JP2019086801A - Audio processing method and audio processing apparatus - Google Patents

Abstract

To facilitate setting of variables for controlling the quality of voices.SOLUTION: A feature quantity identifying unit 22 identifies feature quantities of object voices by analyzing a voice signal X. A section setting unit 24 sets a processable section Q correspondingly to the result of comparison of the feature quantities identified by the feature quantity identifying unit 22 and a threshold. A variables control unit 26 sets a control variable C for controlling voice quality regarding the processable section Q. A voice processing unit 28 generates a voice signal Y of a voice, resulting from control of voice quality in the processable section Q out of the object voices, corresponding to the control variable C.SELECTED DRAWING: Figure 2

Description

  The present invention relates to technology for controlling voice quality of speech.

  Techniques for controlling voice quality of speech have been conventionally proposed. For example, Patent Document 1 discloses a configuration in which voice conversion parameters for controlling voice quality of synthetic speech are temporally changed according to an instruction from a user.

Japanese Patent Application Publication No. 2004-038071

  However, it is practically difficult for the user to properly adjust the voice conversion parameters so that the speech of the desired voice quality that is aurally natural is reproduced. The problem is particularly acute for users who do not have sufficient expertise in speech and voice quality. In view of the above circumstances, the present invention aims to facilitate setting of variables for controlling voice quality of speech.

  In order to solve the above problems, the speech processing apparatus according to the present invention comprises: feature quantity specifying means for specifying a feature quantity of the target speech; and section setting means for setting a processing section according to the comparison result of the feature quantity and the threshold value. And variable control means for setting a control variable for controlling voice quality for the processing section, and voice processing means for generating a voice signal of a voice in which voice quality of the processing section of the target voice is controlled according to the control variable. Do. In the above configuration, the voice quality of the processing section set according to the feature amount of the target voice is controlled. Therefore, even when the user does not have specialized knowledge on voice quality (for example, knowledge of the section of the target voice to be converted to a specific voice quality), it is possible to reproduce natural voice quality audibly.

  In a preferred aspect of the present invention, the feature amount specifying means specifies an elapsed time from a start point in a specific section of the target speech as a feature amount. For example, the section setting means sets a section where the elapsed time exceeds the threshold for the first voice quality as a processing section, and sets a section where the elapsed time falls below the threshold for the second voice quality different from the first voice quality. Do. In the above aspect, a section where the elapsed time exceeds the threshold (for example, a section on the end side of the voiced section) and a section where the elapsed time is less than the threshold (for example, the section on the head of the voiced section) Set as Therefore, there is an advantage that it is possible to reproduce a plurality of voice qualities naturally audible.

  In the configuration in which the elapsed time of the specific section of the target voice is specified as the feature quantity, the feature quantity specifying unit specifies, for example, the pitch or volume of the target voice as the feature quantity, and the section setting unit determines the pitch of the target voice Alternatively, the processing section is set according to the comparison result of the volume and the first threshold and the comparison result of the elapsed time and the second threshold. In the above aspect, since the pitch or volume of the target voice is applied to the setting of the processing section in addition to the elapsed time, the above-described effect of being able to generate a voice with natural voice quality audibly is particularly remarkable. Further, according to the configuration in which the feature amount specifying unit divides a specific section with the point at which the target voice changes in pitch or volume as a boundary, for example, the processing section can be set according to the elapsed time of sound generation for each note (for example, A section on the end side or the top side of each note can be set as a processing section).

  In a preferred aspect of the present invention, the section setting means sets a processing section in accordance with the feature amount of the target voice in the automatic setting mode, and sets a processing section in accordance with an instruction from the user in the manual setting mode. In the above aspect, since the automatic setting mode and the manual setting mode are prepared, for example, a user who has sufficient knowledge of voice quality can reproduce his / her desired voice quality in the manual setting mode, and the knowledge about voice quality is insufficient. Some users have the advantage of being able to audibly reproduce natural voice quality in the automatic setting mode.

  In a preferable aspect of the present invention, the section setting means sets a processing section according to a comparison result of a feature amount according to an instruction from a user among a plurality of types of feature amounts and a threshold. In the above aspect, the feature quantity according to the instruction from the user among the plurality of feature quantities is applied to setting of the processing section, so that there is an advantage that voice quality adapted to the user's intention and preference can be reproduced.

  In a preferred aspect of the present invention, the section setting means variably sets the threshold according to an instruction from the user. In the above aspect, since the threshold value to be compared with the feature value is variably set according to the instruction from the user for setting the processing section, the user is compared with the configuration in which the threshold value is fixed to a predetermined value. There is an advantage that it is possible to reproduce a voice in which the voice quality of the processing section reflecting the intention and preference of the voice is controlled.

  The configuration for the feature amount specifying unit to specify the feature amount is arbitrary. For example, a configuration for specifying a feature amount by analysis of an audio signal of a target sound, or a configuration for specifying a feature amount from music data specifying each note of a music corresponding to the target sound is adopted. According to the configuration for analyzing the audio signal, there is an advantage that the feature amount of the target voice can be specified accurately, and according to the configuration using the music data, there is an advantage that the feature amount of the target voice can be specified easily. Note that, in the first analysis mode, the feature amount specifying unit specifies the feature amount by analysis of the audio signal of the target sound, and in the second analysis mode, music data specifying each note of the music corresponding to the target sound. A configuration for specifying feature quantities from the above is also suitable.

  In a preferred aspect of the present invention, the feature amount specifying unit specifies a feature amount from synthesis data instructing synthesis of a target voice, and the speech processing unit is a speech synthesis process to which the synthesis data is applied. An audio signal of an audio controlled according to the control variable is generated. The above aspect has an advantage that it is possible to generate a voice signal of voice in which the voice quality of the processing section is controlled without requiring the voice signal of the target voice.

  The voice processing device according to each of the above aspects is realized by hardware (electronic circuit) such as DSP (Digital Signal Processor) dedicated to voice processing, and is a general-purpose arithmetic processing device such as CPU (Central Processing Unit) It is also realized by the collaboration of the program and the program. The program of the present invention may be provided in the form of being stored in a computer readable recording medium and installed on the computer. The recording medium is, for example, a non-transitory recording medium, and is preferably an optical recording medium (optical disc) such as a CD-ROM, but any known medium such as a semiconductor recording medium or a magnetic recording medium may be used. Recording media of the form Also, for example, the program of the present invention may be provided in the form of distribution via a communication network and installed on a computer. The present invention is also specified as an operation method (voice processing method) of the voice processing device according to each aspect described above.

