of the language feature quantity 311 shown in FIG. 5 is 3/4 = 0.75.

  The word usage frequency is a probability that a predefined word that can specify individuality appears in one accent phrase. For example, when the word “Tanaka” is defined, the word appearance frequency of the language feature amount shown in FIG. 5 is ¼ = 0.25.

  The grammatical error frequency is the probability that a sentence contains a grammatical error. Dialect appearance frequency is the probability that a word predefined as a dialect will appear in one accent phrase.

  The dialect accent appearance frequency is a probability that an accent defined in advance as a dialect accent appears in one accent phrase. The dialect accent appearance frequency is calculated based on the fundamental frequency of the prosodic feature value 312.

  The average F0 value is an average value of fundamental frequencies calculated as the prosodic feature value 312 and is an average value of the voice pitch of the speaker. For example, the average F0 value of the fundamental frequency shown in FIG. 6 is 230 Hz. The F0 pattern shape is a value calculated based on the fundamental frequency calculated as the prosodic feature value 312 and indicates an average pattern of the speaker's intonation.

  The phoneme feature quantity and the voice quality feature quantity are calculated based on the phoneme spectrum of the phoneme feature quantity 313, and the phoneme feature quantity is a value indicating the voiceprint (speech tract shape) of the speaker.

  3 includes a dialect feature value calculation module 304, an accent feature value calculation module 305, and a vocal tract shape feature value calculation module 306.

  The dialect feature quantity calculation module 304 calculates the dialect feature quantity 321 described above based on the language feature quantity 311.

  The accent feature quantity calculation module 305 calculates the dialect accent appearance frequency as the accent feature quantity 322 based on the fundamental frequency calculated as the prosodic feature quantity 312. Note that the accent feature quantity calculation module 305 may calculate the probability that an accent error is included in one accent phrase as the accent feature quantity 322.

  The vocal tract shape feature amount calculation module 306 calculates a vocal tract shape feature amount 323 based on the phoneme spectrum calculated as the phoneme feature amount 313.

  Next, the individuality degree calculation unit 113 will be described.

  The individuality degree calculation unit 113 calculates the individuality degree 330 of each individuality feature amount 320 calculated by the individuality feature amount calculation unit 112 using a predefined individuality degree calculation model 340. The individuality degree calculation model 340 is defined to correspond to each individuality feature amount 320 and is information for calculating the individuality degree. For example, the individuality degree calculation model 340 is statistical data including the number distribution of the individuality feature amount 320 values.

  The individuality degree calculation model 340 corresponding to the average F0 value calculated by the individuality feature amount calculation unit 112 will be described with reference to FIG. FIG. 9 is an explanatory diagram of the individuality degree calculation model 340 corresponding to the average F0 value according to the first embodiment of this invention.

  In FIG. 9, it is assumed that statistical data of the number distribution of average F0 values of a large amount of audio data is prepared. The statistical data shown in FIG. 9 includes male average F0 value statistical data 901 which is statistical data of male average F0 values, female average F0 value statistical data 902 which is statistical data of female average F0 values, and average F0 of all genders. All gender average F0 value statistical data 903 which is statistical data of values is included. For example, the male average F0 value statistical data 901 indicates that the number of persons whose average F0 value is 120 Hz is the maximum.

  Based on the male average F0 value statistical data 901, a male average F0 value personality degree calculation model 911, which is a personality degree calculation model corresponding to the average F0 value of males, is calculated, and based on the female average F0 value statistical data 902, females are calculated. A female average F0 value individuality degree calculation model 912, which is a personality degree calculation model corresponding to the average F0 value, is calculated, and the individuality corresponding to the average F0 value for all genders based on the average F0 value statistical data 903 for all genders An average F0 value individuality degree calculation model 913 that is a degree calculation model is calculated.

  Each individuality degree calculation model 911 to 913 is generated such that the number distribution and the individuality degree are inversely proportional. In other words, the individuality degree calculation models 911 to 913 are generated so that the individuality degree of the average F0 value having a large number distribution is small and the individuality degree of the average F0 value having a small number distribution is large.

  For example, in the male average F0 value individuality degree calculation model 911, the average F0 value 120 Hz that maximizes the number distribution in the male average F0 value statistical data 901 has the lowest individuality degree. Further, in the male average F0 value statistical data 901, the distribution of the number of people decreases as the average F0 value moves away from 120 Hz. Therefore, in the male average F0 value individuality calculation model 911, the individuality degree increases as the average F0 value moves away from 120 Hz. growing. For example, when the average F0 value of a certain male voice data is 300 Hz, the distribution of the number of persons whose average F0 value is 300 Hz is very small, and thus the individuality degree of the average F0 value of the male is very large. Therefore, the individuality degree calculated using the personality feature quantity distribution as the individuality degree calculation model 340 can be said to be a quantification of the deviation degree of the reference individuality feature quantity having the largest number of people.

  Since the average F0 value depends on gender, the individuality degree is calculated using the individuality degree calculation model 911 or 912 for each sex when the individuality degree calculation unit 113 can determine the gender. Thereby, the individuality degree can be accurately calculated. If the individuality degree calculation unit 113 cannot determine the gender, the individuality degree is calculated using the average F0 value individuality degree calculation model 913 for all genders.

  For example, when the average F0 value calculated from the fundamental frequency shown in FIG. 6B by the personality feature amount calculation unit 112 is 230 Hz and the voice data of the average F0 value is that of a male, the male average F0 value of the individual Based on the sex calculation model 911, the individual personality of the average F0 value of this male is calculated as 200.

  3 includes a dialect individuality calculation module 307, an accent individuality calculation module 308, and a vocal tract shape individuality calculation module 309.

  The dialect individuality degree calculation module 307 calculates a dialect individuality degree 331 that is the individuality degree of the dialect feature quantity 321 using the individuality degree calculation model 340 corresponding to the dialect feature quantity 321. The accent individuality degree calculation module 308 calculates an accent individuality degree 332 that is the individuality degree of the accent feature amount 322 using the individuality degree calculation model 340 corresponding to the accent feature amount 322. The vocal tract shape individuality degree calculation module 309 uses the individuality degree calculation model 340 corresponding to the vocal tract shape feature quantity 323 to calculate the vocal tract shape individuality degree 333 which is the individuality degree of the vocal tract shape feature quantity 323. calculate.

  The individuality degree calculation model 340 is not limited to the statistical data of the personality feature quantity distribution, but may be a calculation formula and a rule defined by an expert or the like. The model may be automatically generated from the variance value.

(Operation of the audio signal processing parameter determination unit 120)
Next, the operation of the audio signal processing parameter determination unit 120 will be described with reference to FIG. FIG. 10 is a functional block diagram of the audio signal processing parameter determination unit 120 according to the first embodiment of this invention.

  The audio signal processing parameter determination unit 120 includes, as the parameter determination data group 1002, the individuality feature amount 320 calculated by the individuality feature amount calculation unit 112, the individuality degree 330 calculated by the individuality degree calculation unit 113, and The audio signal processing parameter 1003 is acquired, and the acquired parameter determination data group 1002 is input to the individuality degree condition determining unit 121. Details of the audio signal processing parameter 1003 will be described with reference to FIG.

