JP2013033103A - Voice quality conversion device and voice quality conversion method - Google Patents

Abstract

[PROBLEMS] To convert an utterance state of input speech while maintaining naturalness.
A vocal tract sound source separation unit that separates input speech into vocal tract information and sound source information, and a volume in the oral cavity when the input speech is uttered from the vocal tract information separated by the vocal tract sound source separation unit. Based on the opening degree calculation unit 102 that calculates the opening degree corresponding to the input sound at every predetermined reproduction time and the utterance state conversion degree that indicates the degree of conversion of the opening degree, the conversion degree of the opening degree at each predetermined reproduction time is calculated. Based on the conversion rate determined by the opening degree conversion rate determination unit 105 and the opening degree conversion rate determination unit 105 to be determined, the opening degree calculation unit 102 calculates the vocal tract information separated by the vocal tract sound source separation unit 101. The vocal tract information conversion unit 103 converts the vocal tract information into the vocal tract information corresponding to the open degree after the open degree is converted at the conversion rate, the vocal tract information after the conversion by the vocal tract information conversion unit 103, and the vocal tract sound source Sound source information separated by the separation unit 101 Using, and a synthesizing unit 104 for generating a synthesized sound.
[Selection] Figure 5

Description

  The present invention relates to a voice quality conversion apparatus and method for converting voice quality of voice based on a voice synthesis technique.

  In recent years, with the development of speech synthesis technology, speech processed in various fields has come to be used.

  For example, the use of voice processed by a method such as fine correction of vocal sounds in music recording is increasing.

  Sound recording requires a dedicated recording studio environment. For this reason, when it is necessary to correct the sound once recorded, it is necessary to re-record the sound, which is expensive.

  When such re-recording is necessary, there are "wrong words" and "accent mistakes". In addition to things that are easy to notice during recording, such as these causes, there are changes in editing work after recording. It may be necessary. For example, there is a demand for changing the way of utterance (the utterance mode described later), for example, when it is desired to change the recorded voice to a slightly more active utterance, or to change to a slightly weaker utterance.

  For example, after recording the voice of a narration, if the producer requests that some of the audio be “like a little more cheerful”, the narrator needs to come back to the studio and start over. . However, enormous costs are required to actually re-record.

  As a countermeasure for such a case, there is a technique for enhancing speech. Conventionally, as a technique for improving speech intelligibility, there is a speech enhancement apparatus as shown in Patent Document 1. FIG. 15 is a configuration diagram of the speech enhancement device of Patent Document 1.

  The speech enhancement apparatus includes a speech enhancement unit 1 to which speech before enhancement from the speech decoder 4 is input, a noise estimation unit 3 to which a digital signal obtained from the analog signal S2 from the microphone 9 is input, and a speech decoder 4. And an emphasis characteristic determination unit 2 that determines an emphasis characteristic based on a signal from the noise estimation unit 3.

  The encoded data C1 is input to the speech decoder 4 and input to the speech enhancement unit 1 and the enhancement characteristic determination unit 2 as pre-enhancement speech.

  On the other hand, the analog signal S2 (ambient noise) input to the microphone 9 is converted into a digital signal by the A / D converter 8 and input to the speech encoder 7 and the noise estimation unit 3 as a transmission signal.

  The noise estimation unit 3 determines whether the transmission signal is a background noise interval or a speech interval. If the transmission signal is a noise interval, the noise estimation unit 3 estimates the noise characteristic and gives the noise characteristic to the enhancement characteristic determination unit 2.

  The emphasis characteristic determination unit 2 determines the emphasis characteristic based on both the characteristic of the speech before enhancement and the noise characteristic. This enhancement characteristic is input to the speech enhancement unit 1, and the speech enhancement unit 1 enhances the decoded speech based on the enhancement characteristic and provides the enhanced speech after enhancement to the D / A converter 5. The emphasized sound is converted into an analog signal by the D / A converter 5 and output from the speaker 6 as an analog signal S1.

  In this way, the nature of the ambient noise collected is estimated, and the speech is adaptively enhanced according to both the nature of the ambient noise and the nature of the unenhanced speech, thereby providing highly clear speech. To do.

JP 2004-289614 A

  In the speech enhancement device of Patent Document 1, speech is enhanced by changing the formant intensity according to the background noise and the characteristics of the input speech (formant center frequency).

  The voice quality of a natural utterance is affected by various factors such as the utterance speed of the voice, the position within the utterance, or the position within the accent phrase. For example, in a natural utterance, the beginning of a sentence is clearly uttered with high clarity, but at the end of the sentence, there is a tendency for pronunciation to be negligible and the intelligibility tends to decrease. Further, when a word is emphasized at the time of utterance, the voice quality of the word tends to be higher in clarity than when not emphasized.

  FIG. 1 shows a human vocal cord and vocal tract. Hereinafter, the principle of human voice generation will be described. As shown in FIG. 1, human speech is narrowed by an articulator such as a tongue when a sound source waveform generated by vibration of the vocal cord 1601 passes through a vocal tract 1604 composed of glottis 1602 to lips 1603. It is generated under the influence of. The analysis and synthesis type speech synthesis method analyzes human speech based on such a speech generation principle. Specifically, it is possible to convert the voice quality of the synthesized sound by separating the voice into vocal tract information and sound source information and modifying the separated parameters. For example, a model called a sound source vocal tract model is used as a speech analysis method. In the analysis by the sound source vocal tract model, the sound is separated into sound source information and vocal tract information based on the generation process. The voice quality can be changed by transforming the separated sound source information and vocal tract information.

  FIG. 2 shows the vocal tract transfer characteristics of the same vowel with the same preceding phoneme by the same speaker. In FIG. 2, the horizontal axis represents frequency, and the vertical axis represents spectral intensity.

  A curve 211 shown in FIG. 2 indicates a vocal tract transmission characteristic of the vowel part / a / of “mathematic” / ma / when uttering “/ memaigashishimasku /”. A curve 212 indicates a vocal tract transmission characteristic of a vowel part / a / of / ma / when uttered “no hot water (/ oyugadaseN /)”. Comparing vowels having the same preceding phoneme, it can be seen that the position and strength of the formant (upward peak) indicating the resonance frequency are greatly different.

  This is because the vowel part / a / corresponding to the curve 211 is close to the beginning of the sentence and is a content word, whereas the vowel part / a / corresponding to the curve 212 is close to the end of the sentence, In addition, it is a function word. In terms of audibility, the vowel part / a / corresponding to the curve 211 can be heard more clearly.

  Thus, in a natural utterance, the utterance method is different in the sentence. In other words, there is a difference in the way of conscious or unconscious utterance, such as “speech and clear speech” and “speaking and unclear speech”. Such a difference in utterance method is hereinafter referred to as “speech mode”.

