JP7200483B2 - Speech processing method, speech processing device and program - Google Patents

Description

The present invention relates to techniques for processing audio signals representing speech.

2. Description of the Related Art Conventionally, various techniques have been proposed for adding speech expressions such as singing expressions to speech. For example, Patent Literature 1 discloses a technique for converting the sound represented by the audio signal into a characteristic voice such as a hoarse voice or a hoarse voice by moving each harmonic component of the audio signal in the frequency domain. ing.

Japanese Unexamined Patent Application Publication No. 2014-2338

However, the technique of Patent Literature 1 has room for further improvement from the viewpoint of generating an audibly natural sound. SUMMARY OF THE INVENTION In view of the above circumstances, the present invention aims at synthesizing acoustically natural speech.

In order to solve the above problems, a sound processing method according to a preferred aspect of the present invention sets a processing period of a first sound signal representing a singing voice to a reference voice having acoustic characteristics different from those of the singing voice. In the second sound signal represented, the representation period to be applied to the deformation of the first sound signal is extended according to the time length of the representation period, and the first sound signal after the extension of the processing period is the second sound signal It transforms according to the representation period.

In order to solve the above problems, a sound processing apparatus according to a preferred aspect of the present invention sets a processing period of a first sound signal representing a singing voice to a reference voice having acoustic characteristics different from those of the singing voice. In the second sound signal represented, the representation period to be applied to the deformation of the first sound signal is extended according to the time length of the representation period, and the first sound signal after the extension of the processing period is the second sound signal A synthesizing unit that transforms according to the expression period is provided.

1 is a block diagram illustrating the configuration of a sound processing device according to an embodiment of the present invention; FIG. 2 is a block diagram illustrating the functional configuration of a sound processing device; FIG. FIG. 4 is an explanatory diagram of a stationary period in the first sound signal; 4 is a flowchart illustrating a specific procedure of signal analysis processing; It is the time change of the fundamental frequency immediately after the pronunciation of the singing voice is started. It is the time change of the fundamental frequency immediately before the pronunciation of the singing voice ends. FIG. 10 is a flowchart illustrating a specific procedure of release processing; FIG. FIG. 11 is an explanatory diagram of release processing; FIG. 4 is an explanatory diagram of a spectral envelope outline; 4 is a flowchart illustrating a specific procedure of attack processing; FIG. 10 is an explanatory diagram of attack processing;

FIG. 1 is a block diagram illustrating the configuration of a sound processing device 100 according to a preferred embodiment of the invention. The sound processing device 100 of the present embodiment is a signal processing device that adds various sound expressions to the voice of a song sung by a user (hereinafter referred to as "singing voice"). A sound expression is an acoustic characteristic added to a singing voice (an example of the first sound). Focusing on singing of a song, the sound expression is a musical expression or facial expression related to the pronunciation of voice (that is, singing). Specifically, singing expressions such as vocal fly, growl, or hoarseness are suitable examples of sound expressions. Note that the sound expression can also be called voice quality.

The sound expression consists of a part where the volume increases immediately after the beginning of the pronunciation (hereinafter referred to as the "attack part") and a part where the volume decreases immediately before the end of the pronunciation (hereinafter referred to as the "release part"). It is particularly noticeable in In consideration of the above tendency, in the present embodiment, sound expressions are added to the attack part and the release part of the singing voice.

As illustrated in FIG. 1, the sound processing device 100 is realized by a computer system including a control device 11, a storage device 12, an operating device 13, and a sound emitting device . For example, a portable information terminal such as a mobile phone or a smart phone, or a portable or stationary information terminal such as a personal computer is preferably used as the sound processing device 100 . The operation device 13 is an input device that receives instructions from a user. For example, a plurality of manipulators operated by the user or a touch panel for detecting contact by the user is preferably used as the operating device 13 .

The control device 11 is a processing circuit such as a CPU (Central Processing Unit), for example, and executes various kinds of arithmetic processing and control processing. The control device 11 of the present embodiment generates a third sound signal Y representing a sound obtained by adding sound expression to the singing voice (hereinafter referred to as "deformed sound"). The sound emitting device 14 is, for example, a speaker or headphones, and emits a modified sound represented by the third sound signal Y generated by the control device 11 . A D/A converter for converting the third sound signal Y generated by the control device 11 from digital to analog is omitted for convenience. Although the configuration in which the sound processing device 100 includes the sound emitting device 14 is illustrated in FIG. good.

The storage device 12 is a memory composed of a known recording medium such as a magnetic recording medium or a semiconductor recording medium, and stores programs executed by the control device 11 and various data used by the control device 11 . Note that the storage device 12 may be configured by combining multiple types of recording media. Alternatively, a storage device 12 (for example, a cloud storage) may be provided separately from the sound processing device 100, and the control device 11 may perform writing and reading to and from the storage device 12 via a communication network. That is, the storage device 12 may be omitted from the sound processing device 100 .

The storage device 12 of this embodiment stores the first sound signal X1 and the second sound signal X2. The first sound signal X1 is an acoustic signal representing the singing voice of a song sung by the user of the sound processing device 100 . The second sound signal X2 is an acoustic signal representing voice (hereinafter referred to as "reference voice") sung by a singer other than the user (for example, a singer) with added sound expression. The first sound signal X1 and the second sound signal X2 have different acoustic characteristics (for example, voice quality). The sound processing device 100 of the present embodiment adds the sound representation of the reference voice (an example of the second sound) represented by the second sound signal X2 to the singing voice represented by the first sound signal X1, so that the modified sound A tritone signal Y is generated. It does not matter whether the songs are different or similar between the singing voice and the reference voice. In the above description, it is assumed that the vocalist of the singing voice and the vocalist of the reference voice are different people, but the vocalist of the singing voice and the vocalist of the reference voice may be the same person. For example, the singing voice is the voice sung by the user without adding the sound expression, and the reference voice is the voice to which the user has added the singing expression.

FIG. 2 is a block diagram illustrating the functional configuration of the control device 11. As shown in FIG. As illustrated in FIG. 2, the control device 11 executes a program stored in the storage device 12 to generate the third sound signal Y from the first sound signal X1 and the second sound signal X2. A plurality of functions (signal analysis unit 21 and synthesis processing unit 22) are realized. The functions of the control device 11 may be realized by a plurality of devices configured separately from each other, or some or all of the functions of the control device 11 may be realized by a dedicated electronic circuit.

The signal analysis unit 21 generates analysis data D1 by analyzing the first sound signal X1, and generates analysis data D2 by analyzing the second sound signal X2. The analysis data D 1 and the analysis data D 2 generated by the signal analysis unit 21 are stored in the storage device 12 .