It is a block diagram of the speech processing unit concerning a 1st embodiment of the present invention. It is a functional block diagram of a speech processing unit. It is a flowchart of the voice analysis process which a feature-value identification part performs. It is explanatory drawing of operation | movement of a speech processing apparatus. It is a flowchart of operation | movement of an audio processing apparatus. It is a schematic diagram of a threshold value setting screen. It is a schematic diagram of the operation mode selection screen in 2nd Embodiment. It is explanatory drawing of the setting of the process area (vocal fly) in 2nd Embodiment. It is a schematic diagram of the feature-value selection screen in 3rd Embodiment. It is a schematic diagram of the threshold value setting screen in 3rd Embodiment. It is a functional block diagram of the speech processing unit in a 4th embodiment. It is a functional block diagram of the speech processing unit in a 5th embodiment. It is a schematic diagram of the operation mode selection screen in 5th Embodiment. It is a schematic diagram of the feature-value selection screen in 5th Embodiment. It is a schematic diagram of the threshold value setting screen in 5th Embodiment. It is a functional block diagram of the speech processing unit in a 6th embodiment. It is a functional block diagram of the speech processing unit in a 7th embodiment. It is explanatory drawing of operation | movement of the speech processing unit in 7th Embodiment. It is a flowchart of operation | movement of 7th Embodiment. It is explanatory drawing of operation | movement of the speech processing unit in 8th Embodiment. It is a flowchart of operation | movement of 8th Embodiment. It is a flowchart of operation | movement of 8th Embodiment.

First Embodiment
FIG. 1 is a block diagram of an audio processing apparatus 100 according to the first embodiment of the present invention. As illustrated in FIG. 1, a signal supply device 200 is connected to the audio processing device 100. The signal supply device 200 supplies the audio processing device 100 with an audio signal X representing a waveform of audio to be processed by the audio processing device 100 (hereinafter referred to as “target audio”). The target voice of the first embodiment is a singing voice that sings a specific music (hereinafter, referred to as “target music”). A sound pickup device that picks up surrounding sound to generate a sound signal X, a reproduction device that obtains the sound signal X from a portable or built-in recording medium and supplies it to the sound processing device 100, sound from a communication network A communication device that receives the signal X and supplies it to the speech processing device 100 may be suitably adopted as the signal supply device 200. It is also possible to configure the signal supply device 200 integrally with the voice processing device 100.

  The voice processing device 100 is a signal processing device that generates the voice signal Y by adjusting the voice quality of the target voice represented by the voice signal X supplied from the signal supply device 200. In the first embodiment, the case of converting the target voice of the voice signal X into breathy sound (breathy) is illustrated. The breath sound is a voice with abundant breathiness (spear voice), and each harmonic on the frequency axis with respect to the harmonic component (basic sound component and multiple harmonic components) caused by vibration of vocal cords The sound component present in the gap of the component means a relatively dominant voice.

  As illustrated in FIG. 1, the audio processing device 100 is realized by a computer system including an arithmetic processing unit 10, a storage device 12, a display device 14, an operation device 16, and a sound emitting device 18. Arithmetic processing unit 10 executes various control processing and arithmetic processing by executing a program stored in storage device 12. The storage unit 12 stores programs executed by the arithmetic processing unit 10 and various data used by the arithmetic processing unit 10. A well-known recording medium such as a semiconductor recording medium or a magnetic recording medium, or a combination of a plurality of types of recording medium is arbitrarily adopted as the storage device 12.

  The display device 14 (for example, a liquid crystal display panel) displays an image instructed from the arithmetic processing unit 10. The operation device 16 is an input device operated by the user for various instructions to the voice processing device 100. In addition to a plurality of operators pressed by the user, a touch panel integrally formed with the display device 14 can be adopted as the operation device 16. The sound emitting device 18 (for example, a speaker or a headphone) reproduces a voice corresponding to the voice signal Y generated by the arithmetic processing unit 10 (that is, a voice obtained by converting the voice quality of the target voice). The illustration of a D / A converter for converting the audio signal Y from digital to analog and an amplifier for amplifying the audio signal Y are omitted for convenience.

  FIG. 2 is a functional block diagram of the speech processing apparatus 100 according to the first embodiment. As exemplified in FIG. 2, the arithmetic processing unit 10 executes a program stored in the storage unit 12 to generate a plurality of functions (a feature amount specifying unit 22 for generating an audio signal Y from the audio signal X). The section setting unit 24, the variable control unit 26, and the voice processing unit 28) are realized. A configuration in which each function of the arithmetic processing unit 10 is distributed to a plurality of units or a configuration in which a dedicated electronic circuit (for example, a DSP) realizes a part of the functions of the arithmetic processing unit 10 may be adopted.

  The feature amount specifying unit 22 sequentially specifies the feature amounts of the target sound. The feature quantity specifying unit 22 of the first embodiment analyzes the audio signal X supplied from the signal supply device 200 to extract the pitch (pitch) P and the elapsed time E of the target audio sequentially. The pitch P is set to any one of a plurality of discrete pitches (for example, a plurality of pitches constituting a scale). The elapsed time E means an elapsed time from the start point in a section (hereinafter referred to as a "voiced section") in which a voiced sound is present in the target voice. Therefore, as the duration of the voiced section is longer, the elapsed time E increases to a larger value from the start to the end of the voiced section. The voiced section is a section in which the harmonic structure of the voiced sound in which each harmonic component is arranged at substantially equal intervals on the frequency axis (a silent section in which a clear harmonic structure is not observed and a silent section in which no voice exists). Section excluded).

  FIG. 3 is a flowchart of an operation (hereinafter referred to as “voice analysis process”) in which the feature amount specifying unit 22 of the first embodiment specifies a feature amount (pitch P, elapsed time E), and FIG. It is explanatory drawing of operation | movement of an apparatus. The speech analysis process of FIG. 3 is sequentially performed for each unit section (frame) in which the speech signal X is divided on the time axis. FIG. 4 exemplarily shows a schematic waveform of the audio signal X of the target voice pronounced “Saita”.

  When the voice analysis process is started, the feature amount specifying unit 22 extracts the pitch p0 in the unit section of the voice signal X (SA1). The pitch p0 is the fundamental frequency (pitch) of the audio signal X. The time change of the pitch p0 is also shown in the waveform of the audio signal X in FIG. A known technique (pitch extraction technique) is arbitrarily adopted to extract the pitch p0 of the audio signal X.

  The feature amount identifying unit 22 determines whether the unit section corresponds to a voiced section (SA2). As illustrated in FIG. 4, significant pitch p0 is extracted in the voiced section v0 in which a clear harmonic structure is observed, whereas it is significant in sections other than the voiced section v0 (silence section and silent section). There is a tendency that the pitch p0 is not extracted. In consideration of the above tendency, the feature amount specifying unit 22 of the first embodiment determines whether the unit section is included in the voiced section v0 according to whether or not the significant pitch p0 is extracted in step SA1. Determine if

  When the unit section corresponds to the voiced section v0 (SA2: YES), the feature amount identifying unit 22 adds a predetermined value (for example, 1) to the elapsed time e0 (SA3). On the other hand, when the unit section does not correspond to the voiced section v0 (SA2: NO), the feature amount identifying unit 22 initializes the elapsed time e0 to zero (SA4). Therefore, as understood from FIG. 4, the elapsed time e0 is set to zero at the start of the voiced section v0 and increases with the passage of time in the voiced section v0, and becomes the end point (SA2: NO) of the voiced section v0. Is initialized to zero.