  Such utterances vary not only in the phonological environment but also in various other linguistic and physiological influences.

  Since the speech enhancement device of Patent Document 1 performs formant intensity conversion based on background noise and the formant center frequency of the input speech without taking into account such a variation in utterance mode, The temporal change pattern of the utterance mode is different from that of the input voice.

  The temporal change of the utterance mode will be described with reference to the conceptual diagram of FIG. FIG. 3A shows a change in the utterance state (intelligibility) of each vowel included in the voice with respect to the voice “/ meigashimashixu /” uttered as the input voice. The region of X is a clear utterance and shows a phoneme with high intelligibility. The area Y indicates a phonation that is lazy and has a low intelligibility. For example, in this way, the first half is an utterance mode with high clarity, and the second half shows an utterance mode with low clarity.

  On the other hand, FIG. 3B shows a temporal change in the utterance mode of the emphasized speech when speech enhancement is performed by emphasizing or weakening formants. Since the formant intensity is converted only by the background noise and the formant center frequency of the input speech, the temporal variation pattern of the utterance mode cannot be maintained. For example, when the utterance mode fluctuates as shown in FIG. 3B, the utterance mode is emphasized such that the section X that clearly utters with high clarity and the section Y that utters with low clarity and utterance alternates alternately. Voice will be obtained.

  Thus, if the temporal change pattern of the utterance mode is different from the temporal change pattern of the input speech, the naturalness of the change of the utterance mode in the voice after voice quality conversion cannot be maintained, and as a result, the voice after voice quality conversion is lost. The problem was that the naturalness of the material deteriorated greatly.

  The present invention solves the above-described conventional problem, and by converting the voice quality while maintaining the time pattern of the utterance mode possessed by the input voice, naturalness at the time of voice quality conversion, in other words, does not reduce fluency. An object of the present invention is to provide a voice quality conversion device capable of converting the utterance mode.

  A voice quality conversion device according to an aspect of the present invention includes a vocal tract sound source separation unit that separates input speech into vocal tract information and sound source information, and the input from the vocal tract information separated by the vocal tract sound source separation unit. Based on the opening degree calculation unit that calculates the opening degree corresponding to the volume in the oral cavity at the time of sound production for each predetermined reproduction time of the input sound, and the utterance mode conversion degree indicating the degree of conversion of the opening degree The opening degree conversion rate determining unit that determines the conversion rate of the opening degree for each reproduction time and the vocal tract sound source separating unit separated based on the conversion rate determined by the opening degree conversion rate determining unit A vocal tract information conversion unit that converts vocal tract information into vocal tract information corresponding to an opening degree after the opening degree calculated by the opening degree calculation unit is converted at the conversion rate; and the vocal tract information conversion Vocal tract information after conversion by the part, and the vocal tract source Using said sound source information separated by the section, and a synthesizing unit for generating synthetic speech.

  According to this configuration, it is possible to save the time pattern of the utterance mode in the input voice when converting the utterance mode of the input voice. That is, when the utterance mode conversion ratio is increased, the opening degree is relatively increased, and the voice is clearly uttered and sounds as if uttered energetically. On the other hand, when the utterance mode conversion ratio is reduced, the opening degree becomes relatively small, and the voice is uttered lazyly and sounds as if uttered without power. As a result, the voice quality-converted voice stores the time pattern of the change in utterance mode, so that voice quality conversion without deteriorating naturalness (fluency) is possible.

  Note that the present invention can be realized not only as a voice quality conversion device including such a characteristic processing unit, but also as a voice quality conversion method including steps executed by the characteristic processing unit included in the voice quality conversion device. Can be realized. Also, it can be realized as a program for causing a computer to function as a characteristic processing unit included in the voice quality conversion device or a program for causing a computer to execute characteristic steps included in the voice quality conversion method. Such a program can be distributed via a computer-readable non-transitory recording medium such as a CD-ROM (Compact Disc-Read Only Memory) or a communication network such as the Internet. .

  According to the voice quality conversion device of the present invention, it is possible to maintain the temporal change pattern of the utterance mode in the input voice when converting the utterance mode of the input voice. In other words, the voice quality-converted voice stores the time pattern of the change in utterance mode, so that voice quality conversion that does not deteriorate naturalness (fluency) is possible.

Diagram showing the speech mechanism of human speech Diagram showing differences in vocal tract transmission characteristics due to different utterance modes Conceptual diagram showing temporal variation of vocalization Diagram showing differences in vocal tract cross-sectional area function due to differences in vocalization modes The block diagram which shows the functional structure of the voice quality conversion apparatus in embodiment of this invention Diagram showing examples of vocal tract cross-sectional area function A diagram showing the time pattern of the degree of opening in utterance A diagram showing an example of conversion of time pattern of aperture Diagram showing examples of conversion coefficients for vocal tract cross-sectional area function Diagram showing conversion example of vocal tract cross section function Diagram showing vocal tract transmission characteristics before and after opening degree conversion The flowchart which shows operation | movement of the voice quality conversion apparatus in embodiment of this invention. The block diagram which shows the functional structure of the voice quality conversion apparatus in the modification of embodiment of this invention. The flowchart which shows operation | movement of the voice quality conversion apparatus in the modification of embodiment of this invention. Configuration of conventional voice quality conversion device

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. Each of the embodiments described below shows a preferred specific example of the present invention. The numerical values, the constituent elements, the arrangement positions and connection forms of the constituent elements, the steps, the order of the steps, and the like shown in the following embodiments are examples, and are not intended to limit the present invention. The invention is limited only by the claims. Therefore, among the constituent elements in the following embodiments, constituent elements that are not described in the independent claims indicating the highest concept of the present invention are not necessarily required to achieve the object of the present invention. It will be described as constituting a preferred form.

  A voice quality conversion device according to an embodiment of the present invention includes: a vocal tract sound source separation unit that separates input speech into vocal tract information and sound source information; and the vocal tract information separated by the vocal tract sound source separation unit, Based on the opening degree calculation unit for calculating the opening degree corresponding to the volume in the oral cavity at the time of utterance of the input sound for each predetermined reproduction time of the input sound, and the utterance mode conversion degree indicating the degree of conversion of the opening degree, Based on the conversion rate determined by the openness conversion rate determination unit and the openness conversion rate determination unit that determines the conversion rate of the openness for each predetermined reproduction time, the vocal tract sound source separation unit A vocal tract information conversion unit that converts the vocal tract information into vocal tract information corresponding to an opening degree after the opening degree calculated by the opening degree calculation unit is converted at the conversion rate; and the vocal tract information The vocal tract information after conversion by the conversion unit and the vocal tract sound source Using said sound source information separated by the releasing unit, and a synthesizing unit for generating synthetic speech.