The analysis data D1 is data representing a plurality of stationary periods Q1 in the first sound signal X1. As illustrated in FIG. 3, each stationary period Q1 indicated by the analysis data D1 is a variable-length period in which the fundamental frequency f1 and spectrum shape of the first sound signal X1 are temporally stable. The analysis data D1 designates a start point time (hereinafter referred to as "start point time") T1_S and an end point time (hereinafter referred to as "end point time") T1_E of each steady period Q1. It should be noted that the fundamental frequency f1 or spectrum shape (that is, phoneme) often changes between two consecutive notes in a piece of music. Therefore, each stationary period Q1 is likely to be a period corresponding to one note in the music.

Similarly, the analysis data D2 is data representing a plurality of stationary periods Q2 in the second sound signal X2. Each stationary period Q2 is a variable-length period in which the fundamental frequency f2 and the spectral shape of the second sound signal X2 are temporally stable. The analysis data D2 specifies the starting point time T2_S and the ending point time T2_E of each steady period Q2. As with the stationary period Q1, each stationary period Q2 is likely to be a period corresponding to one note in the piece of music.

FIG. 4 is a flow chart of processing (hereinafter referred to as "signal analysis processing") S0 for the signal analysis unit 21 to analyze the first sound signal X1. For example, the signal analysis processing S0 of FIG. 4 is started with an instruction from the user to the operation device 13 as a trigger. As illustrated in FIG. 4, the signal analysis unit 21 calculates the fundamental frequency f1 of the first sound signal X1 for each of a plurality of unit periods (frames) on the time axis (S01). A known technique is arbitrarily adopted for calculating the fundamental frequency f1. Each unit period is a sufficiently short period compared to the time length assumed for the steady period Q1.

The signal analysis unit 21 calculates a mel-cepstrum M1 representing the spectral shape of the first sound signal X1 for each unit period (S02). The mel-cepstrum M1 is represented by a plurality of coefficients representing the envelope of the frequency spectrum of the first sound signal X1. The mel-cepstrum M1 is also expressed as a feature quantity representing the phoneme of the singing voice. A known technique is arbitrarily adopted for calculating the mel-cepstrum M1. Note that MFCC (Mel-Frequency Cepstrum Coefficients) may be calculated instead of the mel-cepstrum M1 as the feature quantity representing the spectral shape of the first sound signal X1.

The signal analysis unit 21 estimates the voicing of the singing voice represented by the first sound signal X1 for each unit period (S03). That is, it is determined whether the singing voice corresponds to voiced sound or unvoiced sound. A known technique is arbitrarily adopted for estimating voicedness (voiced/unvoiced). The order of calculation of the fundamental frequency f1 (S01), calculation of the mel-cepstrum M1 (S02) and estimation of voicing (S03) is arbitrary and not limited to the order illustrated above.

The signal analysis unit 21 calculates, for each unit period, a first index δ1 that indicates the degree of temporal change in the fundamental frequency f1 (S04). For example, the difference in the fundamental frequency f1 between two consecutive unit periods is calculated as the first index .delta.1. The greater the temporal change in the fundamental frequency f1, the larger the value of the first index .delta.1.

The signal analysis unit 21 calculates a second index δ2 indicating the degree of temporal change of the mel-cepstrum M1 for each unit period (S05). For example, a numerical value obtained by synthesizing (for example, adding or averaging) differences for each coefficient of the mel-cepstrum M1 between two consecutive unit periods is suitable as the second index δ2. The second index δ2 becomes a larger numerical value as the temporal change in the spectral shape of the singing voice becomes more conspicuous. For example, the second index .delta.2 becomes a large numerical value near the time when the phoneme of the singing voice changes.

The signal analysis unit 21 calculates a fluctuation index Δ corresponding to the first index δ1 and the second index δ2 for each unit period (S06). For example, the weighted sum of the first index δ1 and the second index δ2 is calculated for each unit period as the fluctuation index Δ. A weighted value of each of the first index δ1 and the second index δ2 is set to a predetermined fixed value or a variable value according to an instruction from the user to the operating device 13 . As can be understood from the above description, there is a tendency that the greater the temporal variation of the fundamental frequency f1 or the mel-cepstrum M1 (that is, the spectral shape) of the first sound signal X1, the larger the value of the variation index Δ.

The signal analysis unit 21 identifies a plurality of stationary periods Q1 in the first sound signal X1 (S07). The signal analysis unit 21 of this embodiment specifies the steady period Q1 according to the result of estimating the voicedness of the singing voice (S03) and the fluctuation index Δ. Specifically, the signal analysis unit 21 defines a set of a series of unit periods in which the singing voice is estimated to be voiced and the fluctuation index Δ is below a predetermined threshold as the steady period Q1. A unit period in which the singing voice is estimated to be unvoiced or a unit period in which the fluctuation index Δ exceeds the threshold is excluded from the stationary period Q1. After each steady period Q1 of the first sound signal X1 is defined by the above procedure, the signal analysis unit 21 stores the analysis data D1 specifying the start point time T1_S and the end point time T1_E of each steady period Q1 in the storage device 12. (S08).

The signal analysis unit 21 generates analysis data D2 by executing the above-described signal analysis processing S0 also for the second sound signal X2 representing the reference speech. Specifically, for each unit period of the second sound signal X2, the signal analysis unit 21 calculates the fundamental frequency f2 (S01), calculates the mel-cepstrum M2 (S02), and estimates the voicing (voiced/unvoiced) ( S03) is executed. The signal analysis unit 21 calculates a fluctuation index Δ according to a first index δ1 indicating the degree of temporal change in the fundamental frequency f2 and a second index δ2 indicating the degree of temporal change in the mel-cepstrum M2. (S04-S06). Then, the signal analysis unit 21 specifies each stationary period Q2 of the second sound signal X2 according to the estimation result of the voicedness of the reference speech (S03) and the variation index Δ (S07). The signal analysis unit 21 stores the analysis data D2 specifying the start point time T2_S and the end point time T2_E of each steady period Q2 in the storage device 12 (S08). Note that the analysis data D1 and the analysis data D2 may be edited according to instructions from the user to the operation device 13. FIG.

The synthesis processing unit 22 of FIG. 2 transforms the analysis data D1 of the first sound signal X1 using the analysis data D2 of the second sound signal X2. The synthesis processing section 22 of this embodiment includes an attack processing section 31 , a release processing section 32 and a speech synthesis section 33 . The attack processing unit 31 executes attack processing S1 for adding the sound representation of the attack portion in the second sound signal X2 to the first sound signal X1. The release processing unit 32 performs release processing S2 for adding the sound representation of the release portion in the second sound signal X2 to the first sound signal X1. The voice synthesizing unit 33 synthesizes the third sound signal Y of the modified sound from the analysis data processed by the attack processing unit 31 and the release processing unit 32 .