  The personality condition determination unit 121 includes a personality condition determination module 1004. The individuality degree condition determination module 1004 determines whether or not the individuality degree 330 satisfies a predefined individuality degree condition 1001. Details of the individuality condition 1001 will be described with reference to FIG.

  When the individuality degree 330 satisfies the individuality degree condition 1001, when the individuality degree condition determination module 1004 determines (Yes), the audio signal processing parameter determination unit 120 ends the processing, and the audio signal processing parameter 1003 is changed to the audio signal processing parameter 1003. The signal is input to the signal processing unit 130. On the other hand, if the individuality degree 330 does not satisfy the individuality degree condition 1001 and the individuality degree condition determination module 1004 determines (No), the audio signal processing parameter determination unit 120 sets the audio signal processing parameter adjustment unit 122 to Migrate processing. The audio signal processing parameter adjustment unit 122 adjusts the audio signal processing parameter 1003 based on the audio signal processing parameter adjustment method, which will be described in detail with reference to FIG. 13, and uses the value of the adjusted audio signal processing parameter 1003 for parameter determination data. The audio signal processing parameter 1003 of the group 1002 is updated. Then, the audio signal processing parameter adjustment unit 122 inputs the adjusted audio signal processing parameter 1003 to the individuality degree recalculation unit 123.

  Note that the initial value of the audio signal processing parameter 1003 is set to a value at which the audio signal processing unit 130 does not execute the audio signal processing. That is, when it is determined in the process by the first individuality degree condition determination module 1004 that all the individuality degrees 330 satisfy the individuality degree condition 1001, the audio signal processing unit 130 performs no processing on the audio data. Without executing, the audio data is input to the audio output I / F 105.

  The personality condition 1001 will be described with reference to FIG. FIG. 11 is an explanatory diagram of the individuality condition 1001 according to the first embodiment of this invention.

  The personality condition 1001 is set according to the individuality that the user of the audio signal processing system 100 desires to conceal, and various personality conditions 1001 are set. For example, when it is desired to conceal gender, the individuality condition 1001 is set such that the average F0 value individuality is not more than a predetermined value (for example, 20) as in Example 1 shown in FIG.

  In addition, a plurality of individuality feature quantities such as “average F0 value individuality degree ≦ 15, filler usage frequency individuality degree ≦ 20, and voice quality characteristic individuality degree ≦ 10” in Example 3 shown in FIG. It is also possible to set the individuality degree condition 1001 for the individuality degree 330 corresponding to 320.

  Further, some calculation may be performed on the individuality degree, and a condition may be set for the calculation result. For example, as in Example 4 shown in FIG. 11, the average value of the individuality levels corresponding to all the individuality feature amounts is calculated, and the calculated average value is equal to or less than a predetermined value (for example, 25). The condition 1001 may be set.

  The audio signal processing parameter 1003 will be described with reference to FIG. FIG. 12 is an explanatory diagram of the audio signal processing parameter 1003 according to the first embodiment of this invention.

  The audio signal processing parameter 1003 is a parameter for the audio signal processing unit 130 to execute audio signal processing on the audio data.

  The audio signal processing parameter 1003 includes the type of personality feature amount to be adjusted, the adjustment method, and the adjustment amount (change amount).

  For example, when adjusting the filler appearance frequency of the personality feature amount 320, the filler appearance frequency is registered in the type of personality feature amount to be adjusted, and the type of filler to be adjusted is registered in the adjustment method. The number of fillers to be deleted is registered in.

  Further, when adjusting the average F0 value of the personality feature value 320, the average F0 value is registered as the type of the personality feature value to be adjusted, and the adjustment method uses a publicly known PSOLA (used to adjust the average F0 value). Pitch Synchronous Overlap Add Method) is registered, and a value indicating how much the average F0 value is changed is registered in the adjustment amount.

  Next, an audio signal processing parameter adjustment method will be described with reference to FIG. FIG. 13 is an explanatory diagram of an audio signal processing parameter adjustment method according to the first embodiment of this invention.

  In the audio signal processing parameter adjustment method, the individuality feature amount 320 to be adjusted and the adjustment amount in one process by the audio signal processing parameter adjustment unit 122 are registered for each individuality degree condition 1001.

  When the individuality degree 330 does not satisfy the example 1 shown in FIG. 11 of the individuality degree condition 1001, the audio signal processing parameter adjustment unit 122 performs the audio signal processing based on the example 1 shown in FIG. 13 of the audio signal processing parameter adjustment method. The parameter 1003 is adjusted. That is, the average F0 value is adjusted by 10 Hz in the direction of decreasing the average F0 value individuality. Note that the audio signal processing parameter adjustment unit 122 determines whether to increase or decrease the average F0 value with reference to the individuality degree calculation model 340. For example, when the average F0 value is 230 Hz, the average F0 value personality is 200, and the threshold value of the average F0 value is smaller than 20, the personality condition 1001 is adjusted. The unit 122 refers to the male average F0 value individuality degree calculation model 911 shown in FIG. 9 and adjusts the audio signal processing parameter 1003 so that the average F0 value is reduced by 10 Hz.

  In the audio signal processing parameter adjustment method, from the viewpoint of reducing the processing load of the audio signal processing parameter adjustment unit 122, it is desirable to determine the adjustment amount so that the individuality feature amount 320 monotonously increases or decreases monotonously.

  Returning to FIG. 10, the individuality degree recalculation unit 123 will be described.

  The individuality degree recalculation unit 123 includes an individuality feature amount estimation module 1005 and an individuality degree calculation module 1006.

  The individuality feature amount estimation module 1005 estimates the adjusted individuality feature amount 320 based on the audio signal processing parameter 1003 adjusted by the audio signal processing parameter adjustment unit 122 and is included in the parameter determination data group 1002. The personality feature amount 320 is updated.

  The individuality degree calculation module 1006 recalculates the individuality degree 330 based on the individuality feature quantity 320 estimated by the individuality feature quantity estimation module 1005, and updates the individuality degree 330 of the parameter determination data group 1002. . It is desirable to use the same program as the individuality degree calculation unit 113 included in the voice individuality feature analysis unit 110 for the individuality degree calculation module 1006.

  Next, the overall operation of the audio signal processing parameter determination unit 120 will be described by taking as an example a case where the average F0 value, which is the individuality feature amount 320 of the audio data shown in FIG. Will be described.

  Here, it is assumed that the individuality condition 1001 is set such that the average F0 value individuality is 30 or less, and the initial value of the audio signal processing parameter 1003 is “average F0 correction presence / absence = none, average F0 value correction”. It is assumed that “method = none, average F0 value adjustment amount = 0 Hz”.

  First, the individuality degree condition determination unit 121 determines that the individuality degree condition 1001 is not satisfied because the average F0 value individuality degree is 200, and sets the parameter determination data group 1002 to the audio signal processing parameter adjustment unit 122. input.