  According to this configuration, when converting the utterance mode of the input voice, it is possible to save the time pattern (temporal change sequence) of the utterance mode in the input voice. That is, when the utterance mode conversion ratio is increased, the opening degree is relatively increased, and the voice is clearly uttered and sounds as if uttered energetically. On the other hand, when the utterance mode conversion ratio is reduced, the opening degree becomes relatively small, and the voice is uttered lazyly and sounds as if uttered without power. As a result, the voice quality-converted voice stores the time pattern of the change in utterance mode, so that voice quality conversion without deteriorating naturalness (fluency) is possible.

  Preferably, the opening degree calculation unit calculates a vocal tract cross-sectional area function from the vocal tract information separated by the vocal tract sound source separation unit at each predetermined reproduction time, and divides the oral cavity into a plurality of sections In this case, the opening degree is calculated as the sum of the vocal tract cross-sectional areas of the sections indicated by the calculated vocal tract cross-sectional area function.

  More preferably, the above-described voice quality conversion device further includes a vowel section extraction unit that extracts a vowel section that is a time section of a vowel included in the input speech, and the vocal tract information conversion unit includes the vocal tract sound source separation. Only the vocal tract information in the vowel segment extracted by the vowel segment extraction unit is converted from the vocal tract information separated by the unit.

  According to this configuration, it is possible to prevent excessive conversion of the vocal tract shape by not converting the opening degree to the consonant section in which the vocal tract information changes quickly in time, so that the sound quality is deteriorated during the conversion of the voice quality. Can be suppressed.

  The opening degree conversion rate determination unit may determine the utterance state conversion degree as the conversion rate for each predetermined reproduction time.

  According to this configuration, vocal tract information can be converted by directly using the utterance mode conversion degree inputted from the outside.

  Preferably, the opening degree conversion rate determining unit calculates a product of a dynamic range in the temporal change series of the opening degree calculated by the opening degree calculating unit and the utterance state conversion degree at each predetermined reproduction time. The conversion rate by which the opening degree calculated by the opening degree calculation unit is multiplied to calculate the opening degree after the conversion so as to coincide with the dynamic range in the temporal change sequence of the opening degree after conversion. decide.

  According to this configuration, it is possible to convert the dynamic range of the temporal pattern (temporal change series) of the opening degree of the input voice based on the utterance mode conversion degree input from the outside. For this reason, it is possible to convert the voice quality to a more crisp voice.

  In addition, the opening degree conversion rate determination unit may calculate an average value of the opening degree in the temporal change series of the opening degree calculated by the opening degree calculation unit and the utterance state conversion degree at each predetermined reproduction time. The opening degree calculated by the opening degree calculating unit to calculate the opening degree after the conversion so that the product coincides with the average value of the opening degree in the temporal change series of the opening degree after the conversion. The conversion rate to be multiplied may be determined.

  According to this configuration, it is possible to convert the average value of the time pattern (temporal change series) of the opening degree of the input speech based on the utterance mode conversion degree input from the outside. For this reason, it is possible to perform voice quality conversion in which the utterance mode is converted on average.

  The vocal tract information conversion unit converts a vocal tract cross-sectional area indicating the vocal tract information separated by the vocal tract sound source separation unit based on the conversion rate determined by the opening degree conversion rate determination unit. By doing so, the vocal tract information separated by the vocal tract sound source separation unit may be converted.

  Preferably, the vocal tract information conversion unit, when dividing the oral cavity into a plurality of sections, converts the conversion coefficient corresponding to the conversion rate and determined for each section to the vocal tract cross-sectional area for each section. By multiplying, the vocal tract information separated by the vocal tract sound source separation unit is converted.

  According to this configuration, when the vocal tract cross-sectional area function is deformed, conversion can be performed with a higher degree of freedom, and therefore it is possible to prevent deterioration of the naturalness of the voice whose voice quality has been converted.

  More preferably, the absolute value of the difference between the conversion coefficient in the section closer to the lip and the conversion coefficient in the adjacent section is larger as the conversion coefficient in the section closer to the lips.

  According to this configuration, when the vocal tract cross-sectional area function is converted, it becomes possible to perform a deformation that more closely matches the utterance mechanism, and it is possible to prevent deterioration of the naturalness of the voice whose voice quality has been converted.

(Embodiment)
In the embodiment of the present invention, a method is described in which voice quality conversion is performed on an original voice (input voice) by converting a degree of opening representing an utterance mode described later based on a conversion rate.

  As already described, when performing voice quality conversion, it is important to maintain temporal variation of the utterance mode in the input voice. The utterance mode is, for example, utterance with clear and high intelligibility, or utterance with low intelligibility.

  The utterance mode is affected by, for example, the speech rate, the position in the utterance, or the position in the accent phrase. For example, in a natural utterance, the beginning of a sentence is clearly and clearly uttered, but at the end of the sentence, lazyness occurs and the intelligibility tends to decrease. In addition, in the input speech, the utterance mode when a certain word is emphasized is different from the case where the word is not emphasized.

  FIG. 4 (a) shows the logarithmic vocal tract cross-sectional area function of / ma / vowel part / a / of “vertigo” when the above-mentioned “/ memaigasimasxu /” is spoken. (B) shows the logarithmic vocal tract cross-sectional area function of the vowel part / a / of / ma / when uttered “no hot water (/ oyugadaseN /)”.

  Since the vowel part / a / in FIG. 4A is a sound that is close to the beginning of a sentence and included in a content word (independent word), the utterance is clearly and clearly spoken. On the other hand, since the vowel part / a / in FIG. 4B is close to the end of the sentence, the utterance mode is lazy, and the clarity is low.

  The inventors of the present application have found that the voicing mode is related to the volume in the oral cavity by carefully observing the relationship between the voicing mode and the logarithmic vocal tract cross-sectional area function.

  That is, as the volume in the oral cavity is larger, the utterance state tends to be clearer, and conversely, as the volume in the oral cavity is smaller, the utterance state tends to be lazy and have a lower clarity.

  By using the intraoral volume that can be calculated from the voice as an index of the opening degree, it is possible to control the utterance mode of the voice.

  In the present embodiment, a synthesized sound with little deterioration in naturalness can be generated by using the degree of opening indicating the intraoral volume to preserve the temporal variation of the vocalization mode and perform voice quality conversion.

  FIG. 5 is a block diagram illustrating a functional configuration of the voice quality conversion apparatus according to the embodiment. The voice quality conversion apparatus includes a vocal tract sound source separation unit 101, an opening degree calculation unit 102, an opening degree conversion rate determination unit 105, a vocal tract information conversion unit 103, and a synthesis unit 104.