FIG. 5 shows the change over time of the fundamental frequency f1 immediately after the start of vocalization of the singing voice. As illustrated in FIG. 5, a voiced period Va exists immediately before the stationary period Q1. The voiced period Va is the period of voiced speech that precedes the stationary period Q1. The voiced period Va is a period in which the acoustic characteristics of the singing voice (for example, the fundamental frequency f1 or the spectral shape) fluctuate unstably immediately before the steady period Q1. For example, focusing on the steady period Q1 immediately after the vocalization of the singing voice starts, the attack part from the time τ1_A at which the vocalization of the singing voice starts to the start time T1_S of the steady period Q1 corresponds to the voiced period Va. In the above description, the focus is on the singing voice, but the reference voice also has a voiced period Va immediately before the steady period Q2. In the attack processing S1, the synthesizing unit 22 (specifically, the attack processing unit 31) performs the attack processing of the second sound signal X2 for the voiced period Va and the stationary period Q1 immediately after the first sound signal X1. Add sound expression.

FIG. 6 shows the change over time of the fundamental frequency f1 just before the end of vocalization of the singing voice. As illustrated in FIG. 6, a voiced period Vr exists immediately after the steady period Q1. A voiced period Vr is a period of voiced speech that follows the stationary period Q1. The voiced period Vr is a period in which the acoustic characteristics of the singing voice (for example, the fundamental frequency f2 or the spectral shape) unstably fluctuate immediately after the steady period Q1. For example, focusing on the steady period Q1 immediately before the end of the vocalization of the singing voice, the release part from the end point time T1_E of the steady period Q1 to the time τ1_R at which the singing voice is muted corresponds to the voiced period Vr. Although the above description focused on the singing voice, the reference voice also has a voice period Vr immediately after the stationary period Q2. In the release processing S2, the synthesizing unit 22 (specifically, the release processing unit 32) performs the release part of the second sound signal X2 for the voiced period Vr and the immediately preceding steady period Q1 of the first sound signal X1. Add sound expression.

<Release process S2>
FIG. 7 is a flowchart illustrating specific contents of the release processing S2 executed by the release processing unit 32. As shown in FIG. The release process S2 of FIG. 7 is executed every steady period Q1 of the first sound signal X1.

When the release processing S2 is started, the release processing unit 32 determines whether or not to add the sound expression of the release portion of the second sound signal X2 to the stationary period Q1 to be processed in the first sound signal X1 (S21). . Specifically, the release processing unit 32 determines not to add the sound expression of the release part for the stationary period Q1 corresponding to any one of conditions Cr1 to Cr3 exemplified below. However, the conditions for determining whether or not to add sound representation to the stationary period Q1 of the first sound signal X1 are not limited to the following examples.
[Condition Cr1] The time length of the steady period Q1 is below a predetermined value.
[Condition Cr2] The duration of the silent period immediately after the steady period Q1 is below a predetermined value.
[Condition Cr3] The time length of the voiced period Vr following the steady period Q1 exceeds a predetermined value.

It is difficult to add sound expression with natural voice quality to the stationary period Q1 whose time length is sufficiently short. Therefore, when the length of time of the steady period Q1 is less than a predetermined value (condition Cr1), the release processing unit 32 excludes the steady period Q1 from addition of sound representation. Also, if a sufficiently short silent period exists immediately after the stationary period Q1, the silent period may be a period of unvoiced consonants in the middle of the singing voice. If sound representation is added to the period of unvoiced consonants, there is a tendency that an auditory sense of incongruity is perceived. Considering the above tendency, when the length of time of the silent period immediately after the steady period Q1 is less than a predetermined value (condition Cr2), the release processing unit 32 excludes the steady period Q1 from addition of sound representation. Also, if the time length of the voiced period Vr immediately after the steady period Q1 is sufficiently long, there is a high possibility that sufficient sound expression has already been added to the singing voice. Therefore, when the time length of the voiced period Vr following the stationary period Q1 is sufficiently long (condition Cr3), the release processing unit 32 excludes the stationary period Q1 from addition of sound representation. When it is determined that no sound expression is added to the stationary period Q1 of the first sound signal X1 (S21: NO), the release processing unit 32 performs the release processing S2 without executing the processing (S22-S26) described in detail below. exit.

When it is determined that the sound representation of the release portion of the second sound signal X2 is added to the steady period Q1 of the first sound signal X1 (S21: YES), the release processing section 32 adds a plurality of steady periods Q2 of the second sound signal X2. Among them, the stationary period Q2 corresponding to the sound expression to be added to the stationary period Q1 of the first sound signal X1 is selected (S22). Specifically, the release processing unit 32 selects the steady period Q2 in which the situation in the music is similar to the steady period Q1 to be processed. For example, the situation (context) to be considered for one steady period (hereinafter referred to as the "steady period of interest") includes the time length of the steady period of interest, the time length of the steady period immediately after the steady period of interest, and the steady period of interest. The pitch difference from the stationary period immediately after, the pitch of the stationary period of interest, and the time length of the silent period immediately before the stationary period of interest are exemplified. The release processing unit 32 selects the steady period Q2 that minimizes the difference from the steady period Q1 for the situations illustrated above.

The release processing unit 32 executes processing (S23-S26) for adding the sound expression corresponding to the stationary period Q2 selected in the above procedure to the first sound signal X1 (analysis data D1). FIG. 8 is an explanatory diagram of the processing in which the release processing unit 32 adds the sound representation of the release portion to the first sound signal X1.

FIG. 8 also shows waveforms on the time axis and temporal changes in the fundamental frequency for each of the first sound signal X1, the second sound signal X2, and the third sound signal Y after deformation. In FIG. 8, the start time T1_S and the end time T1_E of the steady period Q1 of the singing voice, the end time τ1_R of the voiced period Vr immediately after the steady period Q1, and the voiced period Va corresponding to the note immediately after the steady period Q1 , the start time T2_S and end time T2_E of the steady period Q2 of the reference speech, and the end time τ2_R of the voiced period Vr immediately after the steady period Q2 are known information.

The release processing unit 32 adjusts the positional relationship on the time axis between the steady period Q1 to be processed and the steady period Q2 selected in step S22 (S23). Specifically, the release processing unit 32 adjusts the position of the steady period Q2 on the time axis to a position based on the end points (T1_S, T1_E) of the steady period Q1. As exemplified in FIG. 8, the release processing unit 32 of the present embodiment generates the second sound signal X2 (stationary The period Q2) is arranged on the time axis of the first sound signal X1.

<Extension of processing period Z1_R (S24)>
The release processing unit 32 expands or contracts a period Z1_R during which the sound expression of the second sound signal X2 is added to the first sound signal X1 (hereinafter referred to as "processing period") on the time axis (S24). As exemplified in FIG. 8, the processing period Z1_R is the period from the time Tm_R at which addition of the sound expression is started (hereinafter referred to as "synthesis start time") to the end point time τ1_R of the voiced period Vr immediately after the steady period Q1. be. The synthesis start time Tm_R is the later time between the start time T1_S of the steady period Q1 of the singing voice and the start time T2_S of the steady period Q2 of the reference voice. As illustrated in FIG. 8, when the start time T2_S of the steady period Q2 is located after the start time T1_S of the steady period Q1, the start time T2_S of the steady period Q2 is set as the synthesis start time Tm_R. However, synthesis start time Tm_R is not limited to start point time T2_S.