  The feature amount specifying unit 22 determines the pitch P by normalizing the pitch p0 of the audio signal X (SA5). Specifically, as illustrated in FIG. 4, the pitch closest to the pitch p0 among the plurality of pitches set discretely is specified as the pitch P after normalization. As understood from the above description, the pitch P is maintained at a constant value in one note of the target music and can be discretely fluctuated for each note. Therefore, the point in time when the pitch P changes on the time axis is likely to correspond to the boundary between adjacent notes in the target music.

  The feature amount specifying unit 22 normalizes the elapsed time e0 of each voiced section v0 to the elapsed time E of the voiced section V corresponding to each note of the target music (SA6). Specifically, as understood from FIG. 4, the feature quantity specifying unit 22 sets the voiced section v0 as the target music with the point at which the pitch P of the audio signal X changes (that is, the boundary of each successive note). The elapsed time e from the start point of each voiced section V is calculated by dividing the voiced section V into the voiced section V for each note and setting (correcting) the elapsed time e0 to be zero at the start point of the voiced section V. Therefore, the elapsed time E is set to zero at the start point of the voiced segment V for each note of the target music and increases with time in the voiced segment V, and is initialized to zero when the end point of the voiced segment V arrives. . The elapsed time E can also be expressed as a length of time (duration) in which one note of the target music continues. The feature amount specifying unit 22 of the first embodiment sequentially specifies the feature amounts (pitch P and elapsed time E) of the audio signal X for each unit section by repeating the voice analysis process exemplified above.

  The section setting unit 24 in FIG. 2 sets a processing section Q in accordance with the feature amount (pitch P, elapsed time E) specified by the feature amount specifying unit 22. The processing section Q is a section in which the voice quality of the target voice of the voice signal X should be changed (a section of the target voice to be converted to the breath sound). The section setting unit 24 of the first embodiment sets a processing section Q according to the comparison result of the feature amount (pitch P, elapsed time E) specified by the feature amount specifying unit 22 and the threshold value. Specifically, as illustrated in FIG. 4, the section setting unit 24 sets the processing section Q according to the comparison result of the pitch P and the threshold value PTH and the comparison result of the elapsed time E and the threshold value ETH. In actual singing, a general tendency is observed that as the pitch of the singing voice is higher and the duration is longer, the spitness of the singing voice is likely to increase. From the viewpoint of reproducing the above tendency, the section setting unit 24 of the first embodiment processes the section in which the pitch P exceeds the threshold PTH and the elapsed time E exceeds the threshold ETH, as illustrated in FIG. 4. Set as section Q. Since the elapsed time E monotonously increases with time in the voiced section V, a section on the end side of the voiced sections V whose duration exceeds the threshold ETH is defined as the processing section Q. The threshold value PTH and the threshold value ETH are variably set in accordance with an instruction from the user to the operation device 16.

  The variable control unit 26 in FIG. 2 sets a control variable C for the processing section Q set by the section setting unit 24. The control variable C is a variable for controlling voice quality. The control variable C in the first embodiment is a variable that indicates the degree of inspiratory sound. As illustrated in FIG. 4, the variable control unit 26 sets the control variable C so as to increase at a predetermined increase rate from zero from the start point to the end point of the processing section Q set by the section setting unit 24. That is, the variable control unit 26 temporally changes the control variable C so that the degree of the breath sound increases as the end point of the processing section Q is approached (the sound of one note is prolonged).

  The voice processing unit 28 in FIG. 2 generates a voice signal Y by executing voice quality conversion processing to which the control variable C set by the variable control unit 26 is applied to the voice signal X. Voice quality conversion processing is voice processing that changes the voice quality of the target voice according to the control variable C. The voice processing unit 28 according to the first embodiment converts the voice signal X in the processing section Q into a breath sound of a degree according to the control variable C (provides a breath characteristic of a degree according to the control variable C) In the processing), an audio signal Y is generated. A well-known technique is arbitrarily employ | adopted for giving breathability. For example, the voice processing unit 28 separates the voice signal X into a harmonic component and an inharmonic component (breath component), and the intensity of the inharmonic component with respect to the harmonic component (that is, breath characteristic) according to the control variable C. As a result, the voice signal Y of the voice in which the processing section Q is converted into the breath sound according to the control variable C is generated.

  FIG. 5 is a flowchart of processing in which the arithmetic processing unit 10 generates an audio signal Y from the audio signal X. For example, in response to an instruction from the user to the operation device 16, the process of FIG. 5 is started, and is repeated for each unit section over the entire section of the audio signal X.

  When the audio signal X of one unit section is taken in from the signal supply device 200 (SB1), the section setting unit 24 variably sets the threshold value PTH and the threshold value ETH according to an instruction from the user to the operation device 16 To do (SB2). Specifically, the arithmetic processing unit 10 causes the display unit 14 to display the setting screen of FIG. 6 (hereinafter referred to as the “threshold setting screen”). The threshold setting screen is an image for the user to designate the threshold PTH of the pitch P (Pitch) and the threshold ETH of the elapsed time E (Duration). The user can arbitrarily adjust the threshold PTH and the threshold ETH by appropriately operating the operation device 16 while visually recognizing the threshold setting screen.

  The feature amount specifying unit 22 specifies the pitch P of the unit section and the elapsed time E by executing the speech analysis process described with reference to FIG. 3 (SB3). Then, the section setting unit 24 determines whether the pitch P of the unit section exceeds the threshold PTH (SB4) and determines whether the elapsed time E of the unit section exceeds the threshold ETH (SB5) . When the result of both step SB4 and step SB5 is affirmative (P> PTH, E> ETH), the variable control unit 26 sets the control variable C for the unit section (SB6), and the voice processing unit 28 determines the variable A voice signal Y is generated from the voice signal X by voice quality conversion processing to which the control variable C set by the control unit 26 is applied (SB7). On the other hand, when the result of one or both of step SB4 and step SB5 is negative, the setting of control variable C (SB6) and the voice quality conversion process (SB7) for voice signal X are not executed. That is, the audio signal X supplied from the signal supply device 200 is output as the audio signal Y. As understood from the above description, the determination in step SB4 and step SB5 in FIG. 5 corresponds to the process in which the section setting unit 24 sets the processing section Q. By executing the process of FIG. 5 for each unit section of the audio signal X, an audio signal Y of an audio in which the processing section Q of the target audio is converted into an exhalation sound is generated.

  In the first embodiment described above, the voice quality of the processing section Q set according to the feature amount (pitch P, elapsed time E) of the target voice is controlled. Therefore, even when the user does not have specialized knowledge on voice quality (knowledge of the section to which breathiness is given in the target voice), it is possible to reproduce natural voice quality (breathing sound) audibly. . That is, there is an advantage that setting of the control variable C is facilitated (for example, designation of the processing section Q by the user and setting of time change of the control variable C are unnecessary).

Second Embodiment
The second embodiment of the present invention will be described below. In addition, about the element in which an effect | action and a function are the same as 1st Embodiment in each form illustrated below, the code | symbol referred by description of 1st Embodiment is diverted and detailed description of each is abbreviate | omitted suitably.