  The vocal tract sound source separation unit 101 receives input speech. The vocal tract sound source separation unit 101 separates the received input speech into vocal tract information and sound source information.

  The opening degree calculation unit 102 uses the vocal tract information separated by the vocal tract sound source separation unit 101 to calculate the opening degree of each frame of the input speech. That is, the opening degree calculation unit 102 uses the vocal tract information separated by the vocal tract sound source separation unit 101 to set the opening degree corresponding to the volume in the oral cavity when the input voice is uttered, to a predetermined reproduction time (frame) of the input voice. Calculate every time.

  The opening degree conversion rate determination unit 105 calculates a rate that is the conversion rate of the opening degree in each frame calculated by the opening degree calculation unit 102 based on the utterance mode conversion degree indicating the conversion degree of the opening degree input from the outside. calculate.

  The vocal tract information conversion unit 103 calculates the vocal tract information separated by the vocal tract sound source separation unit 101 based on the open degree conversion rate determined by the open degree conversion rate determination unit 105 by the open degree calculation unit 102. The opening degree is converted into vocal tract information corresponding to the opening degree after being converted at the opening degree conversion rate.

  The synthesis unit 104 generates a synthesized sound using the vocal tract information converted by the vocal tract information conversion unit 103 and the sound source information separated by the vocal tract sound source separation unit 101.

  With the voice quality conversion apparatus configured as described above, voice quality conversion can be performed while maintaining temporal variation of the utterance mode of the input voice.

  Hereinafter, each component will be described.

<Vocal tract sound source separation unit 101>
The vocal tract sound source separation unit 101 separates vocal tract information and sound source information by applying a vocal tract sound source model (speech generation model that models a speech utterance mechanism) to input speech. There is no limitation on the vocal tract sound source model to be used, and any model may be used.

  For example, there is a linear prediction model (LPC model) as a vocal tract sound source model. The linear prediction model predicts a certain sample value s (n) of a speech waveform from p sample values before the sample value, and the sample value s (n) can be expressed as Equation 1.

The coefficient α _i (i = 1 to p) for p sample values can be calculated by using a correlation method, a covariance method, or the like. When the calculated coefficient α _i is used, the input audio signal can be generated by Equation 2.

  Here, S (z) is a value after the z conversion of the voice signal s (n), U (z) is a value after the z conversion of the voiced sound source signal u (n), and the input voice S ( z) represents a signal obtained by inverse filtering the vocal tract information 1 / A (z).

Further, the PARCOR coefficient (partial autocorrelation coefficient) may be calculated using the linear prediction coefficient α _i analyzed by the LPC analysis. It is known that the PARCOR coefficient has better interpolation characteristics than the linear prediction coefficient. The PARCOR coefficient can be calculated by using the Levinson-Durbin-Itakura algorithm. The PARCOR coefficient has the following characteristics.

  (Characteristic 1) The lower-order coefficient has a greater influence on the spectrum due to the fluctuation, and the higher the order, the smaller the influence of the fluctuation.

  (Characteristic 2) The influence of high-order coefficient fluctuations covers the entire area flatly.

  In the following description, the PARCOR coefficient is used as the vocal tract information. Note that the vocal tract information to be used is not limited to the PARCOR coefficient, and a linear prediction coefficient may be used. Further, a line spectrum pair (LSP) may be used.

  Further, an ARX model (Autogressive with exogenous input) may be used as the vocal tract sound source model. ARX analysis is significantly different from LPC analysis in that a mathematical sound source model is used as a sound source. Also, in the ARX analysis, unlike the LPC analysis, the vocal tract and sound source information can be more accurately separated even when a plurality of basic periods are included in the analysis interval (Non-patent Document 1: “Sound source pulse train is considered. Robust ARX speech analysis method, ”Acoustical Society of Japan, Vol. 58, No. 7, 2002, pp. 386-397).

  In ARX analysis, speech is generated by the generation process shown in Equation 3. In Equation 3, S (z) represents a value after the z conversion of the audio signal s (n). U (z) represents a value after z conversion of the voiced sound source signal u (n). E (z) represents a value after the z conversion of the silent noise source e (n). That is, in ARX analysis, voiced sound is generated by the first term on the right side of Equation 3, and unvoiced sound is generated by the second term on the right side.

  At this time, the sound model shown in Formula 4 is used as a model of the voiced sound source signal u (t) = u (nTs) (Ts is a sampling period).

  Where AV is the voiced sound source amplitude, T0 is the pitch period, and OQ is the glottal opening rate. In the case of voiced sound, the first term on the upper right side of Equation 4 is used, and in the case of unvoiced sound, the second term on the upper right side of Equation 4 is used. The glottal opening rate OQ indicates the rate at which the glottal is opened in one pitch period. It is known that the greater the glottal opening rate OQ, the softer the voice.

  The ARX analysis has the following advantages compared to the LPC analysis.

  (Advantage 1) Since the analysis is performed by arranging sound source pulse trains corresponding to a plurality of pitch periods in the analysis window, vocal tract information can be stably extracted even in high pitch sounds such as women or children.

  (Advantage 2) Particularly, the vocal tract sound source separation performance of narrow vowels such as / i / and / u /, where the pitch frequency F0 and the first formant frequency F1 are close to each other, is high.

  In the voiced sound section, U (z) can be obtained by inverse filtering the input speech S (z) with the vocal tract information 1 / A (z), as in the case of LPC analysis.

  Also in the ARX analysis, the vocal tract information 1 / A (z) has the same format as the system function in the LPC analysis. Therefore, the PARCOR coefficient may be obtained by the same method as the LPC analysis.

<Openness Calculation Unit 102>
The opening degree calculation unit 102 uses the vocal tract information separated by the vocal tract sound source separation unit 101 to calculate an opening degree corresponding to the volume in the oral cavity for each voice frame of the input voice. Each voice frame of the input voice includes information on the time when the input voice was uttered.

  The opening degree calculation unit 102 calculates a vocal tract cross-sectional area function from the vocal tract information separated by the vocal tract sound source separation unit 101 at each predetermined reproduction time, and calculates the calculated voice when the oral cavity is divided into a plurality of sections. The opening degree is calculated as the sum of the vocal tract cross-sectional areas of each section indicated by the road cross-sectional area function.

  Specifically, the opening degree calculation unit 102 calculates the vocal tract cross-sectional area function using Equation 5 from the PARCOR coefficient extracted as the vocal tract information.

Here, k _i represents the i-th order PARCOR coefficient, A _i represents the i-th vocal tract cross-sectional area, and A _{N + 1} = 1.