As illustrated in FIG. 8, the release processing unit 32 of the present embodiment extends the processing period Z1_R of the first sound signal X1 according to the time length of the expression period Z2_R of the second sound signal X2. The expression period Z2_R is a period representing the sound expression of the release portion of the second sound signal X2, and is used to add the sound expression to the first sound signal X1. As illustrated in FIG. 8, the representation period Z2_R is a period from the synthesis start time Tm_R to the end point time τ2_R of the voiced period Vr immediately after the steady period Q2.

A reference voice sung by a skilled singer such as a singer is added with sufficient sound expression over a suitable length of time, whereas a singing voice sung by a user who is unfamiliar with singing lacks sound expression in terms of time. tend to Under the above tendency, as illustrated in FIG. 8, the expression period Z2_R of the reference voice is longer than the processing period Z1_R of the singing voice. Therefore, the release processing unit 32 of the present embodiment extends the processing period Z1_R of the first sound signal X1 to the time length of the expression period Z2_R of the second sound signal X2.

The expansion of the processing period Z1_R is realized by a process (mapping) that associates an arbitrary time t1 of the first sound signal X1 (singing voice) with an arbitrary time t of the third sound signal Y after deformation (deformed sound). be done. FIG. 8 shows the correspondence relationship between time t1 (vertical axis) of the singing voice and time t (horizontal axis) of the modified sound.

The time t1 in the correspondence relationship of FIG. 8 is the time of the first sound signal X1 corresponding to the time t of the modified sound. A reference line L indicated by a chain line in FIG. 8 means a state (t1=t) in which the first sound signal X1 is not expanded or contracted. Also, a section in which the gradient of the time t1 of the singing voice with respect to the time t of the modified sound is smaller than that of the reference line L means a section in which the first sound signal X1 is expanded. A section in which the gradient of time t1 with respect to time t is greater than that of the reference line L means a section in which the singing voice is contracted.

The correspondence between the time t1 and the time t is represented by the nonlinear functions of formulas (1a) to (1c) exemplified below.

The time T_R is a predetermined time located between the synthesis start time Tm_R and the end point time τ1_R of the processing period Z1_R, as illustrated in FIG. For example, the later time between the middle point ((T1_S+T1_E)/2) between the start time T1_S and the end time T1_E of the steady period Q1 and the synthesis start time Tm_R is set as the time T_R. As understood from the formula (1a), the period before the time T_R in the processing period Z1_R is not expanded or contracted. That is, extension of the processing period Z1_R is started from the time T_R.

As understood from the formula (1b), in the period after the time T_R in the processing period Z1_R, the degree of expansion is large at a position close to the time T_R, and the degree of expansion decreases as the end point time τ1_R approaches. Stretched on the time axis. The function η(t) of Expression (1b) is a nonlinear function for extending the processing period Z1_R toward the front on the time axis and reducing the degree of extension of the processing period Z1_R toward the rear on the time axis. Specifically, for example, a quadratic function of time t (η(t)=t ² ) is preferably used as the function η(t). As described above, in the present embodiment, the processing period Z1_R is extended on the time axis so that the closer the position is to the end point time τ1_R of the processing period Z1_R, the smaller the degree of extension. Therefore, it is possible to sufficiently maintain the acoustic characteristics in the vicinity of the ending point time τ1_R of the singing voice even in the deformed sound. At a position close to the time T_R, there is a tendency that an auditory sense of incongruity caused by the expansion is less likely to be perceived than at a position near the end point time τ1_R. Therefore, even if the degree of extension is increased at a position close to time T_R as in the above example, the audible naturalness of the deformed sound hardly deteriorates. Note that the period from the end point time τ2_R of the expression period Z2_R to the start point time τ1_A of the next voiced period Vr in the first sound signal X1 is shortened on the time axis as can be understood from Equation (1c). Note that since there is no sound during the period from the end point time τ2_R to the start point time τ1_A, the first sound signal X1 may be deleted by partial deletion.

As illustrated above, the singing voice processing period Z1_R is extended to the time length of the reference voice representation period Z2_R. On the other hand, the representation period Z2_R of the reference speech is not expanded or contracted on the time axis. That is, the time t2 of the arranged second sound signal X2 corresponding to the time t of the modified sound coincides with the time t (t2=t). As illustrated above, in the present embodiment, the singing voice processing period Z1_R is expanded according to the time length of the expression period Z2_R, so expansion of the second sound signal X2 is unnecessary. Therefore, it is possible to accurately add the sound representation of the release portion represented by the second sound signal X2 to the first sound signal X1.

When the processing period Z1_R is extended by the procedure illustrated above, the release processing unit 32 transforms the processing period Z1_R after the extension of the first sound signal X1 according to the expression period Z2_R of the second sound signal X2 (S25-S26 ). Specifically, fundamental frequency synthesis (S25) and spectral envelope outline synthesis (S26) are performed between the processing period Z1_R after decompressing the singing voice and the representation period Z2_R of the reference voice.

<Fundamental frequency synthesis (S25)>
The release processing unit 32 calculates the fundamental frequency F(t) of the third sound signal Y at each time t using the following formula (2).

The smoothed fundamental frequency F1(t1) in Equation (2) is a frequency obtained by smoothing the time series of the fundamental frequency f1(t1) of the first sound signal X1 on the time axis. Similarly, the smoothed fundamental frequency F2(t2) in Equation (2) is a frequency obtained by smoothing the time series of the fundamental frequency f2(t2) of the second sound signal X2 on the time axis. The coefficient λ1 and the coefficient λ2 in equation (2) are set to non-negative values of 1 or less (0≤λ1≤1, 0≤λ2≤1).

As can be seen from the formula (2), the second term of the formula (2) expresses the difference between the fundamental frequency f1(t1) of the singing voice and the smoothed fundamental frequency F1(t1) according to the coefficient λ1, This is a process for reducing the fundamental frequency f1(t1) of the first sound signal X1. Also, the third term of the equation (2) expresses the difference between the fundamental frequency f2(t2) of the reference sound and the smoothed fundamental frequency F2(t2) by the degree according to the coefficient λ2, the fundamental frequency of the first sound signal X1 This is a process to add to f1(t1). As can be understood from the above description, the release processing unit 32 calculates the difference between the fundamental frequency f1(t1) of the singing voice and the smoothed fundamental frequency F1(t1) as the fundamental frequency f2(t2) of the reference voice and the smoothed fundamental frequency It functions as an element that replaces the difference with F2(t2). That is, the time change of the fundamental frequency f1(t1) within the processing period Z1_R after expansion of the first sound signal X1 approaches the time change of the fundamental frequency f2(t2) within the representation period Z2_R of the second sound signal X2.