  Here, it is assumed that the audio signal processing parameter adjustment method is an adjustment method for adjusting the average F0 value by 10 Hz in the direction of decreasing the individuality. Since the average F0 value (230 Hz) is larger than the average value (120 Hz) of male average F0 values shown in FIG. “F0 value adjustment method = PSOLA, average F0 value adjustment amount = −10 Hz”, the audio signal processing parameter 1003 of the parameter determination data group 1002 is updated, and the parameter determination data group 1002 is recalculated. Input to the unit 123.

  Based on the adjusted audio signal processing parameter 1003, the personality feature amount estimation module 1005 subtracts the average F0 value by 10 Hz, and updates the average F0 value of the parameter determination data group 1002 to 220 Hz.

  Next, the individuality degree calculation module 1006 calculates the updated individuality degree 330 of the average F0 value (220 Hz), and updates the average F0 value individuality degree to 170.

  As described above, the parameter determination data group 1002 includes “average F0 value = 220 Hz”, “average F0 value individuality = 170”, “average F0 adjustment presence / absence = present, average F0 value adjustment method = PSOLA, average F0 value adjustment”. The amount is updated to “−10 Hz”.

  The updated parameter determination data group 1002 is input to the individuality degree condition determination unit 121 and is described above until the individuality degree condition determination unit 121 determines that the individuality degree 330 satisfies the individuality degree condition 1001. Repeat the process.

  Finally, the parameter determination data group 1002 includes “average F0 value = 150 Hz”, “average F0 value individuality = 30”, “average F0 adjustment presence / absence = present, average F0 value adjustment method = PSOLA, average F0 value adjustment” In the case of “amount = −80 Hz”, the individuality degree condition determining unit 121 determines that the individuality degree 330 satisfies the individuality degree condition 1001, and all the processes of the audio signal processing parameter determining unit 120 are completed.

(Operation of the audio signal processing unit 130)
Next, the operation of the audio signal processing unit 130 will be described with reference to FIG. FIG. 14 is a functional block diagram of the audio signal processing unit 130 according to the first embodiment of this invention.

  The voice signal processing unit 130 includes a personality feature removal unit 1401 and a voice synthesis unit 1405.

  The personality feature removal unit 1401 executes speech signal processing on speech data based on the speech signal processing parameter 1003, and performs a language personality feature removal module 1402, a prosodic personality feature removal module 1403, and a phonological personality feature removal module. Including at least one of 1404.

  The language individuality feature removal module 1402 executes voice signal processing for adjusting the language feature quantity 311 on the voice data input from the voice input I / F 102 based on the voice signal processing parameter 1003. The prosodic personality feature removal module 1403 executes speech signal processing for adjusting the prosodic feature value 312 on the speech data input from the speech input I / F 102 based on the speech signal processing parameter 1003. The phoneme personality feature removal module 1404 executes voice signal processing for adjusting the phoneme feature quantity 313 on the voice data input from the voice input I / F 102 based on the voice signal processing parameter 1003.

  However, the language personality feature removal module 1402, the prosodic personality feature removal module 1403, and the phonological personality feature removal module 1404 use initial values for parameters that are not defined in the speech signal processing parameter 1003, Usually, the initial value of the speech signal processing parameter 1003 is set not to execute the language personality feature removal module 1402, the prosodic personality feature removal module 1403, and the phonological personality feature removal module 1404. .

  The speech synthesizer 1405 again synthesizes the speech using the speech feature quantity 310 that has been subjected to speech signal processing by the personality feature remover 1401 and outputs the synthesized speech to the speech output I / F 105. As a speech synthesis method of the speech synthesizer 1405, there are a method of reconnecting the disconnected speech after cutting the original speech waveform, a parameter synthesis method using a vocal tract filter, and the like, as in PSOLA.

  In FIG. 10, when the average F0 value, which is the personality feature amount 320 of the audio data shown in FIG. 4, is 230 Hz and the average F0 value individuality is 200, the audio signal processing parameter determination unit 120 performs the audio signal processing. It has been described that the parameter 1003 is determined as “average F0 adjustment presence / absence = present, average F0 value adjustment method = PSOLA, average F0 value adjustment amount = −80 Hz”. The operation of the audio signal processing unit 130 when the audio signal processing parameter 1003 is input to the audio signal processing unit 130 will be described.

  Since the average F0 value is a prosodic personality feature amount, in order to adjust the average F0 value, the prosodic personality feature amount must be adjusted. For this reason, when the audio signal processing parameter 1003 described above is input to the audio signal processing unit 130, only the prosodic personality feature removal module 1403 executes the audio signal processing.

  Specifically, the prosodic personality feature removal module 1403 detects a periodic waveform from the waveform of the voiced sound of the voice input from the voice input I / F 102. Then, the prosodic personality feature removal module 1403 reduces the average F0 value by 80 Hz by superimposing the detected periodic waveform so that the interval between the detected periodic waveforms is shortened by 1/80 seconds.

  The prosodic personality feature removal module 1403 may reduce the F0 value of all voiced sound waveforms by 80 Hz, but in order to ensure the inflection and naturalness of the sound itself, the amplitude of the voiced sound waveform is adjusted. You may change based on F0 value. Specifically, the prosodic personality feature removal module 1403 reduces the amplitude of the voiced sound waveform when the F0 value of the voiced sound waveform is reduced, and increases the F0 value of the voiced sound waveform when the F0 value of the voiced sound waveform is increased. Increase the waveform amplitude.

  FIG. 15A is time-series data of the fundamental frequency before the average F0 value is adjusted according to the first embodiment of this invention. FIG. 15A is the same as FIG. 6A and shows that the average F0 value is 230 Hz.

  FIG. 15B is time-series data of the fundamental frequency after the average F0 value of the first embodiment of the present invention is adjusted. The average F0 value is reduced by 80 Hz by the prosodic personality feature removal module 1403, and the average F0 value becomes 150 Hz.

  Note that, in the audio signal processing parameter 1003 described above, only the average F0 value is adjusted, so that the audio feature quantity 310 other than the fundamental frequency does not change from FIGS. 5, 6A, 6C, and 7.

  In the present embodiment, when there is a personality degree 330 that does not satisfy the predetermined personality degree condition 1001, the audio signal processing parameter 1003 is determined so that the individuality degree 330 satisfies the personality degree condition 1001, and the determined voice Based on the signal processing parameter 1003, the audio signal processing is executed so that the individuality degree 330 satisfies the individuality degree condition 1001. As a result, it is possible to remove only personality information with strong personality and prevent deterioration of the audio data as much as possible.

(Second Embodiment)
A second embodiment of the present invention will be described with reference to FIGS.

  In this embodiment, the audio signal processing system of the first embodiment is applied to a server that stores uploaded audio data and transfers the audio data in response to a download request. In this case, using the sufficient processing speed of the server and the sufficient capacity of the storage device, at the time of uploading the voice data, the individuality condition is satisfied for each of the plurality of individuality conditions, and the deterioration of sound quality is minimized. Generate voice data. As a result, it is possible to provide audio data that satisfies the individuality condition and minimizes deterioration in sound quality.

  FIG. 16 is an explanatory diagram of the configuration of the information processing system according to the first embodiment of this invention.