  FIG. 6 is a diagram illustrating a logarithmic vocal tract cross-sectional area function of a vowel / a / of a certain utterance. The vocal tract from the glottis to the lips is divided into 11 sections, where section 11 represents the glottis and section 1 represents the lips.

  In FIG. 6, the shaded area can be considered to be generally in the oral cavity. Therefore, when section 1 to section T are considered to be in the oral cavity (T = 5 in FIG. 6), the opening degree V can be defined by Equation 6. It is desirable to change T according to the order of LPC analysis or ARX analysis. For example, in the case of 10th order LPC analysis, about 3 to 5 is desirable. However, the specific order is not limited.

  The opening degree calculation unit 102 calculates the opening degree V defined as described above for each frame of the input sound.

  FIG. 7 shows a temporal change in the degree of opening calculated by Expression 6 in the utterance “Dizzy (/ memaigasimasxu /)”.

  As described above, the opening degree fluctuates with time. If this time fluctuation pattern is broken, the naturalness deteriorates.

  By using the opening degree (volume in the oral cavity) calculated using the vocal tract cross-sectional area function in this way, not only the opening of the lips but also the shape of the oral cavity that cannot be observed directly from the outside (for example, the tongue) Position) can also be considered.

  In the above description, the unit of calculation of the opening degree is the frame, but is not limited to the frame. For example, the aperture may be calculated in units of phonemes. In this case, the opening degree of the phoneme center frame may be the opening degree of the phoneme. Further, an average value of the opening degree in the phoneme may be used as the opening degree of the phoneme. When a phoneme is used as a unit, the opening degree in each frame can be calculated by interpolating from the opening degree of phonemes before and after the frame. In addition to this, a mora or a syllable may be used as a unit for calculating the opening degree.

<Openness Conversion Rate Determination Unit 105>
The opening degree conversion rate determining unit 105 calculates the opening degree conversion rate r _i in each frame i calculated by the opening degree calculating unit 102. Thus, by changing the utterance mode of the input voice, it is possible to obtain a voice having a utterance mode different from the input voice. Specifically, it is possible to convert to a clearer and more active utterance mode, or to convert a uttered and lesser utterance mode. The opening degree conversion rate r _i is a coefficient by which the opening degree is multiplied, and the opening degree can be converted by multiplying the opening degree by the opening degree conversion rate r _i .

As a method of calculating the opening degree conversion rate r _i , for example, the utterance mode conversion degree q may be directly used as the opening degree conversion rate r _i for each frame. Specifically, the opening degree conversion rate determination unit 105 may calculate the opening degree conversion rate r _i of each frame according to Equation 7.

  Here, i indicates a frame number.

When the aperture degree conversion rate r _i calculated in this way is used, the aperture degree V ′ _i after conversion (FIG. 8) with respect to the opening degree calculated by the opening degree calculating unit 102 (broken line in FIG. 8A). The solid line (a) can be obtained by Equation 8.

  By converting the opening degree in this way, the opening degree of the input voice is converted according to the utterance mode conversion degree. As a result, it is possible to generate an opening degree pattern corresponding to a case where utterances having different utterance modes are made.

Further, the dynamic range in the change sequence of the aperture may be adjusted while maintaining the time pattern of the change in the aperture. That is, the opening degree conversion rate determination unit 105, for each frame, the product of the dynamic range and the utterance aspect conversion degree of temporal change sequence of the opening degree V _i where the opening degree calculation unit 102 calculates the opening after conversion The opening degree conversion rate r _i is determined so as to coincide with the dynamic range in the temporal change series of the degree V ′ _i .

Specifically, as shown by the broken line in FIG. 8B, when the aperture time pattern calculated by the aperture calculation unit 102 is V _i , the aperture conversion rate determination unit 105 calculates the input voice according to Equation 9 as follows. An opening degree conversion rate r _i for converting the dynamic range of the opening degree is calculated.

Here, E (x) represents a function for calculating an average value of x. The opening degree V ′ _i obtained by converting the opening degree V _i using r _i calculated by Expression 9 defines the dynamic range by Expression 10. The relationship between the dynamic range DR before the conversion and the dynamic range DR ′ after the conversion is expressed as in Expression 11 using the utterance mode conversion degree q. By using the opening degree conversion rate r _i calculated in this way, it is possible to convert the dynamic range while maintaining the movement of the time pattern of the opening degree.

When the aperture degree conversion rate r _i calculated in this way is used, the aperture degree V ′ _i after conversion can be expressed by Expression 8.

  The aperture time pattern converted as described above is as shown by the solid line in FIG. 8B, and the dynamic range of the aperture time pattern can be converted.

  For example, when the utterance state conversion degree q is larger than 1, the dynamic range of the pattern of the opening degree is expanded. Therefore, the voice of the section with a large opening degree included in the input voice is converted so that the opening degree becomes larger. Audio in a section with a small opening degree is converted so that the opening degree becomes smaller. As a result, it is possible to convert the input voice to a sharp voice. Specifically, when the opening degree increases, the voice is clearly uttered and sounds as if it was uttered energetically.

  On the other hand, when the utterance state conversion degree q is smaller than 1, the dynamic range is reduced. Therefore, the voice of the section with a large opening degree included in the input voice is converted so that the opening degree becomes smaller. For this reason, the variation of the utterance mode is reduced, and the whole voice is less crisp and can be converted into a calm voice. Specifically, when the opening degree is small, the voice is uttered lazyly and sounds as if it was uttered without power.

Further, the opening degree conversion rate determining unit 105 may adjust the opening degree so as to shift the average value in the changing series of the opening degree while holding the time pattern of the opening degree change. That is, the opening degree conversion rate determination unit 105, for each frame, the product of the average value and the utterance aspect conversion degree of the opening degree V _i in the time variation sequence of aperture of the opening degree calculation unit 102 is calculated, converted The opening degree conversion rate r _i is determined so as to coincide with the average value of the opening degree in the temporal change series of the later opening degree V ′ _i .

Specifically, as shown by the broken line in FIG. 8C, when the aperture time pattern calculated by the aperture calculation unit 102 is V _i , the aperture conversion rate determination unit 105 uses the expression 12 to calculate the input voice. An opening degree conversion rate r _i that shifts (converts) the average value of the opening degree is calculated.

When the aperture degree conversion rate r _i calculated in this way is used, the aperture degree V ′ _i after conversion can be expressed by Expression 8.

  The aperture time pattern converted as described above is as shown by the solid line in FIG. 8C, and can be converted so as to shift the average value of the aperture time pattern.

  When the utterance state conversion degree q is larger than 1, it can be uttered uniformly and clearly and converted into a sound with a healthy feeling. When the utterance mode conversion degree q is smaller than 1, it is possible to convert the voice into a voice uttered without any difficulty by uttering evenly lazy.