<Synthesis of Spectrum Envelope (S26)>
The release processing unit 32 synthesizes a spectral envelope outline between the processing period Z1_R after decompression of the singing voice and the expression period Z2_R of the reference voice. As illustrated in FIG. 9, the spectral envelope outline G1 of the first sound signal X1 means the intensity distribution obtained by further smoothing the spectrum envelope g2, which is the outline of the frequency spectrum g1 of the first sound signal X1, in the frequency domain. do. Specifically, the spectral envelope outline G1 is the intensity distribution obtained by smoothing the spectral envelope g2 to such an extent that phonology (differences depending on phonemes) and individuality (differences depending on the speaker) cannot be perceived. For example, the spectrum envelope outline G1 is represented by a predetermined number of coefficients positioned on the lower order side among the plurality of coefficients of the mel-cepstrum representing the spectrum envelope g2. In the above description, attention is focused on the spectral envelope outline G1 of the first sound signal X1, but the same applies to the spectrum envelope outline G2 of the second sound signal X2.

The release processing unit 32 calculates a spectral envelope outline (hereinafter referred to as a "synthetic spectral envelope outline") G(t) of the third sound signal Y at each time t using the following formula (3).

The symbol G1_ref in Equation (3) is the reference spectral envelope outline. Of the plurality of spectral envelope outlines G1 of the first sound signal X1, one spectral envelope outline G1 at a specific point in time is used as a reference spectral envelope outline G1_ref (an example of the first reference spectrum envelope outline). be. Specifically, the reference spectral envelope outline G1_ref is the spectral envelope outline G1(Tm_R) of the first sound signal X1 at the synthesis start time Tm_R (example of the first time point). That is, the time point at which the reference spectral envelope outline G1_ref is extracted is positioned later than the starting point time T1_S of the steady period Q1 and the starting point time T2_S of the steady period Q2. Note that the time at which the reference spectral envelope outline G1_ref is extracted is not limited to the synthesis start time Tm_R. For example, the spectral envelope outline G1 at any time point within the stationary period Q1 is used as the reference spectral envelope outline G1_ref.

Similarly, the reference spectral envelope outline G2_ref in Equation (3) is one spectral envelope outline G2 at a specific point in time among the plurality of spectral envelope outlines G2 of the second sound signal X2. Specifically, the reference spectral envelope outline G2_ref is the spectral envelope outline G2(Tm_R) of the second sound signal X2 at the synthesis start time Tm_R (example of the second time point). In other words, the point of time at which the reference spectral envelope outline G2_ref is extracted is positioned later than the starting point time T1_S of the steady period Q1 and the starting point time T2_S of the steady period Q2. Note that the point at which the reference spectral envelope outline G2_ref is extracted is not limited to the synthesis start time Tm_R. For example, the spectral envelope outline G2 at an arbitrary time point within the stationary period Q1 is used as the reference spectral envelope outline G2_ref.

The coefficient μ1 and the coefficient μ2 in equation (3) are set to non-negative values of 1 or less (0≤μ1≤1, 0≤μ2≤1). The second term of Equation (3) expresses the difference between the spectral envelope outline G1(t1) of the singing voice and the reference spectral envelope outline G1_ref by the degree corresponding to the coefficient μ1 (example of the first coefficient), the first This is a process of reducing from the spectral envelope outline G1(t1) of the sound signal X1. In addition, the third term of Equation (3) expresses the difference between the spectral envelope outline G2(t2) of the reference speech and the reference spectral envelope outline G2_ref in accordance with the coefficient μ2 (an example of the second coefficient), This is a process of reducing from the spectral envelope outline G2(b) of the second sound signal X2. As can be understood from the above description, the release processing unit 32 converts the difference (example of the first difference) between the spectral envelope outline G1(t1) of the singing voice and the reference spectral envelope outline G1_ref into the spectral envelope of the reference voice. It functions as an element that replaces the difference (an example of the second difference) between the outline G2(t2) and the reference spectrum envelope outline G2_ref.

<Attack processing S1>
FIG. 10 is a flowchart illustrating specific contents of the attack processing S1 executed by the attack processing section 31. As shown in FIG. The attack process S1 of FIG. 10 is executed for each stationary period Q1 of the first sound signal X1. The specific procedure of the attack processing S1 is the same as that of the release processing S2.

When the attack processing S1 is started, the attack processing section 31 determines whether or not to add the sound representation of the attack portion of the second sound signal X2 to the stationary period Q1 to be processed in the first sound signal X1 (S11). . Specifically, the attack processing unit 31 determines not to add the sound representation of the attack portion for the stationary period Q1 corresponding to any one of conditions Ca1 to Ca5 illustrated below. However, the conditions for determining whether or not to add sound representation to the stationary period Q1 of the first sound signal X1 are not limited to the following examples.
[Condition Ca1] The time length of the steady period Q1 is below a predetermined value.
[Condition Ca2] The fluctuation width of the smoothed fundamental frequency f1 within the steady period Q1 exceeds a predetermined value.
[Condition Ca3] The fluctuation range of the smoothed fundamental frequency f1 within a predetermined length of period including the start point of the steady period Q1 exceeds a predetermined value.
[Condition Ca4] The time length of the voiced period Va immediately preceding the steady period Q1 exceeds a predetermined value.
[Condition Ca5] The fluctuation width of the fundamental frequency f1 in the voiced period Va immediately before the steady period Q1 exceeds a predetermined value.

Condition Ca1, like condition Cr1, is a condition that takes into consideration the fact that it is difficult to add sound expression with natural voice quality during a sufficiently short stationary period Q1. Further, when the fundamental frequency f1 fluctuates greatly within the stationary period Q1, there is a high possibility that sufficient sound expression is added to the singing voice. Therefore, the stationary period Q1 in which the fluctuation range of the smoothed fundamental frequency f1 exceeds a predetermined value is excluded from the objects to which sound expression is added (condition Ca2). The condition Ca3 has the same content as the condition Ca2, but is a condition focused on a period particularly close to the attack part in the steady period Q1. Further, when the time length of the voiced period Va immediately before the stationary period Q1 is sufficiently long, or when the fundamental frequency f1 fluctuates greatly within the voiced period Va, sufficient sound expression is already added to the singing voice. Probability is high. Therefore, the steady period Q1 (condition Ca4) in which the time length of the immediately preceding voiced period Va exceeds a predetermined value and the steady period Q1 (condition Ca5) in which the fluctuation range of the fundamental frequency f1 within the voiced period Va exceeds a predetermined value are defined. , are excluded from addition of sound expressions. If it is determined that the sound expression is not added to the stationary period Q1 (S11: YES), the attack processing section 31 ends the attack processing S1 without executing the processing (S12-S16) described in detail below.