  The processing unit 10 according to the second embodiment causes the display unit 14 to display the setting screen (hereinafter referred to as the “operation mode selection screen”) shown in FIG. 7. The operation mode selection screen is an image for the user to select one of the manual setting mode (manual) and the automatic setting mode (auto). The automatic setting mode is an operation mode in which the processing section Q and the control variable C are set automatically (without requiring an instruction from the user to the operation device 16). That is, in the automatic setting mode, as in the first embodiment, the processing section Q and the control variable C in the processing section Q are automatically set according to the feature amount (pitch P, elapsed time E) of the target voice. Be done. On the other hand, the manual setting mode is an operation mode in which the processing zone Q and the control variable C are set in accordance with an instruction from the user to the operation device 16. That is, in the manual setting mode, the section setting unit 24 sets the section instructed by the user by the operation on the operation device 16 as the processing section Q, and the variable control unit 26 processes in accordance with the instruction from the user for the operation apparatus 16 The time change of the control variable C in the section Q is set.

  As illustrated in FIG. 7, the user can select one of the manual setting mode and the automatic setting mode for each of a plurality of types of voice quality (breath sound, vocal fly,...). That is, under any operation mode of the manual setting mode and the automatic setting mode, the processing section Q and the control variable C are individually set for each of a plurality of voice qualities. The vocal fly in FIG. 7 is a voice (edge voice) which is pronounced by repeating the closing and releasing of the vocal cords when singing in the low frequency range, and is typically pronounced immediately after the start of the utterance.

  FIG. 8 is an explanatory diagram of the operation of the section setting unit 24 when the automatic setting mode is set for the vocal fly. As exemplified in FIG. 8, the voiced section v0 corresponding to the pitch p0 of the target voice is set as the voiced section V, and the process of dividing the voiced section v0 into notes (normalization of the elapsed time e0) is omitted. . That is, for the vocal fly, the elapsed time e0 in the first embodiment corresponds to the elapsed time E.

  From the viewpoint of reproducing the above-mentioned tendency that vocal fly is likely to occur immediately after the start of vocalization in the low tone range, the section setting unit 24 of the second embodiment has the pitch P equal to the threshold PTH as illustrated in FIG. A section below which the elapsed time E (e0) falls below the threshold value ETH is set as a processing section Q in which the target voice is converted into a vocal fly. Since the elapsed time E monotonously increases with time, as understood from FIG. 8, a section on the head side of the voiced section V (section immediately after the start of sound generation) is defined as the processing section Q. The threshold value PTH and the threshold value ETH are individually set for each type of voice quality (for each of the breath sound and the vocal fly) in accordance with an instruction from the user to the operation device 16.

  As understood from the above description, the processing section Q differs depending on the type of voice quality. Specifically, for voice quality such as breath sounds that are likely to occur at the end of the utterance, a section where the elapsed time E exceeds the threshold ETH (that is, the section at the end of the voiced section V) is set as the processing section Q. As voice quality such as vocal fly which tends to occur immediately after the start, a section where the elapsed time E falls below the threshold ETH (that is, a section on the head side of the voiced section V) is set as the processing section Q.

  The variable control unit 26 sets the control variable C of the vocal fly to an effective value (for example, 1) inside the processing section Q, as illustrated in FIG. For example, it is set to 0). The voice processing unit 28 performs voice quality conversion processing of the processing section Q to which the control variable C is applied independently of each other for each of a plurality of types of voice quality. Although the specific process of converting the target voice into the vocal fly is optional, for example, a method of reducing the sampling frequency by resampling the audio signal X is preferably employed.

  Also in the second embodiment, the same effect as that of the first embodiment is realized. In the second embodiment, a section in which the elapsed time E exceeds the threshold ETH (a section at the end of the voiced section V) and a section in which the elapsed time E is below the threshold ETH (a section at the beginning of the voiced section V) It is set according to the kind of voice quality given to. Therefore, there is an advantage that it is possible to reproduce a plurality of voice qualities naturally audible. Further, in the second embodiment, since the automatic setting mode and the manual setting mode are prepared, a user who has sufficient knowledge about voice quality reproduces his / her desired voice quality in the manual setting mode and knowledge about voice quality. Users who do not have sufficient voice quality have the advantage of being able to audibly reproduce natural voice quality in the automatic setting mode.

Third Embodiment
In the first embodiment, the processing section Q is set according to the pitch P of the target voice and the elapsed time E, but the feature quantity applied to the setting of the processing section Q is not limited to the above example. For example, in addition to the pitch P and the elapsed time E, the volume (dynamics) D can be applied to the setting of the processing section Q. For example, in actual singing, there is a tendency that the lower the volume D, the easier it is to increase the breathiness of the singing voice. From the viewpoint of reproducing the above tendency, in addition to the conditions (P> PTH, E> ETH) regarding the pitch P and the elapsed time E, the section setting unit 24 is a section in which the condition that the volume D falls below the threshold DTH Is set as the processing section Q. Also, in actual singing, there is a tendency that as the volume D is smaller, vocal fly is more likely to occur. From the viewpoint of reproducing the above tendency, in addition to the conditions (P <PTH, E <ETH) regarding the pitch P and the elapsed time E, the section setting unit 24 is a section in which the condition that the volume D falls below the threshold DTH Is set as the processing section Q.

  FIG. 9 is a schematic view of a setting screen (hereinafter, referred to as a “feature amount selection screen”) displayed on the display device 14 in the third embodiment. The feature amount selection screen is an image for the user to select the feature amount applied to the setting of the processing section Q. Specifically, for each of a plurality of types of feature quantities (pitch P, elapsed time E, volume D), an enabled state (a state in which a check is added) and an invalid state are made according to an instruction from the user to the operation device 16 A state is selected. The section setting unit 24 compares one or more feature amounts designated by the user in the feature amount selection screen among the plurality of types of feature amounts with the threshold value (PTH, ETH, DTH) corresponding to the feature amounts. A processing zone Q is set according to the result. On the other hand, the feature amount set in the invalid state on the feature amount selection screen is not added to the setting of the processing section Q. As in the second embodiment, in the configuration in which multiple types of voice quality are added to the target voice, a separate feature value selection screen is displayed for each voice quality set in the automatic setting mode, and applied to the setting of the processing section Q. Combinations of feature quantities are individually selected for each voice quality.

  FIG. 10 is a schematic view of a threshold setting screen in the third embodiment. The threshold setting screen in FIG. 10 is an image for the user to set the threshold (PTH, ETH, DTH) for each of a plurality of types of feature quantities. As for the feature amounts set in the valid state on the feature amount selection screen of FIG. 9, a threshold value is set according to an instruction from the user to the operation device 16 as in the threshold value setting screen of FIG. On the other hand, with respect to the feature amount set in the invalid state on the feature amount selection screen, the change of the threshold value is prohibited on the threshold value setting screen. For example, with regard to the feature amount in the invalid state, the display on the threshold setting screen is displayed in gray (an aspect representing that it is excluded from the operation target).