  The information processing system includes a server 1600, an upload terminal 1610 that uploads audio data to the server 1600, and a download terminal 1620 that downloads audio data from the server 1600. The server 1600, the upload terminal 1610, and the download terminal 1620 are connected via a network 1605 that can communicate with each other. In FIG. 16, one upload terminal 1610 and one download terminal 1620 are shown, but a plurality of upload terminals 1610 and download terminals 1620 may be provided.

  The server 1600 functions as the audio signal processing system 100 and includes a CPU 103, a memory 104, an auxiliary storage device 101, and a communication I / F (interface) 1601. The CPU 103, the memory 104, and the auxiliary storage device 101 are the same as the audio signal processing system 100 shown in FIG. The CPU 103, the memory 104, the auxiliary storage device 101, and a communication I / F (interface) 1601 are provided. The CPU 103, the memory 104, and the auxiliary storage device 101 are connected to each other via the bus 106.

  The communication I / F 1601 is an interface that connects the audio signal processing system 100 to the network 1605. The communication I / F 1601 functions as a voice input I / F 102 when receiving voice data from the upload terminal 1610. The communication I / F 1601 functions as an audio output I / F 105 when the download terminal 162 downloads audio data from the server 1600.

  The upload terminal 1610 is a terminal that uploads input voice data to the server 1600, and includes a CPU 103, a memory 104, a communication I / F 1611, and a voice input I / F 1612. The CPU 103, the memory 104, the communication I / F 1611, and the voice input I / F 1612 are connected via the bus 106. Since the CPU 103 and the memory 104 are the same as the server 1600, the description thereof is omitted.

  The communication I / F 1611 is an interface that connects the upload terminal 1610 to the network 1605. The voice input I / F 1612 is an interface for inputting voice to the upload terminal 1610.

  The download terminal 1620 is a terminal that downloads audio from the server 1600, and includes a CPU 103, a memory 104, a communication I / F 1621, and an audio output I / F 1622. The CPU 103, the memory 104, the communication I / F 1611, and the audio output I / F 1613 are connected via the bus 106. Since the CPU 103 and the memory 104 are the same as the server 1600, the description thereof is omitted.

  The communication I / F 1621 is an interface that connects the download terminal 1620 to the network 1605. The audio output I / F 1622 is an interface for the download terminal 1620 to output audio.

  The communication I / Fs 1601, 1611, and 1621 can communicate audio data with the server 1600, the upload terminal 1610, and the download terminal 1620.

  Note that the server 1600, the upload terminal 1610, and the download terminal 1620 may be integrated into the same computer.

  FIG. 17 is a functional block diagram of the server 1600 according to the second embodiment of this invention. Of the functional blocks of the server 1600 shown in FIG. 17, the same functional blocks as those of the audio signal processing system 100 shown in FIG.

  An outline of the operation of the server 1600 will be described.

  Individuality authority is set for the download terminal 1620 or the user of the download terminal 1620. The individuality authority is information indicating how many individuality characteristics are disclosed, and an individuality degree condition 1001 is set corresponding to the individuality authority.

  When the audio data is uploaded, the server 1600 adjusts the uploaded audio data so as to satisfy the individuality condition 1001 for each individuality condition 1001 corresponding to all individuality authorities, and the adjusted audio data Is stored for each individuality authority.

  Further, when the server 1600 receives a download request from the download terminal 1620, the server 1600 specifies the personality authority of the download terminal 1620 that transmitted the download request or the user of the download terminal 1620, and the audio data corresponding to the specified personality authority And the searched voice data is transmitted to the download terminal 1620.

  The server 1600 includes a communication I / F 1601, a voice personality feature analysis unit 110, a voice signal processing parameter determination unit 1700, a voice signal processing unit 130, a user information extraction unit 1730, a personality authority setting unit 1750, and a voice search unit 1780. . The auxiliary storage device 101 of the server 1600 stores a personality condition list 1710 and an audio database 1720.

  First, processing of the server 1600 when audio data is uploaded will be described.

  When the communication I / F 1601 receives voice data uploaded by the upload terminal 1610, the communication I / F 1601 inputs the received voice data to the voice personality characteristic analysis unit 110.

  The voice personality feature analysis unit 110 calculates a voice feature value 310, a personality feature value 320, and a personality degree 330 based on the voice data, and inputs them to the voice signal processing parameter determination unit 1700.

  Here, in the personality degree condition list 1710, personality degree conditions 1001 corresponding to all personality authorities that can be set for the user of the download terminal 1620 are registered.

  The audio signal processing parameter determination unit 1700 determines an audio signal processing parameter 1003 such that the individuality degree 330 satisfies the individuality degree condition 1001 for each individuality degree condition 1001 registered in the individuality degree condition list 1710. The determined audio signal processing parameter 1003 is input to the audio signal processing unit 130.

  The audio signal processing parameter determination unit 1700 executes processing different from that of the audio signal processing parameter determination unit 120 of the first embodiment shown in FIG. 3 and will be described in detail with reference to FIG.

  The audio signal processing unit 130 performs audio signal processing on the audio data for each audio signal processing parameter 1003 input from the audio signal processing parameter determination unit 1700, and stores the audio data on which the audio signal processing has been performed in the audio database 1720. Store.

  A method for storing voice data in the voice database 1720 will be described.

  The voice database 1720 includes personality degree condition voice databases 1721 to 1723 corresponding to the individuality degree conditions. The audio signal processing unit 130 stores the audio data after execution of the audio signal processing in the individuality condition audio databases 1721 to 1723 corresponding to the individuality condition 1001 used for determining the audio signal processing parameter 1003.

  As described above, the audio data uploaded from the upload terminal 1610 is adjusted so as to satisfy each individuality condition 1001 registered in the individuality condition list 1710, and the audio data satisfying each individuality condition 1001 is Stored for each individuality condition 1001.

  Next, the operation of server 1600 when audio data is downloaded from server 1600 will be described.

  In the download operation of the server 1600, in response to the download request received by the communication I / F 1601 from the download terminal 1620, the voice database 1720 searches the voice database 1720 for voice data having personality authority corresponding to the user of the download terminal 1620. Is transmitted via the communication I / F 1601 to the download terminal 1620 that has transmitted the download request.

  The server 1600 includes a user information extraction unit 1730, a personality authority setting unit 1750, and a voice search unit 1780 as a configuration for realizing the download operation.

  First, the user information extraction unit 1730 extracts user information 1740 from the download request input from the communication I / F 1601 and inputs the extracted user information 1740 to the personality authority setting unit 1750. The user information 1740 is included in the download request and includes, for example, the type of the download terminal 1620 and user identification information that can identify the user of the download terminal 1620.

  The personality authority setting unit 1750 refers to the personality authority setting rule 1760, specifies the personality authority 1770 corresponding to the user information 1740 input from the user information extraction unit 1730, and performs a voice search for the specified personality authority 1770 Input to the unit 1780. In the personality authority setting rule 1760, the correspondence between the user information 1740 of the download terminal 1620 and the personality authority 1770 is registered. The personality authority setting rule 1760 will be described in detail with reference to FIG.

  The voice search unit 1780 searches for voice data matching the download request from the personality degree condition voice database corresponding to the inputted individuality authority 1770 among the individuality degree condition voice databases 1721 to 1723 included in the voice database 1720. To do. The voice search unit 1780 transmits the searched voice data to the download terminal 1620 that has transmitted the download request via the communication I / F 1601.