When it is determined that the sound representation of the attack portion of the second sound signal X2 is added to the steady period Q1 of the first sound signal X1 (S11: YES), the attack processing section 31 adds a plurality of steady periods Q2 of the second sound signal X2. Among them, the stationary period Q2 corresponding to the sound expression to be added to the stationary period Q1 is selected (S12). The method by which the attack processing unit 31 selects the steady period Q2 is the same as the method by which the release processing unit 32 selects the steady period Q2.

The attack processing unit 31 executes processing (S13-S16) for adding the sound expression corresponding to the stationary period Q2 selected in the above procedure to the first sound signal X1. FIG. 11 is an explanatory diagram of processing in which the attack processing unit 31 adds the sound expression of the attack portion to the first sound signal X1.

The attack processing unit 31 adjusts the positional relationship on the time axis between the steady period Q1 to be processed and the steady period Q2 selected in step S12 (S13). Specifically, as illustrated in FIG. 11, the attack processing unit 31 generates the second sound signal X2 ( The stationary period Q2) is arranged on the time axis of the first sound signal X1.

<Extension of processing period Z1_A>
The attack processing unit 31 extends the processing period Z1_A in which the sound representation of the second sound signal X2 is added to the first sound signal X1 on the time axis (S14). The processing period Z1_A is a period from the starting point time τ1_A of the voiced period Va immediately before the steady period Q1 to the time Tm_A at which the addition of the sound expression is finished (hereinafter referred to as "synthesis end time"). The synthesis end time Tm_A is, for example, the start time T1_S of the steady period Q1 (the start time T2_S of the steady period Q2). That is, in the attack processing S1, the voiced period Va preceding the steady period Q1 is extended as the processing period Z1_A. As described above, the stationary period Q1 is a period corresponding to musical notes of music. According to the configuration in which the voiced period Va is extended but the steady period Q1 is not extended, the change in the start point time T1_S of the steady period Q1 is suppressed. That is, it is possible to reduce the possibility that the head of the note in the singing voice moves back and forth.

As illustrated in FIG. 11, the attack processing unit 31 of this embodiment extends the processing period Z1_A of the first sound signal X1 according to the time length of the expression period Z2_A of the second sound signal X2. The representation period Z2_A is a period representing the sound representation of the attack portion of the second sound signal X2, and is used to add the sound representation to the first sound signal X1. As illustrated in FIG. 11, the expression period Z2_A is the voiced period Va immediately preceding the stationary period Q2.

Specifically, the attack processing unit 31 extends the processing period Z1_A of the first sound signal X1 to the time length of the expression period Z2_A of the second sound signal X2. FIG. 11 shows the correspondence relationship between time t1 (vertical axis) of the singing voice and time t (horizontal axis) of the modified sound.

As exemplified in FIG. 11, in the present embodiment, the processing period Z1_A is extended on the time axis so that the closer the position is to the starting point time τ1_A of the processing period Z1_A, the smaller the degree of extension. Therefore, it is possible to sufficiently maintain the acoustic characteristics in the vicinity of the starting point time τ1_A of the singing voice even in the deformed sound. On the other hand, the representation period Z2_A of the reference speech is not expanded or contracted on the time axis. Therefore, it is possible to accurately add the sound representation of the attack portion represented by the second sound signal X2 to the first sound signal X1.

When the processing period Z1_A is extended by the procedure illustrated above, the attack processing unit 31 transforms the processing period Z1_A after extension of the first sound signal X1 according to the expression period Z2_A of the second sound signal X2 (S15-S16 ). Specifically, fundamental frequency synthesis (S25) and spectrum envelope synthesis (S26) are performed between the processing period Z1_A after decompression of the singing voice and the representation period Z2_A of the reference voice.

Specifically, the attack processing unit 31 calculates the following from the fundamental frequency f1(t1) of the first sound signal X1 and the fundamental frequency f2(t2) of the second sound signal X2 by the same calculation as the formula (2). The fundamental frequency F(t) of the third sound signal Y is calculated. That is, the attack processing unit 31 reduces the difference between the fundamental frequency f1(t1) and the smoothed fundamental frequency F1(t1) from the fundamental frequency f1(t1) of the first sound signal X1 at a degree corresponding to the coefficient λ1. , the difference between the fundamental frequency f2(t2) and the smoothed fundamental frequency F2(t2) is added to the fundamental frequency f1(t1) of the first sound signal X1 at a degree corresponding to the coefficient λ2 to obtain the third sound signal Calculate the fundamental frequency of Y, F(t). Therefore, the time change of the fundamental frequency f1(t1) within the processing period Z1_A after expansion of the first sound signal X1 approaches the time change of the fundamental frequency f2(t2) within the representation period Z2_A of the second sound signal X2.

Also, the attack processing unit 31 synthesizes a spectral envelope outline between the processing period Z1_A after decompression of the singing voice and the expression period Z2_A of the reference voice. Specifically, the attack processing unit 31 calculates the spectral envelope outline G1(t1) of the first sound signal X1 and the spectral envelope outline G2(t1) of the second sound signal X2 by a calculation similar to the above-described formula (3). t2) and the synthesized spectral envelope outline G(t) of the third sound signal Y is calculated. The reference spectral envelope outline G1_ref applied to Equation (3) in the attack process S1 is the spectral envelope outline G1(Tm_A) at the synthesis end time Tm_A (example of the first point in time) of the first sound signal X1. That is, the time point at which the reference spectral envelope outline G1_ref is extracted is located at the start point time T1_S of the stationary period Q1.

Similarly, the reference spectral envelope outline G2_ref applied to Equation (3) in the attack processing S1 is the spectral envelope outline G2(Tm_A) at the synthesis end time Tm_A (exemplification of the second point in time) of the second sound signal X2. is. That is, the time point at which the reference spectral envelope outline G2_ref is extracted is located at the start point time T1_S of the stationary period Q1.

As can be understood from the above description, each of the attack processing unit 31 and the release processing unit 32 of the present embodiment performs The first sound signal X1 (analysis data D1) is transformed using the second sound signal X2 (analysis data D2). By the attack processing S1 and the release processing S2 illustrated above, the time series of the fundamental frequency F(t) of the third sound signal Y representing the modified sound and the time series of the synthetic spectral envelope outline G(t) are generated. . The speech synthesizer 33 of FIG. 2 generates the third sound signal Y from the time series of the fundamental frequency F(t) of the third sound signal Y and the time series of the synthesized spectrum envelope outline G(t).