  Also in the third embodiment, the same effect as that of the first embodiment is realized. Further, in the third embodiment, since each of the plurality of feature quantities is selectively applied to the setting of the processing section Q, the type of the feature quantity applied to the setting of the processing section Q is compared with a fixed configuration. There is an advantage of being able to reproduce various voice quality. In the third embodiment, in particular, since the feature quantity according to the instruction from the user among the plurality of feature quantities is applied to the setting of the processing section Q, it is possible to reproduce voice quality adapted to the user's intention and preference. Effect is realized. The configuration of the second embodiment is similarly applied to the third embodiment.

Fourth Embodiment
FIG. 11 is a functional block diagram of the arithmetic processing unit 10 of the speech processing apparatus 100 according to the fourth embodiment. As illustrated in FIG. 11, in the fourth embodiment, the audio signal X and the music data Z are supplied in parallel from the signal supply device 200 to the audio processing device 100. The music data Z is time series data for specifying a pitch (note number), an intensity (velocity), and a sound generation period (a start point and an end point) for each note constituting the music. For example, time series data conforming to the MIDI (Musical Instrument Digital Interface) standard is suitably used as the music data Z.

  The music data Z designates each note of the target music to be sung in the target sound represented by the audio signal X in time series. Therefore, each note of the target voice of the audio signal X and each note designated by the music data Z correspond to each other. In consideration of the above relationship, the feature amount specifying unit 22 of the fourth embodiment specifies the feature amounts (volume P, elapsed time E, volume D) of the target sound from the music data Z. Specifically, the feature amount specifying unit 22 specifies the pitch (note number) of each note designated by the music data Z as the pitch P of the target sound. Also, the feature amount specifying unit 22 specifies the intensity (velocity) of each note designated by the music data Z as the volume D, and specifies the elapsed time E from the sound generation period of each note. An operation in which the section setting unit 24 sets the processing section Q by applying the feature amount specified by the feature quantity specifying unit 22 and an operation in which the variable control unit 26 sets the control variable C in the processing section Q are the first embodiment It is similar. The voice processing unit 28 generates a voice signal Y from the voice signal X by voice quality conversion processing to which the control variable C is applied, as in the first embodiment.

  Also in the fourth embodiment, the same effect as that of the first embodiment is realized. Further, in the fourth embodiment, since the feature amount of the target voice is specified by referring to the music data Z, the feature amount is compared with the configuration of the first embodiment in which the feature amount is specified by analysis of the audio signal X. There is an advantage that the processing load required to identify the is reduced. On the other hand, according to the first embodiment in which the feature amount is specified by analysis of the audio signal X, the feature amount of the target voice can be identified more accurately than in the fourth embodiment in which the feature amount is estimated from the music data Z There is an advantage. The configurations of the second and third embodiments are also applied to the fourth embodiment.

Fifth Embodiment
FIG. 12 is a functional block diagram of the arithmetic processing unit 10 of the speech processing apparatus 100 according to the fifth embodiment. As understood from FIG. 12, in the fifth embodiment, the audio signal X and the music data Z are supplied in parallel from the signal supply device 200 to the audio processing device 100 as in the fourth embodiment. The feature amount identifying unit 22 of the fifth embodiment identifies the feature amounts (volume P, elapsed time E, volume D) of the target audio using one or both of the audio signal X and the music data Z. Specifically, when the operation mode of either the manual setting mode or the automatic setting mode is selected by the user as in the second embodiment and the automatic setting mode is selected, the first analysis mode and the second analysis mode are selected. Either of the analysis mode is selected by the user. The first analysis mode is an operation mode for specifying the feature quantities (pitch P, elapsed time E, volume D) of the target voice by analysis of the audio signal X as in the first embodiment, and the second analysis mode is As in the fourth embodiment, this is an operation mode for specifying the feature amount of the target sound from the music data Z.

  The processing unit 10 according to the fifth embodiment causes the display unit 14 to display the operation mode selection screen of FIG. 13. The operation mode selection screen of the fifth embodiment accepts selection of the manual setting mode (manual) and the automatic setting mode (auto) from the user as in the second embodiment (FIG. 7), and also selects the automatic setting mode. It is an image which accepts the selection of the 1st analysis mode and the 2nd analysis mode from the user about the voice quality which was carried out. As illustrated in FIG. 13, the user selects an operation mode (manual setting mode / automatic setting mode, first analysis mode / second analysis mode) for each of a plurality of types of voice quality (breathiness, vocal fly). It is possible.

  Specifically, for voice quality for which the user has selected the automatic setting mode, an operation image (check box) 42 for accepting selection between the first analysis mode and the second analysis mode can receive an instruction from the user. It is set to the enabled state. The user can select the second analysis mode (MIDI) by adding a check to the operation image 42, and can select the first analysis mode by canceling the check of the operation image 42. On the other hand, the operation image 42 corresponding to the voice quality for which the user has selected the manual setting mode is set to an invalid state (for example, gray out) in which the user's operation is not accepted.

  Further, for the voice quality set in the automatic setting mode, the arithmetic processing unit 10 causes the display device 14 to display the feature amount selection screen of FIG. 14 and the threshold setting screen of FIG. In FIG. 14 and FIG. 15, "audio" represents the audio signal X used to specify a feature in the first analysis mode, and "MIDI" represents music used to specify a feature in the second analysis mode. Data Z is represented. In addition, the pitch P (Pitch) specified from the audio signal X in the first analysis mode and the pitch P (Note Number) specified from the music data Z in the second analysis mode reflect the difference in significance between the two. And the notation is different. The same applies to the volume D (first analysis mode: Dynamics, second analysis mode: Velocity).

  The feature amount selection screen of FIG. 14 is configured to include a first area 51 corresponding to the first analysis mode (audio signal X) and a second area 52 corresponding to the second analysis mode (music data Z). . Each of the first area 51 and the second area 52 is an image for the user to select a feature to be applied to setting of the processing section Q, as in the example of FIG. 9. Specifically, the first area 51 is used to select the feature quantity (that is, the feature quantity specified from the audio signal X) to be applied to the setting of the processing section Q in the first analysis mode, and the second area 52 is The second analysis mode is used to select the feature amount (that is, the feature amount specified from the music data Z) to be applied to the setting of the processing section Q. In the state where the first analysis mode is selected on the operation mode selection screen of FIG. 13, the first area 51 is set to the valid state (the state for receiving an instruction from the user) and the second area 52 is disabled (use Set to not receive instructions from the On the other hand, in a state where the second analysis mode is selected on the operation mode selection screen of FIG. 13, as shown in FIG. 14, the second area 52 is set to the valid state and the first area 51 is set to the invalid state. Be done.