Note that the calculation method of the aperture conversion rate r _i is not limited to the above-described method, and any method may be used as long as the size of the aperture conversion is maintained while maintaining the time pattern of the aperture.

<Vocal tract information conversion unit 103>
The vocal tract information conversion unit 103 converts the vocal tract information calculated by the vocal tract sound source separation unit 101 using the opening degree conversion rate r _i of the frame i calculated as described above. That is, the vocal tract information conversion unit 103 converts the vocal tract cross-sectional area indicating the vocal tract information separated by the vocal tract sound source separation unit 101 based on the open degree conversion rate determined by the open degree conversion rate determination unit 105. As a result, the vocal tract information separated by the vocal tract sound source separation unit 101 is converted.

Specifically, the vocal tract information conversion unit 103 calculates, in each frame, a conversion function C ^k _i indicating a conversion coefficient for the vocal tract cross-sectional area indicated by the vocal tract cross-sectional area function using Expression 13.

Here, k is a section number in the vocal tract cross-sectional area function, and r _i represents the opening degree conversion rate at the lip position. Further, β is a predetermined constant representing the strength of the degree of change in the opening degree in each section. The larger β is, the steeper change in the conversion function C ^k _i is in the lip side section.

9, beta = 0.7 and the _{r i} -0.4, -0.2,0,0.2,0.4, the conversion in the case of changing the function ^C _{k i} is the transform coefficients shown Ci It is a graph which shows.

As can be seen from the figure, when the opening degree conversion rate r _i is negative, the vocal tract cross-sectional area gradually decreases as the conversion coefficient Ci decreases toward the lips, and the opening degree conversion rate r _i In the positive case, the vocal tract cross-sectional area increases as the conversion coefficient Ci increases. As shown in Expression 14, such a conversion coefficient Ci can be adjusted by multiplying the vocal tract cross-sectional area function A _i of the input speech by opening. Note that Expression 14 shows an expression ^obtained by multiplying the conversion function C ^k _i by the vocal tract cross-sectional area function A _i ^k .

  In other words, when the oral cavity is divided into a plurality of sections, the vocal tract information conversion unit 103 multiplies the vocal tract cross-sectional area for each section by a conversion coefficient corresponding to the opening degree conversion rate and determined for each section. Thus, the vocal tract information separated by the vocal tract sound source separation unit 101 is converted. The conversion function has a shape as shown in FIG. For this reason, as for the conversion coefficient, the absolute value of the difference with the conversion coefficient of an adjacent area is so large that the conversion coefficient of the area close | similar to a lip is.

Note that the conversion function C ^k _i for the vocal tract cross-sectional area function A _i ^k is not limited to Equation 13, and the value of the conversion coefficient C _i greatly changes toward the lip for the vocal tract cross-sectional area function A _i ^k . Any function can be used. Alternatively, the conversion function C ^k _i may be designed so as to uniformly convert the cross-sectional area of the section corresponding to the oral cavity.

  FIG. 10 shows how the vocal tract cross-sectional area changes when the aperture is converted using Equation 14 for the vowel / a /. FIG. 10A is a logarithmic vocal tract cross-sectional area function of a uttered vowel / a /. The horizontal axis indicates the section number, and the vertical axis indicates the value of the vocal tract cross-sectional area. On the other hand, FIG. 10B is a vocal tract cross-sectional area function when the opening degree is converted by Expression 14 with respect to the vocal tract cross-sectional area function of FIG. The horizontal axis and the vertical axis are the same as in FIG. As can be seen from FIG. 10, when comparing the vocal tract cross-sectional area before and after conversion, the vocal tract cross-sectional area gradually decreases from the vicinity of section 5 to the lips (section 1). In this way, by converting the vocal tract cross-sectional area, the movement in the oral cavity accompanying actual speech is simulated. That is, it simulates the movement in the oral cavity in which the vocal tract cross-sectional area changes more as it is closer to the lips.

  FIG. 11 is a diagram showing the vocal tract transmission characteristics of the vocal tract information obtained as described above. In the figure, the horizontal axis indicates the frequency, and the vertical axis indicates the spectrum intensity.

  In FIG. 11, a solid line 1101 indicates a vocal tract transfer characteristic corresponding to the vocal tract cross-sectional area function of FIG. A broken line 1102 indicates a vocal tract transfer characteristic corresponding to the vocal tract cross-sectional area function of FIG. That is, the broken line 1102 indicates the vocal tract transmission characteristic of the vocal tract information after being converted by the vocal tract information conversion unit 103. Since the vocal tract information conversion unit 103 converts the opening degree to be small, the formant strength tends to be weakened. On the other hand, the formant center frequency of the vocal tract transfer characteristic indicated by the broken line 1102 is also moved in comparison with the vocal tract transfer characteristic before conversion (solid line 1101) as the opening degree is converted. As described above, it is possible to change not only the formant intensity but also the formant center frequency by converting the opening degree as in the conventional speech enhancement apparatus. In addition, since the temporal variation pattern of the opening degree retains the same temporal variation as the variation pattern of the input sound, it is possible to retain naturalness as a sound.

<Combining unit 104>
The synthesizing unit 104 synthesizes speech using the vocal tract information (the vocal tract cross-sectional area function) A ′ _i ^k converted by the vocal tract information conversion unit 103 and the sound source information separated by the vocal tract sound source separation unit 101. To do. The synthesis method is not particularly limited. However, when the PARCOR coefficient is used as the vocal tract information, the vocal tract cross-sectional area function A ′ _i ^k is converted into the PARCOR coefficient using Equation 5, and the PARCOR synthesis is used. That's fine. Alternatively, synthesis may be performed after conversion from PARCOR coefficients to LPC coefficients, or formants may be extracted and synthesized by formant synthesis. Furthermore, the LSP coefficient may be calculated from the PARCOR coefficient and synthesized by LSP synthesis.

<Flowchart>
The specific operation of the voice quality conversion apparatus according to the present embodiment will be described with reference to the flowchart shown in FIG.

  In step S101, the vocal tract sound source separation unit 101 separates the input sound into vocal tract information and sound source information.

  In step S102, the opening degree calculation unit 102 calculates the opening degree in each frame included in the input speech using the vocal tract information separated in step S101.

  In step S103, the opening degree conversion rate determination unit 105 determines the opening degree conversion rate based on the opening degree of each frame of the input speech calculated in step S102 and the utterance mode conversion degree that is input separately.

  In step S104, the vocal tract information conversion unit 103 converts the vocal tract information calculated in step S101 based on the opening degree conversion rate determined in step S103.