The speech synthesizing unit 33 in FIG. 2 synthesizes the third sound signal Y of the deformed sound using the results of the attack processing S1 and the release processing S2 (that is, the analysis data after deformation). Specifically, the speech synthesizer 33 adjusts each frequency spectrum g1 calculated from the first sound signal X1 so as to conform to the synthesized spectrum envelope outline G(t), and Adjust the frequency f1 to the fundamental frequency F(t). The adjustment of the frequency spectrum g1 and the fundamental frequency f1 is performed, for example, in the frequency domain. The speech synthesizing unit 33 synthesizes the third sound signal Y by transforming the adjusted frequency spectrum illustrated above into the time domain.

As described above, in this embodiment, the difference (G1(t1)-G1_ref) between the spectral envelope outline G1(t1) of the first sound signal X1 and the reference spectral envelope outline G1_ref, and the second sound signal X2 and the difference (G2(t2)-G2_ref) between the spectral envelope outline G2(t2) and the reference spectral envelope outline G2_ref are combined into the spectral envelope outline G1(t1) of the first sound signal X1. Therefore, it is perceptually natural that the acoustic characteristics are continuous at the boundaries between the periods (processing periods Z1_A and Z1_R) in which the first sound signal X1 is modified using the second sound signal X2 and the periods before and after the relevant periods. It can generate various deformation sounds.

Further, in the present embodiment, the steady period Q1 in which the fundamental frequency f1 and the spectral shape of the first sound signal X1 are temporally stable is specified, and the end point (start point time T1_S or end point time T1_E) of the steady period Q1 is specified. The first sound signal X1 is transformed using the second sound signal X2 placed as a reference. Therefore, an appropriate period of the first sound signal X1 is deformed according to the second sound signal X2, and an acoustically natural deformed sound can be generated.

In the present embodiment, the processing period (Z1_A, Z1_R) of the first sound signal X1 is expanded according to the time length of the expression period (Z2_A, Z2_R) of the second sound signal X2. is unnecessary. Therefore, the acoustic characteristics (for example, sound expression) of the reference voice are accurately added to the first sound signal X1, and an acoustically natural deformed sound can be generated.

<Modification>
Specific modified aspects added to the above-exemplified aspects will be exemplified below. Two or more aspects arbitrarily selected from the following examples may be combined as appropriate within a mutually consistent range.

(1) In the above embodiment, the steady period Q1 of the first sound signal X1 is specified using the fluctuation index Δ calculated from the first index δ1 and the second index δ2. The method of identifying the steady period Q1 according to the index .delta.2 is not limited to the above examples. For example, the signal analysis unit 21 identifies the first provisional period corresponding to the first index δ1 and the second provisional period corresponding to the second index δ2. The first provisional period is, for example, a period of voiced speech in which the first index δ1 is below the threshold. That is, the period during which the fundamental frequency f1 is temporally stable is specified as the first provisional period. The second provisional period is, for example, a period of voiced speech in which the second index δ2 is below the threshold. That is, a period in which the spectral shape is temporally stable is specified as the second provisional period. The signal analysis unit 21 identifies a period in which the first provisional period and the second provisional period overlap each other as the steady period Q1. That is, a period in which both the fundamental frequency f1 and the spectral shape of the first sound signal X1 are temporally stable is specified as the stationary period Q1. As can be understood from the above explanation, the calculation of the fluctuation index Δ may be omitted in identifying the steady period Q1. In the above description, attention is paid to specifying the steady period Q1, but the same applies to specifying the steady period Q2 in the second sound signal X2.

(2) In the above embodiment, the period in which both the fundamental frequency f1 and the spectral shape of the first sound signal X1 are temporally stable is specified as the stationary period Q1. A period in which one of the spectral shapes is temporally stable may be specified as the stationary period Q1. Similarly, a period during which one of the fundamental frequency f2 and the spectral shape of the second sound signal X2 is temporally stable may be specified as the stationary period Q2.

(3) In the above embodiment, the spectral envelope outline G1 at the synthesis start time Tm_R or the synthesis end time Tm_A of the first sound signal X1 is used as the reference spectral envelope outline G1_ref, but the reference spectrum envelope outline G1_ref is extracted. The point in time (first point in time) is not limited to the above example. For example, the spectral envelope outline G1 at the end point (start point time T1_S or end point time T1_E) of the stationary period Q1 may be used as the reference spectral envelope outline G1_ref. However, the first point in time at which the reference spectral envelope outline G1_ref is extracted is preferably a point in the stationary period Q1 in which the spectral shape of the first sound signal X1 is stable.

The same is true for the reference spectral envelope outline G2_ref. That is, in the above embodiment, the spectral envelope outline G2 at the synthesis start time Tm_R or the synthesis end time Tm_A of the second sound signal X2 is used as the reference spectral envelope outline G2_ref, but the reference spectral envelope outline G2_ref is extracted. The point in time (second point in time) is not limited to the above example. For example, the spectral envelope outline G2 at the end point (start point time T2_S or end point time T2_E) of the stationary period Q2 may be used as the reference spectral envelope outline G2_ref. However, it is desirable that the second point in time at which the reference spectral envelope outline G2_ref is extracted be a point in the stationary period Q2 in which the spectral shape of the second sound signal X2 is stable.

Further, the first time point at which the reference spectral envelope outline G1_ref is extracted from the first sound signal X1 and the second time point at which the reference spectrum envelope outline G2_ref is extracted from the second sound signal X2 are separated from each other on the time axis. different points in time.

(4) In the above embodiment, the first sound signal X1 representing the singing voice sung by the user of the sound processing device 100 is processed, but the voice represented by the first sound signal X1 is not limited to the singing voice of the user. . For example, the first sound signal X1 synthesized by a known segment connection type or statistical model type speech synthesis technique may be processed. Alternatively, the first sound signal X1 read from a recording medium such as an optical disk may be processed. Similarly, the second sound signal X2 is obtained by any method.

Moreover, the sounds represented by the first sound signal X1 and the second sound signal X2 are not limited to sounds in a narrow sense (that is, speech sounds uttered by humans). For example, the present invention can be applied to adding various sound expressions (for example, musical performance expressions) to the first sound signal X1 representing the performance sound of a musical instrument. For example, a performance expression such as vibrato is added to a first sound signal X1 representing a monotonous performance sound to which no performance expression is added, using the second sound signal X2.

<Appendix>
For example, the following configuration can be grasped from the form illustrated above.

In the sound processing method according to a preferred aspect (first aspect) of the present invention, the processing period of the first sound signal representing the singing voice is set to the second sound signal representing the reference voice having acoustic characteristics different from those of the singing voice. A sound signal is extended according to a time length of an expression period to be applied to transform the first sound signal, and the first sound signal after the extension of the processing period is applied to the expression period of the second sound signal. Transform and synthesize accordingly. In the above aspect, since the processing period of the first sound signal is extended according to the time length of the representation period of the second sound signal, it is unnecessary to extend or shorten the second sound signal representing the reference sound. Therefore, it is possible to accurately add the acoustic characteristics of the reference speech to the first sound signal.