  The threshold setting screen of FIG. 15 is configured to include a first area 61 corresponding to the first analysis mode and a second area 62 corresponding to the second analysis mode. Each of the first area 61 and the second area 62 is an image for the user to set the threshold (PTH, ETH, DTH) applied to the setting of the processing section Q, as in the example of FIG. Specifically, the first area 61 receives the indication of the threshold applied in the first analysis mode, and the second area 62 receives the indication of the threshold applied in the second analysis mode. When the first analysis mode is selected, the first area 61 is set to the valid state, and when the second analysis mode is selected, the second area 62 is set to the valid state as illustrated in FIG. Ru. Regarding the feature amount ("Velocity" in the second area 62 in FIG. 15) set to the invalid state in the feature amount selection screen in FIG. It is similar to the example of FIG.

  Also in the fifth embodiment, the same effect as that of the first embodiment is realized. Further, in the fifth embodiment, the first analysis mode for specifying the feature amount of the target sound from the audio signal X and the second analysis mode for specifying the feature amount of the target sound from the music data Z are prepared. It has the advantage of being able to reproduce a variety of voice quality adapted to the intentions and preferences of the person. The configurations of the second to fourth embodiments are similarly applied to the fifth embodiment.

Sixth Embodiment
FIG. 16 is a functional block diagram of the arithmetic processing unit 10 of the speech processing apparatus 100 according to the sixth embodiment. As illustrated in FIG. 16, the arithmetic processing unit 10 of the sixth embodiment generates a speech signal Y using synthesis data S instructing synthesis of target speech. The synthetic data S is, for example, time-series data (for example, a file in VSQ format) for specifying a pitch, a sound generation period, and a sound generation content (lyric) for each note constituting a music. The composite data S is generated in response to an instruction from the user to the operation device 16 and stored in the storage device 12. It is also possible to supply the synthetic data S from the outside of the speech processing device 100.

  The feature amount specifying unit 22 of the sixth embodiment specifies the feature amount (volume P and elapsed time E) of the target voice from the synthetic data S. Specifically, the feature amount specifying unit 22 specifies the pitch P of the target voice according to the pitch of each note designated by the synthetic data S, and specifies the elapsed time E from the sound generation period of each note. The section setting unit 24 sets a processing section Q in accordance with the feature amount specified by the feature quantity specifying unit 22, and the variable control unit 26 sets a control variable C for the processing section Q set by the section setting unit 24.

  The speech processing unit 28 of the sixth embodiment generates a speech signal Y by speech synthesis processing to which the synthetic data S is applied. A known technique is arbitrarily adopted for the speech synthesis process. For example, it is estimated by an HMM (Hidden Markov Model) or an HMM (Hidden Markov Model) that is connected by adjusting the pitch and the sound generation period of each speech segment according to the pronunciation content specified by the synthetic data S and connecting them together. A statistical model-type speech synthesis process is preferably employed, which performs a filtering process according to the phonetic characters (phonemes) for the pitch. The voice processing unit 28 applies the control variable C set by the variable control unit 26 to the voice synthesis process, whereby the voice signal Y of the voice in which the voice quality of the processing section Q is controlled according to the control variable C is generated.

  Also in the sixth embodiment, the same effect as that of the first embodiment is realized. Further, in the sixth embodiment, since the feature amount of the target voice is specified with reference to the synthetic data S, there is an advantage that the voice signal X of the target voice is unnecessary. The configurations of the second to fifth embodiments can be applied to the sixth embodiment as well.

Seventh Embodiment
FIG. 17 is a functional block diagram of the arithmetic processing unit 10 of the speech processing apparatus 100 in the seventh embodiment, and FIG. 18 is an explanatory diagram of the operation of the arithmetic processing unit 10 in the seventh embodiment. As illustrated in FIG. 17, the arithmetic processing unit 10 of the seventh embodiment implements a feature amount specifying unit 22, a section setting unit 24, a variable control unit 26, an audio processing unit 28, and a reference sound analysis unit 72. The feature amount specifying unit 22 sequentially extracts the pitch p0 of the audio signal X as a feature amount of the target audio for each unit section.

  The reference sound analysis unit 72 analyzes the audio signal XREF of an exemplary or standard singing voice (hereinafter referred to as “reference voice”) recorded in advance for the target music. Specifically, the reference sound analysis unit 72 analyzes the speech signal XREF to extract the pitch pREF of the reference speech for each unit section, and the threshold value RH and the threshold value RL according to the pitch pREF of the reference speech. It is set variably for each unit section. As understood from FIG. 18, the threshold RH is set to a value above the pitch pREF, and the threshold RL is set to a number below the pitch pREF. For example, the reference sound analysis unit 72 calculates the threshold RH by adding a predetermined value (a positive number) to the pitch pREF, and calculates the threshold RL by subtracting a predetermined value from the pitch pREF. The reference sound analysis unit 72 sequentially specifies the pitch pREF of the reference sound from the music data designating the notes of the singing part of the target music (song music) in time series, and the threshold RH and threshold corresponding to the pitch pREF It is also possible to set RL.

  The section setting unit 24 in FIG. 17 sets, as a processing section Q, a section in which the pitch p0 of the target voice exceeds the threshold RH and a section in which the pitch p0 is below the threshold RL, as illustrated in FIG. That is, the processing section Q of the seventh embodiment is a section in which the pitch p0 of the target speech deviates from the pitch pREF of the reference speech. The variable control unit 26 sets a control variable C for each processing section Q set by the section setting unit 24. The control variable C in the seventh embodiment is a correction value for bringing the pitch p0 of the target speech in the processing section Q closer to the pitch pREF of the reference speech. Specifically, the variable control unit 26 calculates a difference value between the pitch p0 of the target voice and the threshold value RH or the threshold value RL as the control variable C for each unit section in the processing section Q.

  The voice processing unit 28 generates a voice signal Y by executing voice quality conversion processing (voice processing) to which the control variable C set by the variable control unit 26 is applied to the voice signal X. The voice processing unit 28 of the seventh embodiment generates a voice signal Y by processing of changing the pitch p0 of the voice signal X in the processing section Q by the control variable C (pitch conversion processing). Therefore, as exemplified by the broken line in FIG. 18, the voice p0 in the processing section Q of the voice signal X is corrected to the threshold value RH, and the voice maintained in the pitch p0 of the target voice outside the processing section Q A signal Y is generated. That is, the pitch p0 of the audio signal X is maintained in the section (outside of the processing section Q) where the pitch p0 of the audio signal X approximates the pitch pREF of the reference audio, and the pitch p0 is from the pitch pREF of the reference audio. In the diverging interval (in the processing interval Q), the pitch p0 approaches the pitch pREF of the reference speech.

  FIG. 19 is a flowchart of processing executed by the arithmetic processing unit 10 of the seventh embodiment for each unit section. When the process of FIG. 19 starts, the feature amount specifying unit 22 specifies the pitch p0 of the target sound by analysis of the sound signal X (SC1). Further, the reference sound analysis unit 72 specifies the pitch pREF of the reference sound by analysis of the sound signal XREF (SC2), and sets the threshold RH and the threshold RL according to the pitch pREF (SC3).