  In step S105, the synthesis unit 104 generates a synthesized sound using the vocal tract information converted in step S104 and the sound source information separated in step S101.

<Effect>
According to this configuration, when converting the utterance mode of the input voice, it is possible to convert the utterance mode based on the utterance mode conversion degree while preserving the temporal change pattern of the utterance mode in the input voice. That is, when the utterance state conversion ratio is increased, the opening degree is relatively increased, and the voice is clearly uttered, and the sound can be heard as if uttered energetically. On the other hand, when the utterance mode conversion ratio is made small, the opening degree becomes relatively small, and there is an effect that the voice is uttered lazyly and sounds as if uttered without power. In addition, the voice quality-converted voice by the voice quality conversion apparatus according to the embodiment stores the time pattern of the change of the utterance mode, so that voice quality conversion without deteriorating the naturalness (fluency) at the time of voice quality conversion becomes possible. .

  Specifically, as shown by the broken lines in FIGS. 8A to 8C, the change in the utterance mode (intelligibility) of the input speech and the utterance mode of the voice after voice quality conversion as shown by the solid line Therefore, the sound quality is not deteriorated due to the unnaturalness of the voice utterance.

  In the present embodiment, the description is made using Japanese speech, but the present invention is not limited to Japanese, and voice quality conversion can be performed in other languages such as English as well.

  For example, when normal utterance is “Can I make a phone call this plain?”, The plain [ei] at the end of the sentence when uttered and “May I have a thermometer?” At the beginning of the sentence when saying “May I have a thermometer?” The utterance mode differs from ei] (inside [], an international phonetic symbol (IPA International Phonetic Alphabet)). In addition, as in Japanese, the utterance mode changes depending on the position in the sentence, the content word or function word, or the presence or absence of emphasis, so if only the formant intensity is converted, the utterance mode will change over time as in Japanese. The change pattern collapses. Therefore, the naturalness of the voice quality converted speech deteriorates. Therefore, even in English, by converting the voice quality based on the opening degree, it is possible to convert to the target voice quality while preserving the temporal change pattern of the utterance mode in the input voice. As a result, the voice quality-converted voice stores the time pattern of the change of the utterance mode, so that voice quality conversion without deteriorating the naturalness (fluency) at the time of voice quality conversion can be performed.

(Modification)
Next, a modification of the present embodiment will be described. In this modification, the voice quality of the vowel section of the input speech is converted.

  FIG. 13: is a block diagram which shows the modification of the voice quality conversion apparatus of embodiment of this invention. In FIG. 13, the same components as those in FIG.

  The voice quality conversion device according to the present modification is provided with a new vowel segment extraction unit 201 in the configuration of the voice quality conversion device shown in FIG. 5, and uses a vowel vocal tract information conversion unit 202 instead of the vocal tract information conversion unit 103. It has the composition which was.

<Vowel section extraction unit 201>
The vowel section extraction unit 201 extracts a vowel section that is a time section of a vowel included in the input speech. The extraction method of the vowel section is not particularly limited. For example, the vowel segment extraction unit 201 may recognize a phoneme sequence of the input speech using a speech recognition technique and extract a vowel segment from the recognition result. Alternatively, the vowel section extraction unit 201 calculates the similarity between the speech waveform template of the vowel held in advance and the speech waveform of the input speech, and when the calculated similarity is greater than a preset threshold, May be extracted as

<Vowel vocal tract information conversion unit 202>
The vowel vocal tract information conversion unit 202 converts only the vocal tract information in the vowel segment extracted by the vowel segment extraction unit 201 among the vocal tract information separated by the vocal tract sound source separation unit 101.
A specific vocal tract information conversion method is the same as the vocal tract information conversion method performed by the vocal tract information conversion unit 103. Therefore, detailed description thereof will not be repeated.

<Flowchart>
The specific operation of the voice quality conversion apparatus according to this modification will be described with reference to the flowchart shown in FIG.

  In step S201, the vowel section extraction unit 201 extracts a vowel section from the input speech.

  In step S101, the vocal tract sound source separation unit 101 separates the input sound into vocal tract information and sound source information.

  In step S102, the opening degree calculation unit 102 calculates the opening degree in each frame included in the input speech using the vocal tract information separated in step S101.

  In step S103, the opening degree conversion rate determination unit 105 determines the opening degree conversion rate based on the opening degree of each frame of the input speech calculated in step S102 and the utterance mode conversion degree that is input separately.

  In step S202, the vowel vocal tract information conversion unit 202 extracts the vocal tract in the vowel section extracted in step S201 from the vocal tract information calculated in step S101 based on the opening degree conversion rate determined in step S103. Transform information.

  In step S105, the synthesis unit 104 generates a synthesized sound using the vocal tract information converted in step S202 and the sound source information separated in step S101. The vocal tract information is not converted for the consonant section. For this reason, the synthesis unit 104 generates a synthesized sound for the consonant section using the vocal tract information and the sound source information separated by the vocal tract sound source separation unit 101.

<Effect>
According to this configuration, when converting the utterance mode of the input voice, it is possible to convert the utterance mode based on the utterance mode conversion ratio while preserving the temporal change pattern of the utterance mode in the input voice. As a result, the voice quality-converted speech stores the time pattern of the change in the utterance mode, so that it is possible to perform voice quality conversion without deteriorating the naturalness (fluency) at the time of voice quality conversion. In addition, since the vocal tract information changes quickly in time and is large in the consonant section, the sound quality deteriorates due to the deformation of the vocal tract information compared to the vowel section where the duration is relatively long and the movement of the vocal tract information is slow. Is big. From this, it is possible to suppress deterioration in sound quality during voice quality conversion by converting only the vowel section. Further, since the vowel section has a longer duration than the consonant section, it is possible to maintain the effect of voice quality conversion.

  An input voice acquisition unit (not shown) may acquire the input voice. The input sound acquisition unit acquires the time when the input sound is acquired and the input sound in association with each other. This time corresponds to a predetermined reproduction time. The input voice acquisition unit transmits the acquired input voice to the vocal tract sound source separation unit 101. The input voice acquisition unit may be configured with a microphone or the like. Alternatively, the input voice may be stored in a storage unit included in the input voice acquisition unit.

  Note that each of the above devices may be specifically configured as a computer system including a microprocessor, a ROM, a RAM, a hard disk drive, a display unit, a keyboard, a mouse, and the like. A computer program is stored in the RAM or hard disk drive. Each device achieves its functions by the microprocessor operating according to the computer program. Here, the computer program is configured by combining a plurality of instruction codes indicating instructions for the computer in order to achieve a predetermined function.