  As described above, according to the personality authority of the download terminal 1620 that transmitted the download request, it is possible to provide the voice data with confidentiality to the user of the download terminal 1620.

(Operation of Audio Signal Processing Parameter Determination Unit 1700)
Next, the operation of the audio signal processing parameter determination unit 1700 will be described with reference to FIG. FIG. 18 is a functional block diagram of the audio signal processing parameter determination unit 1700 according to the second embodiment of this invention. In the audio signal processing parameter determination unit 1700 shown in FIG. 18, the same functional blocks as those of the audio signal processing parameter determination unit 120 shown in FIG.

  The audio signal processing parameter determination unit 1700 according to the present embodiment includes, in addition to the audio signal processing parameter determination unit 120 according to the first embodiment, a parameter adjustment end condition 1801, an adjustment end determination module 1802, a minimum sound quality degradation prediction value output unit 1803, and a sound quality degradation. An estimation unit 1810 is added.

  The audio signal processing parameter determination unit 1700 selects one individuality degree condition 1001 from the individuality degree conditions 1001 registered in the individuality degree condition list 1710 as the individuality degree condition 1001 to be processed, and the individuality degree condition list The subsequent processing is repeated until the subsequent processing is executed for all personality condition 1001 registered in 1710.

  The personality condition determining unit 121 receives a personality condition 1001 to be processed and a parameter determination data group 1002. The parameter determination data group 1002 includes an individuality feature amount 320, an individuality degree 330, and an audio signal processing parameter 1003.

  When the individuality condition 1001 to be processed and the parameter determination data group 1002 are input to the individuality condition determination unit 121, the individuality condition determination module 1004 has the individuality degree 330 to be processed. It is determined whether or not the condition 1001 is satisfied.

  If the individuality degree condition determination module 1004 determines that the individuality degree 330 does not satisfy the individuality degree condition 1001 to be processed, the adjustment end determination module 1802 indicates that the value corresponding to the parameter adjustment end condition 1801 is a parameter adjustment end condition. It is determined whether or not 1801 is satisfied. The parameter adjustment end condition 1801 is a condition for the audio signal processing parameter adjustment unit 122 to end the adjustment of the audio signal processing parameter 1003. For example, the average F0 value is 50 Hz or less, or the number of adjustments is 10 or more. There are things. Details of the parameter adjustment end condition 1801 will be described with reference to FIG.

  When the adjustment end determination module 1802 determines that the value corresponding to the parameter adjustment end condition 1801 does not satisfy the parameter adjustment end condition 1801, the audio signal processing parameter adjustment unit 122 performs the audio signal processing parameter 1003 as in the first embodiment. Adjust. Then, the individuality degree recalculation unit 123 calculates the individuality degree 330 again based on the adjusted audio signal processing parameter 1003 as in the first embodiment. The personality condition determination module 1004 determines whether or not the recalculated personality 330 satisfies the personality condition 1001 to be processed.

  A case where the individuality degree condition determination module 1004 determines that the individuality degree 330 satisfies the individuality degree condition 1001 to be processed will be described.

  In this case, the personality degree condition determination module 1004 inputs the parameter determination data group 1002 to the sound quality degradation estimation unit 1810. When the parameter determination data group 1002 is input, the sound quality deterioration estimation unit 1810 performs the subsequent processing. The sound quality deterioration estimation unit 1810 includes a sound quality deterioration estimation module 1811, a sound quality improvement module 1813, and a minimum sound quality deterioration predicted value storage module 1814.

  First, the sound quality deterioration estimation module 1811 estimates the sound quality deterioration predicted value 1812 based on the individuality feature amount 320 of the parameter determination data group 1002 and inputs the estimated sound quality deterioration predicted value 1812 to the sound quality improvement module 1813. Since the method for estimating the deterioration amount of the sound quality is known, the description thereof is omitted.

  The sound quality improvement module 1813 determines whether or not the input sound quality deterioration prediction value 1812 is smaller than the minimum sound quality deterioration prediction value. When the predicted sound quality deterioration value 1812 is smaller than the predicted minimum sound quality deterioration value, the sound quality improvement module 1813 determines that the sound quality has improved, and uses the input parameter determination data group 1002 and the input predicted sound quality deterioration value 1812 as the minimum sound quality deterioration. The predicted value storage module 1814 is input.

  The minimum sound quality deterioration predicted value storage module 1814 stores the input sound quality deterioration predicted value 1812 as the minimum sound quality deterioration predicted value in the memory 104, and the input parameter determination data group 1002 as the minimum deterioration parameter determination data group. 104, and the process proceeds to the audio signal processing parameter adjustment unit 122.

  On the other hand, when the predicted sound quality degradation value 1812 is equal to or greater than the predicted minimum sound quality degradation value, the sound quality improvement module 1813 determines that the sound quality is not improved, and shifts the processing to the audio signal processing parameter adjustment unit 122.

  Next, a case where the adjustment end determination module 1802 determines that the value corresponding to the parameter adjustment end condition 1801 satisfies the parameter adjustment end condition 1801 will be described.

  In this case, the minimum sound quality degradation predicted value output unit 1803 inputs the minimum sound quality degradation predicted value and the minimum degradation parameter determination data group stored in the memory 104 to the audio signal processing unit 130. Then, the audio signal processing parameter determination unit 1700 selects the unprocessed personality degree condition 1001 if there is an unprocessed personality degree condition 1001 in the personality degree condition 1001 registered in the personality degree condition list 1710. The above-described processing is executed for the selected personality condition 1001, and if there is no unprocessed personality condition 1001 in the personality condition 1001 registered in the personality condition list 1710, the process ends. .

  As described above, the audio signal processing parameter determination unit 1700 according to the present embodiment has the audio signal processing parameter 1003 that minimizes the deterioration in sound quality from the audio signal processing parameter 1003 that satisfies the personality degree condition 1001 to be processed. Only to the audio signal processing unit 130. As a result, the download terminal 1620 is provided with audio data in which the desired personality is concealed and the deterioration in sound quality is minimized, and the content of the user who uploaded the audio data is concealed while the content is concealed. It is possible to prevent reducing the attractiveness of.

  FIG. 19 is an explanatory diagram of a specific example of the parameter adjustment end condition 1801 according to the second embodiment of this invention.

  FIG. 19 shows five specific examples of the parameter adjustment end condition 1801.

  The first example of the parameter adjustment end condition 1801 is “all parameter combinations have been tested”.

  The second example of the parameter adjustment end condition 1801 is “sufficient sound quality”. If the predicted sound quality degradation value 1812 estimated by the sound quality degradation estimation module 1811 is equal to or less than a predetermined value, the sound quality of the audio data provided to the user of the download terminal 1620 is sufficiently secured, and parameter adjustment is terminated.

  A third example of the parameter adjustment end condition 1801 is “time out”. When a predetermined time has elapsed after the processing by the audio signal processing parameter determination unit 1700 is started, the parameter adjustment is ended.