In a preferred example of the first aspect (second aspect), the processing period is a period including the release portion of the singing voice, and in extending the processing period, the closer the position to the end point of the processing period, the greater the degree of extension. The processing period is extended so that . According to the above aspect, since the degree of expansion decreases closer to the end point of the processing period, it is possible to add the acoustic characteristics of the reference voice while maintaining the acoustic characteristics near the end point of the singing voice.

In a preferred example of the first aspect (third aspect), the processing period is a period including an attack portion of the singing voice, and in extending the processing period, the closer the position to the starting point of the processing period, the greater the degree of extension. The processing period is extended so that . According to the above aspect, since the degree of expansion is smaller at a position closer to the starting point of the processing period, it is possible to add the acoustic characteristics of the reference voice while maintaining the acoustic characteristics near the starting point of the singing voice.

A sound processing apparatus according to a preferred aspect (fourth aspect) of the present invention is configured such that the processing period of the first sound signal representing the singing voice is set to the second sound signal representing the reference voice having acoustic characteristics different from those of the singing voice. A sound signal is extended according to a time length of an expression period to be applied to transform the first sound signal, and the first sound signal after the extension of the processing period is applied to the expression period of the second sound signal. A synthesizing unit that deforms accordingly is provided.

In a preferred example of the fourth aspect (fifth aspect), the processing period is a period including a release portion of the singing voice, and the synthesis processing unit expands the position closer to the end point of the processing period. The processing period is extended so that

In a preferred example of the fourth aspect (sixth aspect), the processing period is a period including an attack portion of the singing voice, and the synthesizing section expands the position closer to the starting point of the processing period. The processing period is extended so that

DESCRIPTION OF SYMBOLS 100... Sound processing apparatus, 11... Control apparatus, 12... Storage device, 13... Operation device, 14... Sound emitting device, 21... Signal analysis part, 22... Synthesis process part, 31... Attack process part, 32... Release process part , 33 . . . speech synthesizing unit.

Claims (10)

At the end point of the first steady period in which the fundamental frequency and spectrum shape are temporally stable in the first sound signal representing the singing voice, and at the second sound signal representing the reference voice having different acoustic characteristics from the singing voice The positions on the time axis of the first steady period and the second steady period were adjusted so that the end point of the second steady period in which the frequency and spectral shape were temporally stable coincided on the time axis. state, the processing period from a point in the first sound signal that is ahead of the end point of the first steady period by a specific time to a point in time when the singing voice is muted is the second sound signal. Extending to the time length of the expression period from the time point ahead of the end point of the two stationary periods by the specific time to the time point when the reference voice is silenced,
adding acoustic characteristics related to voice quality in the representation period of the second sound signal to the processing period in the decompressed first sound signal;
A speech processing method implemented by a computer. 2. The audio processing method according to claim 1, wherein in extending the processing period, the processing period is extended so that the degree of extension decreases as the position nearer to the end point of the processing period is extended. The processing period from the time when the singing voice of the first sound signal representing the singing voice starts to the starting point of the first steady period in which the fundamental frequency and the spectrum shape are temporally stable in the first sound signal, Of the second sound signal representing the reference voice having acoustic characteristics different from the singing voice, the second sound signal whose fundamental frequency and spectrum shape are temporally stable in the second sound signal from the time the reference voice starts. extended to the length of time of the representation period to the beginning of the stationary period,
adding acoustic characteristics related to voice quality in the representation period of the second sound signal to the processing period in the decompressed first sound signal;
A speech processing method implemented by a computer. 4. The audio processing method according to claim 3, wherein, in extending the processing period, the processing period is extended so that the closer the position is to the starting point of the processing period, the smaller the degree of extension. At the end point of the first steady period in which the fundamental frequency and spectrum shape are temporally stable in the first sound signal representing the singing voice, and at the second sound signal representing the reference voice having different acoustic characteristics from the singing voice The positions on the time axis of the first steady period and the second steady period were adjusted so that the end point of the second steady period in which the frequency and spectral shape were temporally stable coincided on the time axis. state, the processing period from a point in the first sound signal that is ahead of the end point of the first steady period by a specific time to a point in time when the singing voice is muted is the second sound signal. Extending to the time length of the expression period from the point in time preceding the end point of the two stationary periods by the specific time to the point in time when the reference voice is muted, the acoustic characteristics related to the voice quality in the expression period of the second sound signal are obtained. , a synthesis processing unit added to the processing period in the decompressed first sound signal. 6. The audio processing device according to claim 5, wherein the synthesizing unit extends the processing period so that the degree of extension decreases as the position nearer to the end point of the processing period. The processing period from the time when the singing voice of the first sound signal representing the singing voice starts to the starting point of the first steady period in which the fundamental frequency and the spectrum shape are temporally stable in the first sound signal, Of the second sound signal representing the reference voice having acoustic characteristics different from the singing voice, the second sound signal whose fundamental frequency and spectrum shape are temporally stable in the second sound signal from the time the reference voice starts. A synthesis processing unit that extends the time length of the expression period up to the start point of the stationary period, and adds acoustic characteristics related to voice quality in the expression period of the second sound signal to the processing period of the first sound signal after the expansion. A speech processing device comprising: 8. The audio processing device according to claim 7, wherein the synthesizing unit extends the processing period so that the closer the position is to the starting point of the processing period, the smaller the degree of extension. At the end point of the first steady period in which the fundamental frequency and spectrum shape are temporally stable in the first sound signal representing the singing voice, and at the second sound signal representing the reference voice having different acoustic characteristics from the singing voice The positions on the time axis of the first steady period and the second steady period were adjusted so that the end point of the second steady period in which the frequency and spectral shape were temporally stable coincided on the time axis. state, the processing period from a point in the first sound signal that is ahead of the end point of the first steady period by a specific time to a point in time when the singing voice is muted is the second sound signal. Extending to the time length of the expression period from the point in time preceding the end point of the two stationary periods by the specific time to the point in time when the reference voice is muted, the acoustic characteristics related to the voice quality in the expression period of the second sound signal are obtained. , a synthesis processing unit added to the processing period in the decompressed first sound signal;
A program that makes a computer function as a The processing period from the time when the singing voice of the first sound signal representing the singing voice starts to the starting point of the first steady period in which the fundamental frequency and the spectrum shape are temporally stable in the first sound signal, Of the second sound signal representing the reference voice having acoustic characteristics different from the singing voice, the second sound signal whose fundamental frequency and spectrum shape are temporally stable in the second sound signal from the time the reference voice starts. A synthesis processing unit that extends the time length of the expression period up to the start point of the stationary period, and adds acoustic characteristics related to voice quality in the expression period of the second sound signal to the processing period of the first sound signal after the expansion. ,
A program that makes a computer function as a

Images