  Furthermore, some or all of the constituent elements constituting each of the above-described devices may be configured by one system LSI (Large Scale Integration). The system LSI is an ultra-multifunctional LSI manufactured by integrating a plurality of components on a single chip, and specifically, a computer system including a microprocessor, ROM, RAM, and the like. . A computer program is stored in the RAM. The system LSI achieves its functions by the microprocessor operating according to the computer program.

  Furthermore, some or all of the constituent elements constituting each of the above-described devices may be configured from an IC card that can be attached to and detached from each device or a single module. The IC card or module is a computer system that includes a microprocessor, ROM, RAM, and the like. The IC card or the module may include the super multifunctional LSI described above. The IC card or the module achieves its function by the microprocessor operating according to the computer program. This IC card or this module may have tamper resistance.

  Further, the present invention may be the method described above. Further, the present invention may be a computer program that realizes these methods by a computer, or may be a digital signal that is formed by the computer program.

  Furthermore, the present invention relates to a non-transitory recording medium that can read the computer program or the digital signal, such as a flexible disk, a hard disk, a CD-ROM, an MO, a DVD, a DVD-ROM, a DVD-RAM, a BD ( It may be recorded on a Blu-ray Disc (registered trademark)), a semiconductor memory, or the like. The digital signal may be recorded on these non-temporary recording media.

  In the present invention, the computer program or the digital signal may be transmitted via an electric communication line, a wireless or wired communication line, a network represented by the Internet, data broadcasting, or the like.

  The present invention may be a computer system including a microprocessor and a memory, wherein the memory stores the computer program, and the microprocessor operates according to the computer program.

  Further, by recording and transferring the program or the digital signal on the non-temporary recording medium, or transferring the program or the digital signal via the network or the like, another independent computer It may be implemented by the system.

  Furthermore, the above embodiment and the above modification examples may be combined.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  The voice quality conversion apparatus according to the present invention has a function of converting to a target voice quality while preserving the temporal change pattern of the utterance mode in the input voice, and is a user interface of information equipment and home appliances that require various voice qualities. In addition, it is useful in applications such as entertainment such as ringtones converted into personal voice quality. It can also be applied to voice changers in voice communications using mobile phones.

DESCRIPTION OF SYMBOLS 101 Vocal tract sound source separation part 102 Opening degree calculation part 103 Vocal tract information conversion part 104 Composition part 105 Opening degree conversion rate determination part 201 Vowel section extraction part 202 Vowel vocal tract information conversion part

Claims (11)

A vocal tract sound source separation unit that separates input speech into vocal tract information and sound source information;
An opening degree calculation unit that calculates an opening degree corresponding to a volume in the oral cavity at the time of utterance of the input sound from the vocal tract information separated by the vocal tract sound source separation part for each predetermined reproduction time of the input sound; ,
An opening degree conversion rate determining unit that determines a conversion rate of the opening degree for each predetermined reproduction time based on a utterance state conversion degree indicating a degree of opening degree conversion;
Based on the conversion rate determined by the opening degree conversion rate determination unit, the opening degree calculated by the opening degree calculation unit is used as the conversion rate of the vocal tract information separated by the vocal tract sound source separation unit. A vocal tract information conversion unit for converting into vocal tract information corresponding to the opening degree after being converted by
A voice quality conversion device comprising: a synthesis unit that generates a synthesized sound using the vocal tract information after conversion by the vocal tract information conversion unit and the sound source information separated by the vocal tract sound source separation unit. The opening degree calculation unit calculates a vocal tract cross-sectional area function from the vocal tract information separated by the vocal tract sound source separation unit at each predetermined reproduction time, and when the oral cavity is divided into a plurality of sections, The voice quality conversion apparatus according to claim 1, wherein the opening degree is calculated as a sum of vocal tract cross-sectional areas of each section indicated by the calculated vocal tract cross-sectional area function. Furthermore, a vowel section extracting unit that extracts a vowel section that is a time section of a vowel included in the input speech is provided,
The vocal tract information conversion unit converts only the vocal tract information in the vowel segment extracted by the vowel segment extraction unit from the vocal tract information separated by the vocal tract sound source separation unit. The voice quality conversion device described in 1. The voice quality conversion device according to any one of claims 1 to 3, wherein the opening degree conversion rate determination unit determines the utterance state conversion degree as the conversion rate for each predetermined reproduction time. The opening degree conversion rate determination unit calculates a product of a dynamic range in the temporal change sequence of the opening degree calculated by the opening degree calculation unit and the utterance state conversion degree at each predetermined reproduction time after conversion. The conversion rate to be multiplied by the opening degree calculated by the opening degree calculating unit is determined to calculate the opening degree after the conversion so as to coincide with the dynamic range in the temporal change sequence of the opening degree. Item 4. The voice quality conversion device according to any one of Items 1 to 3. The opening degree conversion rate determining unit calculates a product of an average value of the opening degree in the temporal change series of the opening degree calculated by the opening degree calculating unit and the utterance state conversion degree at each predetermined reproduction time. The opening degree calculated by the opening degree calculation unit to calculate the opening degree after the conversion so as to coincide with the average value of the opening degree in the temporal change series of the opening degree after the conversion. The voice quality conversion device according to claim 1, wherein a conversion rate is determined. The vocal tract information conversion unit converts a vocal tract cross-sectional area indicating the vocal tract information separated by the vocal tract sound source separation unit based on the conversion rate determined by the opening degree conversion rate determination unit. The voice quality conversion apparatus according to claim 1, wherein the vocal tract information separated by the vocal tract sound source separation unit is converted. The vocal tract information conversion unit, when dividing the oral cavity into a plurality of sections, multiplies the vocal tract cross-sectional area for each section by a conversion coefficient corresponding to the conversion rate and determined for each section. The voice quality conversion apparatus according to claim 7, wherein the vocal tract information separated by the vocal tract sound source separation unit is converted. The voice quality conversion apparatus according to claim 8, wherein the conversion coefficient has a larger absolute value of a difference from the conversion coefficient in an adjacent section as the conversion coefficient in a section closer to the lips. Separating input speech into vocal tract information and sound source information;
From the separated vocal tract information, calculating an opening degree corresponding to the volume in the oral cavity at the time of uttering the input sound for each predetermined reproduction time of the input sound;
Determining a conversion rate of the opening degree at each predetermined reproduction time based on a utterance state conversion degree indicating a conversion degree of the opening degree;
Converting the separated vocal tract information based on the determined conversion rate into vocal tract information corresponding to the open degree after the calculated open degree is converted at the conversion rate; and
A voice quality conversion method comprising: generating a synthesized sound using the converted vocal tract information and the separated sound source information.   A program for causing a computer to execute the voice quality conversion method according to claim 10.

Images