  The fourth example of the parameter adjustment end condition 1801 is “the specified number of adjustments has been reached”. When the number of adjustments of the audio signal processing parameter 1003 by the audio signal processing parameter adjustment unit 122 reaches a predetermined number, the parameter adjustment ends.

  The fifth example of the parameter adjustment end condition 1801 is “the parameter limit value has been reached”. For example, when adjusting the average F0 value to be small, when the average F0 value becomes 0, the audio signal processing parameter adjustment unit 122 cannot adjust the average F0 value, and thus the parameter adjustment ends.

  FIG. 20 is a graph showing the relationship between the sound quality deterioration and the change in individuality of the average F0 value according to the second embodiment of the present invention.

  As shown in FIG. 20, the sound quality deterioration amount is minimized when the average F0 value change amount is 0, and increases in proportion to the average F0 value change amount.

  Here, the specific example described in FIG. 10 will be described. In FIG. 10, when the average F0 value is 150 Hz, the individuality degree is 30, and the individuality degree condition 1001 is satisfied, and the adjustment of the audio signal processing parameter 1003 is completed. However, in this embodiment, the parameter adjustment end condition 1801 The adjustment of the audio signal processing parameter 1003 is continued until the condition is satisfied.

  Here, it is assumed that the parameter adjustment end condition 1801 is that the average F0 value is 50 Hz or less.

  In this case, the audio signal processing parameter 1003 having an average F0 value of 140 Hz to 50 Hz is calculated. As shown in FIG. 20, the sound quality deterioration amount increases in proportion to the change amount of the average F0 value. It is when the average F0 value is 150 Hz (the amount of change in the average F0 value is −80 Hz) that satisfies the quality condition 1001 and the sound quality deterioration amount is minimized. Therefore, the parameter determination data group 1002 when the change amount of the average F0 value is −80 Hz and the sound quality deterioration prediction value when the change amount of the average F0 value is −80 Hz are input to the audio signal processing unit 130.

  FIG. 21 is an explanatory diagram of a specific example of the individuality authority setting rule 1760 according to the second embodiment of this invention.

  In the personality authority setting rule 1760, the correspondence between the user information 1740 and the personality authority 1770 is registered.

  In the example 1 shown in FIG. 21, when the IP address of the download terminal 1620 indicates Japan, it is specified as personality authority 1, and when the IP address of the download terminal 1620 indicates China, it is specified as personality authority 3. Show.

  In Example 2 shown in FIG. 21, when the user attribute of the download terminal 1620 is a section manager, it is specified as individuality authority 1, and when the user attribute of the download terminal 1620 is a department manager, it is specified as individuality authority 2, and the download terminal If the user attribute 1620 is president, it indicates that the personality authority 3 is specified.

  In Example 3 shown in FIG. 21, when the user attribute of the download terminal 1620 is a friend, it is specified as personality authority 1, and when the user attribute of the download terminal 1620 is a colleague, it is specified as personality authority 2, and the download terminal If the user attribute 1620 is family, it indicates that the personality authority 3 is specified.

  In the example 4 shown in FIG. 21, when the type of the download terminal 1620 is OS1, it is specified as individuality authority 1, and when the type of the download terminal 1620 is OS2, it is specified as individuality authority 2, and the download terminal 1620 When the type is OS3, it indicates that the personality authority 3 is specified.

  As described above, according to the audio signal processing system 100 of the present embodiment, the audio data of the best sound quality from the audio data satisfying the personality condition 1001 corresponding to the user information 1740 of the download terminal 1620 is obtained. Can be provided to users.

(Third embodiment)
A third embodiment of the present invention will be described with reference to FIGS.

  The present embodiment is an embodiment in which the audio signal processing system 100 is applied to an aspect in which audio signal processing is executed in real time on spoken voice. As an example of a device on which the audio signal processing system 100 of this embodiment is mounted, a mobile phone, a fixed phone, and the like are assumed. The execution time (delay time) of audio signal processing based on the audio signal processing parameter 1003 differs depending on the type of audio signal processing and the amount of change in audio data due to audio signal processing. If speech signal processing takes too long to execute, actual conversation may not be established.

  Therefore, in the present embodiment, it is possible to prevent the actual conversation from being established by estimating the execution time of the audio signal processing that satisfies the individuality condition 1001 and executing the audio signal processing that minimizes the execution time. The processing load of the audio signal processing system 100 is reduced.

  Details of the relationship between the delay time and the amount of change in audio data due to audio signal processing will be described with reference to FIG.

  The hardware configuration of the audio signal processing system 100 of this embodiment is the same as the hardware configuration of the audio signal processing system 100 shown in FIG.

  Also, the audio signal processing system 100 according to the present embodiment is similar to the first embodiment and the second embodiment in that the audio individuality feature analysis unit 110, the audio signal processing parameter determination unit 2200 (see FIG. 22), and the audio signal processing Part 130 is provided.

  In the present embodiment, since the processing contents executed by the audio signal processing parameter determination unit 2200 are different from those in the first embodiment and the second embodiment, the audio signal processing parameter determination unit 2200 will be described with reference to FIG. FIG. 22 is a functional block diagram of an audio signal processing parameter determination unit 2200 according to the third embodiment of this invention. Of the functional blocks of the audio signal processing parameter determination unit 2200 illustrated in FIG. 22, the same components as those of the audio signal processing parameter determination unit 1700 illustrated in FIG. 18 of the second embodiment are denoted by the same reference numerals, and description thereof is omitted.

  The audio signal processing parameter determination unit 2200 includes a minimum delay time predicted value output unit 2201 instead of the minimum sound quality deterioration predicted value output unit 1803, and includes a delay time estimation unit 2210 instead of the sound quality deterioration estimation unit 1810.

  The delay time estimation unit 2210 includes a delay time estimation module 2211, a delay time determination module 2213, a minimum delay time predicted value storage module 2214, and a device state acquisition module 2215.

  The device status acquisition module 2215 acquires the status of the device on which the audio signal processing system 100 is mounted at a predetermined timing. The device status acquired by the device status acquisition module 2215 includes, for example, the usage status of the CPU 103, the usage status of the memory 104, and the communication status. The usage status of the CPU 103 may be the usage rate of the CPU 103, for example. The usage status of the memory 104 may be, for example, the usage rate of the memory 104. The communication status may be a bandwidth, for example.

  When it is determined that the individuality degree 330 satisfies the individuality degree condition 1001, the individuality degree condition determination module 1004 inputs the parameter determination data group 1002 to the delay time estimation module 2211 of the delay time estimation unit 2210.

  When the parameter determination data group 1002 is input, the delay time estimation module 2211 causes the device state acquisition module 2215 to acquire the state of the device on which the audio signal processing system 100 is mounted. Next, the delay time estimation module 2211 performs audio signal processing based on the audio signal processing parameter 1003 corresponding to the individuality degree satisfying the individuality degree condition 1001 based on the device state acquired by the device state acquisition module 2215. A predicted delay time value 2212 indicating time is estimated. Then, the delay time estimation module 2211 inputs the estimated delay time prediction value 2212 to the delay time determination module 2213.