  The section setting unit 24 determines whether the pitch p0 of the target voice exceeds the threshold RH (SC4) and whether the pitch p0 is below the threshold RL (SC5). If the pitch p0 exceeds the threshold RH (SC4: YES), the variable control unit 26 calculates the difference between the pitch p0 and the threshold RH as the control variable C (SC6). Similarly, when the pitch p0 falls below the threshold value RL (SC5: YES), the variable control unit 26 calculates the difference between the pitch p0 and the threshold value RL as the control variable C (SC7). The audio processing unit 28 generates the audio signal Y having the threshold RH or the threshold RL as the pitch by changing the pitch p0 of the audio signal X by the control variable C (SC8). On the other hand, when the pitch p0 is a numerical value between the threshold RH and the threshold RL (SC4, SC5: NO), the setting of the control variable C (SC6, SC7) and the correction of the pitch p0 (SC8) are executed. Therefore, the voice signal X is made the voice signal Y while maintaining the pitch p0. Then, the audio processing unit 28 outputs the audio signal Y to the sound emission device 18 (SC9). As understood from the above description, the determination in step SC4 and step SC5 in FIG. 19 corresponds to the process in which the section setting unit 24 sets the processing section Q.

  In the seventh embodiment, in the processing section Q in which the pitch p0 of the audio signal X deviates from the pitch pREF of the reference speech, the pitch p0 of the target speech is corrected so as to approach the pitch pREF, while the pitch p0 In a section close to the pitch pREF of the reference voice, the pitch p0 is maintained. Therefore, even if the user does not have specialized knowledge on the section to which pitch p0 should be corrected (knowledge of the section to which pitch p0 should be corrected), the audibly natural voice quality of which the pitch is close to the reference speech It is possible to reproduce speech. On the other hand, in the section where the pitch p0 is close to the pitch pREF of the reference speech, the pitch p0 of the target speech is maintained, so the feature of the target speech (for example, variation of the pitch p0 unique to the singer) is lost. There is also an advantage that such an excessive correction can be avoided.

  In the above description, although the difference value between the pitch p0 of the target speech and the threshold RH or the threshold RL is calculated as the control variable C, the difference between the pitch p0 of the target speech and the pitch pREF of the reference speech is calculated. A configuration in which the pitch p0 in the processing section Q is corrected to the pitch pREF of the reference voice by calculating as the control variable C may also be adopted.

Eighth Embodiment
FIG. 20 is an explanatory diagram of an operation of the arithmetic processing unit 10 in the eighth embodiment. The arithmetic processing unit 10 of the eighth embodiment functions as the same elements as the seventh embodiment (a feature amount specifying unit 22, a section setting unit 24, a variable control unit 26, a voice processing unit 28, a reference sound analysis unit 72). .

  As illustrated in FIG. 20, the reference sound analysis unit 72 of the eighth embodiment specifies the pitch pREF of the reference sound as in the seventh embodiment, and further, a threshold RH_A and a threshold RH_B exceeding the pitch pREF, A threshold value RL_A and a threshold value RH_B below the pitch pREF are variably set according to the pitch pREF. The threshold RH_A exceeds the threshold RH_B, and the threshold RL_A is below the threshold RL_B. As understood from FIG. 20, the section setting unit 24 in the eighth embodiment processes a section from the time point T1 at which the pitch p0 of the target voice exceeds the threshold RH_A to the time T2 at which the pitch p0 falls below the threshold RH_B. Set as. That is, the threshold RH_A applied when the pitch p0 increases and the threshold RH_B applied when the pitch p0 decreases are different (hysteresis characteristics). Similarly, the section setting unit 24 sets, as a processing section Q, a section from the time when the pitch p0 of the target voice falls below the threshold RL_A to the time when the pitch p0 exceeds the threshold RL_B.

  FIG. 21 and FIG. 22 are flowcharts of processing executed by the processing unit 10 of the eighth embodiment for each unit section. The control information F exemplified in the following description is whether or not the unit section to be processed is included in the processing section Q (setting of the control variable C by the variable control unit 26 and pitch p0 by the voice processing unit 28). Information (flag) for identifying whether or not correction is in progress, and is initialized to the numerical value 0 which means that the unit section is not included in the processing section Q at the start of the first unit section Be done.

  When the process of FIG. 21 starts, the feature amount specifying unit 22 specifies the pitch p0 of the target sound (SD1) and the reference sound analysis unit 72 specifies the pitch pREF of the reference sound (SD2) according to the seventh embodiment Similarly, the section setting unit 24 determines whether the control information F is the numerical value 0 (SD3). If the control information F is the numerical value 0 (SD3: YES), the reference sound analysis unit 72 variably sets the threshold RH_A and the threshold RL_A according to the pitch pREF of the reference sound (SD4). For example, the reference sound analysis unit 72 calculates the threshold value RH_A by adding a predetermined value to the pitch pREF, and calculates the threshold value RL_A by subtracting a predetermined value from the pitch pREF.

  The section setting unit 24 determines whether the pitch p0 of the target voice exceeds the threshold RH_A (SD5) and whether the pitch p0 is below the threshold RL_A (SD6). If the pitch p0 exceeds the threshold RH_A (SD5: YES), the variable control unit 26 calculates a difference value between the pitch p0 and the threshold RH_A as a control variable (correction value) C (SD7). On the other hand, when the pitch p0 falls below the threshold value RL_A (SD6: YES), the variable control unit 26 calculates the difference between the pitch p0 and the threshold value RL_A as the control variable C (SD8). The voice processing unit 28 generates the voice signal Y having the threshold RH_A or the threshold RL_A as the pitch by changing the pitch p0 of the target voice by the control variable C (SD9). Further, the section setting unit 24 changes the control information F from numerical value 0 to numerical value 1 (SD10). The numerical value 1 of the control information F means that the pitch p0 of the target voice is being corrected. On the other hand, when the pitch p0 is a numerical value between the threshold RH_A and the threshold RL_A (SD5, SD6: NO), setting of the control variable C (SD7, SD8) and correction of the pitch p0 (SD9) are executed. I will not. The audio processing unit 28 outputs the processed audio signal Y exemplified above to the sound emitting device 18 (SD11).

  When the control information F is set to the numerical value 1 (SD10), the determination result in step SD3 is negative in the subsequent processing of the unit section. When the control information F is the numerical value 1 (SD3: NO), as illustrated in FIG. 22, the reference sound analysis unit 72 determines that the threshold RH_A and threshold RH_B above the pitch pREF of the reference voice and the threshold below the pitch pREF. RL_A and threshold value RL_B are set (SD20).

  The section setting unit 24 determines whether the pitch p0 of the target voice exceeds the threshold RH_B (SD21) and whether the pitch p0 is below the threshold RL_B (SD22). If the pitch p0 exceeds the threshold RH_B (SD21: YES) and if the pitch p0 falls below the threshold RL_B (SD22: YES), the correction of the pitch p0 is continued similarly to the immediately preceding unit section. Specifically, when the pitch p0 exceeds the threshold RH_B, the variable control unit 26 calculates the difference between the pitch p0 and the threshold RH_A as the control variable C (SD23), and the pitch p0 is the threshold RL_B. If the difference is below the threshold value p0, the difference between the pitch p0 and the threshold value RL_A is calculated as the control variable C (SD24). Then, the voice processing unit 28 generates the voice signal Y by changing the pitch p0 of the target voice by the control variable C (SD25).