  The delay time determination module 2213 determines whether or not the input delay time prediction value 2212 is smaller than a predetermined threshold value. When the delay time determination module 2213 determines that the input delay time prediction value 2212 is smaller than a predetermined threshold, the delay time determination module 2213 determines whether or not the input delay time prediction value 2212 is smaller than the minimum delay time prediction value. .

  When it is determined that the input delay time prediction value 2212 is smaller than the minimum delay time prediction value, the delay time determination module 2213 stores the input delay time prediction value 2212 as the minimum delay time prediction value, and a delay time estimation unit The parameter determination data group 1002 input to 2210 is stored as the minimum delay time parameter determination data group, and the process proceeds to the audio signal processing parameter adjustment unit 122.

  On the other hand, when the delay time determination module 2213 determines that the input delay time prediction value 2212 is greater than or equal to a predetermined threshold, or the input delay time prediction value 2212 is greater than or equal to the minimum delay time prediction value. If it is determined, the process proceeds to the audio signal processing parameter adjustment unit 122.

  When the adjustment end determination module 1802 determines that the value corresponding to the parameter adjustment end condition 1801 satisfies the parameter adjustment end condition 1801, the minimum delay time predicted value output unit 2201 determines the minimum delay time predicted value and the minimum delay time parameter. The data group is input to the audio signal processing unit 130, and the process ends.

  Thus, the audio signal processing parameter 1003 corresponding to the individuality degree satisfying the individuality degree condition 1001, the execution time of the audio signal processing is smaller than the predetermined threshold value, and the execution time of the audio signal processing is minimized. The audio signal processing parameter 1003 is input to the audio signal processing unit 130. For this reason, since the execution time of the audio signal processing of the audio signal processing unit 130 becomes smaller than the predetermined threshold value, it is possible to prevent the actual conversation from being established. In addition, since the execution time of the audio signal processing of the audio signal processing unit 130 is minimized, the processing load of the audio signal processing system 100 can be minimized.

  Similar to the audio signal processing parameter determination unit 1700 of the second embodiment, the audio signal processing parameter determination unit 2200 illustrated in FIG. 22 continues until the parameter adjustment end condition is satisfied even if the individuality degree 330 satisfies the individuality degree condition 1001. The audio signal processing parameter 1003 is adjusted. However, the audio signal processing parameter determination unit 220 shown in FIG. 22 predicts the delay time of the audio signal processing based on the audio signal processing parameter 1003 corresponding to the individuality degree 330 when the individuality degree 330 satisfies the individuality degree condition 1001. If the value 2212 is a predetermined threshold value, the audio signal processing parameter 1003 and the estimated delay time value 2212 may be input to the audio signal processing unit 130.

  FIG. 23 is a graph showing the relationship between the delay time and the amount of change in audio data due to audio signal processing according to the third embodiment of the present invention.

  The delay time shown in FIG. 23 increases in proportion to the change amount of the average F0 value. This is because a large amount of arithmetic processing must be performed to execute the audio signal processing in which the amount of change in the average F0 value is large, and the time required for the audio signal processing increases and the delay time also increases. On the other hand, since the average F0 value can be executed with a small amount of arithmetic processing to execute the audio signal processing with a small change amount, the time required for the audio signal processing is shortened and the delay time is also shortened. For this reason, when the amount of change in the average F0 value is 0, nothing is processed, so that the delay time is minimized.

  In addition, this invention is not limited to embodiment mentioned above, Various modifications are included. For example, the above-described embodiment has been described in detail for easy understanding of the present invention, and is not necessarily limited to one having all the configurations described.

DESCRIPTION OF SYMBOLS 110 Voice personality feature analysis part 111 Voice feature-value calculation part 112 Personality feature-value calculation part 113 Personality degree calculation part 120 Voice signal processing parameter determination part 121 Personality degree condition determination part 122 Voice signal processing parameter adjustment part 123 Personality Degree recalculation unit 130 speech signal processing unit 131 personality feature removal unit 132 speech synthesis unit

Claims (10)

In an audio signal processing method executed on input audio data,
The method
Executed in a system comprising a CPU, a storage area, and an interface;
A feature amount calculating step of calculating at least one feature amount obtained by quantifying the input voice data;
Based on the feature amount calculated by the feature Ryosan out step, and personal property calculation step of calculating a quantified individuality of the person of the intensity of the characteristic quantity,
If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing step for executing voice signal processing on the input voice data such that the personality degree satisfies the personality degree condition; only including,
The audio signal process is a process of changing the audio data,
In the method, when there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination step of determining a change amount of the audio data such that the personality degree satisfies the personality degree condition. In addition,
In the audio signal processing step, audio signal processing for changing the audio data based on the amount of change determined in the change amount determination step is executed.
In the change amount determining step,
A process of determining whether or not the characteristic amount of the changed sound data satisfies the personality degree condition while changing the change amount of the sound data in a predetermined amount unit, and the state after the change is a predetermined end condition Run repeatedly until
When it is determined that the characteristic amount of the voice data after the change satisfies the personality condition, the amount of deterioration of the sound quality of the voice data due to the amount of change is estimated, and the amount of change with the least amount of deterioration is determined as the minimum deterioration change. As a quantity, stored in the storage area,
When the state after the change satisfies the termination condition, the minimum deterioration change amount stored in the storage area is determined as the change amount of the audio data .   In an audio signal processing method executed on input audio data,
  The method
  Executed in a system comprising a CPU, a storage area, and an interface;
  A feature amount calculating step of calculating at least one feature amount obtained by quantifying the input voice data;
  A personality degree calculating step for calculating a personality degree obtained by quantifying the strength of the individuality of the feature quantity based on the feature quantity calculated in the feature quantity calculation step;
  If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing step for executing voice signal processing on the input voice data such that the personality degree satisfies the personality degree condition; Including
  The audio signal process is a process of changing the audio data,
  In the method, when there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination step of determining a change amount of the audio data such that the personality degree satisfies the personality degree condition. In addition,
  In the audio signal processing step, audio signal processing for changing the audio data based on the amount of change determined in the change amount determination step is executed.
  The storage area stores, for each of a plurality of feature quantities obtained by quantifying the input voice data, personality degree calculation information used for calculating the individuality degree of the feature quantity from the feature quantity,
  The individuality degree calculation information indicates that, for each of the plurality of feature amounts, the individuality degree of the feature amount is smaller as the value of the number distribution of the feature amount is larger,
  The individuality degree indicates that the greater the value, the stronger the individuality of the feature amount used for calculating the individuality degree,
  In the individuality degree calculation step, based on the feature quantity calculated in the feature quantity calculation step and the individuality degree calculation information, the individuality degree of each feature quantity calculated in the feature quantity calculation step is calculated. Calculate
  In the change amount determining step,
  If there is a personality degree that is equal to or greater than the first threshold, it is determined that there is a personality degree that does not satisfy the personality degree condition,
  An audio signal processing method, wherein the change amount of the audio data is determined so that the individuality degree is smaller than the first threshold value.   The system is connected to an upload terminal for uploading the audio data and a download terminal for downloading the audio data,
  A user who views the audio data via the download terminal is set with a personality authority associated with the personality condition.
  In the change amount determination step, the minimum deterioration change amount is determined as a change amount of the audio data for each individuality degree condition corresponding to an individuality authority that can be set,
  In the audio signal processing step,
  For each amount of change determined by the change amount determination step, execute the audio signal processing on the audio data,
  The audio signal processing method according to claim 1, wherein the audio data after the audio signal processing is stored in the storage area for each individuality authority.   The method
  When the system receives a download request from the download terminal, identifying the personality authority set for the user of the download terminal that transmitted the download request;
  Retrieving audio data corresponding to the specified individuality authority from the audio data after the audio signal processing registered in the storage area;
  The audio signal processing method according to claim 3, further comprising: transmitting the searched audio data to a download terminal that has transmitted the download request.   In an audio signal processing method executed on input audio data,
  The method
  Executed in a system comprising a CPU, a storage area, and an interface;
  A feature amount calculating step of calculating at least one feature amount obtained by quantifying the input voice data;
  A personality degree calculating step for calculating a personality degree obtained by quantifying the strength of the individuality of the feature quantity based on the feature quantity calculated in the feature quantity calculation step;
  If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing step for executing voice signal processing on the input voice data such that the personality degree satisfies the personality degree condition; Including
  The audio signal process is a process of changing the audio data,
  In the method, when there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination step of determining a change amount of the audio data such that the personality degree satisfies the personality degree condition. In addition,
  In the audio signal processing step, audio signal processing for changing the audio data based on the amount of change determined in the change amount determination step is executed.
  In the change amount determining step,
  A process of determining whether or not the characteristic amount of the changed sound data satisfies the personality degree condition while changing the change amount of the sound data in a predetermined amount unit, and the state after the change is a predetermined end condition Run repeatedly until
  When it is determined that the feature amount of the audio data after the change satisfies the personality degree condition, an execution time required to execute the audio signal processing based on the change amount is predicted,
  When the predicted execution time is smaller than a second threshold value, the change amount of the execution time is determined as the change amount of the sound data.   In an audio signal processing system that executes audio signal processing on input audio data,
  A feature amount calculation unit for calculating at least one feature amount obtained by quantifying the input voice data;
  Based on the feature amount calculated by the feature amount calculation unit, a personality degree calculation unit that calculates a personality degree that quantifies the strength of the individuality of the feature amount;
  If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing unit that performs audio signal processing on the input voice data such that the individuality degree satisfies the personality degree condition; With
  The audio signal process is a process of changing the audio data,
  When there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination unit that determines a change amount of the audio data such that the personality degree satisfies the personality degree condition,
  The audio signal processing unit executes audio signal processing for changing the audio data based on a change amount determined by the change amount determining unit,
  The change amount determination unit
  A process of determining whether or not the characteristic amount of the changed sound data satisfies the personality degree condition while changing the change amount of the sound data in a predetermined amount unit, and the state after the change is a predetermined end condition Run repeatedly until
  When it is determined that the characteristic amount of the voice data after the change satisfies the personality condition, the amount of deterioration of the sound quality of the voice data due to the amount of change is estimated, and the amount of change with the least amount of deterioration is determined as the minimum deterioration change. The amount is stored in a storage area provided for the audio signal processing,
  When the state after the change satisfies the termination condition, the audio signal processing system determines the minimum deterioration change amount stored in the storage area as the change amount.   In an audio signal processing system that executes audio signal processing on input audio data,
  A feature amount calculation unit for calculating at least one feature amount obtained by quantifying the input voice data;
  Based on the feature amount calculated by the feature amount calculation unit, a personality degree calculation unit that calculates a personality degree that quantifies the strength of the individuality of the feature amount;
  If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing unit that performs audio signal processing on the input voice data such that the individuality degree satisfies the personality degree condition; With
  The audio signal process is a process of changing the audio data,
  When there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination unit that determines a change amount of the audio data such that the personality degree satisfies the personality degree condition,
  The audio signal processing unit executes audio signal processing for changing the audio data based on a change amount determined by the change amount determining unit,
  For each of the plurality of feature quantities obtained by quantifying the input voice data, the storage area stores personality degree calculation information used for calculating the individuality degree of the feature quantity from the feature quantity;
  The individuality degree calculation information indicates that, for each of the plurality of feature amounts, the individuality degree of the feature amount is smaller as the value of the number distribution of the feature amount is larger,
  The individuality degree indicates that the greater the value, the stronger the individuality of the feature amount used for calculating the individuality degree,
  The individuality degree calculation unit calculates the individuality degree of each feature amount calculated by the feature amount calculation unit based on the feature amount calculated by the feature amount calculation unit and the individuality degree calculation information. ,
  The change amount determination unit
  If there is a personality degree equal to or greater than the first threshold, it is determined that there is a personality degree that does not satisfy the personality degree condition,
  An audio signal processing system, wherein the change amount of the audio data is determined so that the individuality degree is smaller than the first threshold value.   The audio signal processing system includes:
  Connected to an upload terminal for uploading the audio data and a download terminal for downloading the audio data;
  A user who views the audio data via the download terminal is set with a personality authority associated with the personality condition.
  The change amount determination unit determines the minimum deterioration change amount as a change amount of the audio data for each individuality condition corresponding to the individuality authority set for the user,
  The audio signal processor is
  For each change amount determined by the change amount determination unit, the audio signal processing is performed on the audio data,
  The audio signal processing system according to claim 6, wherein the audio data after the audio signal processing is stored in the storage area for each individuality authority.   When a download request is received from the download terminal, a personality authority specifying unit that specifies the personality authority set for the user of the download terminal that transmitted the download request;
  The audio data corresponding to the specified individuality authority is searched from the audio data after the audio signal processing stored for each individuality authority in the storage area, and the search is performed by the download terminal that transmitted the download request. The voice signal processing system according to claim 8, further comprising: a voice search unit that transmits the voice data.   In an audio signal processing system that executes audio signal processing on input audio data,
  A feature amount calculation unit for calculating at least one feature amount obtained by quantifying the input voice data;
  Based on the feature amount calculated by the feature amount calculation unit, a personality degree calculation unit that calculates a personality degree that quantifies the strength of the individuality of the feature amount;
  If there is a personality degree that does not satisfy the predetermined personality degree condition, an audio signal processing unit that performs audio signal processing on the input voice data such that the individuality degree satisfies the personality degree condition; With
  The audio signal process is a process of changing the audio data,
  When there is a personality degree that does not satisfy the predetermined personality degree condition, a change amount determination unit that determines a change amount of the audio data such that the personality degree satisfies the personality degree condition,
  The audio signal processing unit executes audio signal processing for changing the audio data based on a change amount determined by the change amount determining unit,
  The change amount determination unit
  A process of determining whether or not the characteristic amount of the changed sound data satisfies the personality degree condition while changing the change amount of the sound data in a predetermined amount unit, and the state after the change is a predetermined end condition Run repeatedly until
  When it is determined that the feature amount of the voice data after the change satisfies the personality degree condition, an execution time required to execute the voice signal processing based on the change amount is predicted,
  When the predicted execution time is smaller than a second threshold value, the change amount of the execution time is determined as the change amount of the sound data.