  On the other hand, if the pitch p0 falls below the threshold RH_B (SD21: NO) and if the pitch p0 exceeds the threshold RL_B (SD22: N0), the processing section Q ends. That is, the setting of the control variable C (SD23, SD24) and the correction of the pitch p0 (SD25) are not performed, and the section setting unit 24 changes the control information F from numerical value 1 to numerical value 0 (SD26).

  As understood from the above description, in the eighth embodiment, a section from the time when the pitch p0 of the target voice exceeds the threshold RH_A to the time when it falls below the threshold RH_B, and the time when the pitch p0 falls below the threshold RL_A, the threshold RL_B The section up to the point in time is set as a processing section Q for correcting the pitch p0. Therefore, even if the pitch p0 fluctuates in the vicinity of each threshold (RH_A, RH_B, RL_A, RL_B), the presence or absence of correction for the pitch p0 does not change. That is, according to the eighth embodiment, the same effects as the seventh embodiment are realized, and the possibility that the presence or absence of correction for the pitch p0 of the target voice can be frequently switched in a short time can be reduced. is there.

  In the above description, although the pitch p0 of the target voice is corrected to the threshold RH_A or the threshold RL_A in the processing section Q, the configuration for correcting the pitch p0 to the threshold RH_B or the threshold RL_B in the processing section Q, or A configuration may also be employed in which the pitch p0 is corrected to the pitch pREF of the reference voice in the section Q. In addition, it is also possible to set the processing section Q and the control variable C while suppressing minute fluctuations of the pitch p0 of the target voice or the pitch pREF of the reference voice. For example, a low pass filter is preferably used to suppress minute fluctuations in the pitch p0 or the pitch pREF.

<Modification>
Each form described above can be variously modified. The aspect of a specific deformation | transformation is illustrated below. It is also possible to appropriately merge two or more aspects arbitrarily selected from the following exemplifications.

(1) The mode of change of the control variable C in the processing zone Q is arbitrary. For example, in each of the above-described embodiments, the configuration in which the control variable C linearly increases in the processing section Q is illustrated, but it is also possible to change the control variable C in a curvilinear manner (for example, non-linearly) in the processing section Q is there.

(2) The types of feature amounts specified by the feature amount specifying unit 22 are not limited to the above-described examples (pitch P, elapsed time E, volume D). For example, it is also possible to calculate the differential value (time change rate) or the second-order differential value of the feature quantity illustrated in each of the above-described embodiments as the feature quantity. Moreover, although any one of a plurality of discrete pitches is specified as the pitch P in the above-described embodiments, it is also possible to specify the pitch P (pitch curve) so as to change continuously in time. .

(3) In the above embodiments, the threshold (PTH, ETH, DTH) applied to the setting of the processing zone Q is variably set according to the instruction from the user, but the method of setting the threshold is arbitrary. . For example, according to a configuration in which the threshold value of the feature amount is set according to the past value of the feature amount specified by the feature amount specifying unit 22, or a numerical value calculated by statistical processing for the feature amount specified by the feature amount specifying unit 22. A configuration in which the threshold is set accordingly or a configuration in which the threshold of the feature is set according to the value of another feature may be adopted. However, the configuration in which the threshold is a variable value is not essential, and it is also possible to fix the threshold to a predetermined value. In addition, a configuration in which the upper limit value and the lower limit value of the range of feature amounts determined to correspond to the processing section Q are set (configuration in which the upper limit threshold and the lower limit threshold are separately set) or the processing section Q A configuration in which the range of the feature amount is divided into a plurality and set may also be adopted.

(4) In the configuration in which a plurality of types of feature quantities are applied to setting of the processing section Q, it is also possible to individually weight each feature quantity (set superiority or inferiority). For example, with respect to a unit section in which a feature amount having a large weight value exceeds a threshold, it is determined that the unit section corresponds to the processing section Q even when another feature amount is below the threshold.

(5) In each of the above-described embodiments, the elapsed time E is calculated from the start point of the voiced segment V, but the target of calculation of the elapsed time E is not limited to the voiced segment V. For example, it is also possible to calculate the elapsed time E from the start point of the section in which speech is present (hereinafter referred to as "voice section") without distinguishing between voiced and unvoiced. The voice section is a section other than the silent section in the target voice. In addition, for example, it is also possible to calculate the elapsed time E from the start point of a section in which a phoneme that can be pronounced continuously exists (hereinafter referred to as a "continuous sound section"). A typical example of a sustainable phoneme present in a sustained tone interval is a voiced tone (eg, a vowel), but also includes a consonant (eg, a frictional tone) in which the pronunciation can be continued temporally. As understood from the above description, the elapsed time E is comprehensively expressed as an elapsed time from the start point of a specific section in the target voice, and the voiced section V, the voice section and the sustained sound section have the elapsed time E It is an illustration of the specific section calculated.

(6) In the above-described embodiments, the voiced section v0 is divided into the voiced section V with the point at which the pitch P of the audio signal X changes as a boundary, but the voiced section with the point at which the volume D of the audio signal X changes It is also possible to divide v0 into a voiced section V for each note of the target music.

(7) The type of voice quality given to the audio signal X is not limited to the above-mentioned examples (breath sound, vocal fly). For example, a configuration for converting the processing section Q of the audio signal X into a hoarse voice (cloudy voice), a throat voice or a growlling voice (Growl), a processing section Q of the audio signal X is a tense voice or no tension. A configuration for converting to voice (lux) is also adopted. For the addition of hoarse voice and throat clog, for example, the techniques of JP-A-2010-191042 and JP-A-2006-145867 are suitably used. In addition, the target voice is converted into a robust voice by emphasizing the section of the audio signal X immediately after the start of pronunciation, and the target voice is converted into a flat voice by suppressing the section immediately after the start of pronunciation. It is possible.

(8) It is also possible to realize the voice processing device 100 by a server device that communicates with a terminal device such as a mobile telephone. For example, the audio processing device 100 generates the audio signal Y by executing the processing exemplified in the above-described embodiments for the audio signal X (music data Z and synthetic data S) received from the terminal device via the communication network. The voice signal Y is transmitted to the communication network with the terminal device as a destination.

100: voice processing device, 200: signal supply device, 10: arithmetic processing device, 12: storage device, 14: display device, 16: operating device, 18: sound emitting device, 22: feature Amount specifying unit, 24: section setting unit, 26: variable control unit, 28: voice processing unit.

Claims (1)

Feature amount specifying means for specifying a feature amount of the target voice;
A section setting unit configured to set a processing section according to the comparison result of the feature amount and the threshold value;
Variable control means for setting control variables for controlling voice quality for the processing section;
A voice processing apparatus comprising: voice processing means for generating a voice signal of voice in which voice quality of the processing section of the target voice is controlled according to the control variable.