JP2006189544A - Interpretation system, interpretation method, recording medium with interpretation program recorded thereon, and interpretation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an interpretation system capable of outputting synthesized speech increasingly approximate to speaker's vocal quality, an interpretation method, a recording medium with an interpretation program recorded thereon, and an interpretation program. <P>SOLUTION: The interpretation system is equipped with a speech recognition section 101 which recognizes the input speech entered by a first language, an interpretation section 102 which interprets the result of the speech recognition to a second language, a speech synthesis section 103 which subjects the interpreted second language to speech synthesis, a vocal quality analysis section 104 which analyzes the a vocal quality of the first language, a vocal quality similarity metering section 105 which meters the similarity of the vocal quality of the first language, and vocal quality of the second language, and a vocal quality controller 106 which controls the vocal quality of the second language undergoing the speech synthesis in the speech synthesis section 103 based on the result of the vocal quality similarity metering obtained in the vocal quality similarity metering section 105, and thereby the vocal quality of the first language and the vocal quality of the second language are made relatively similar and the occurrence of sense of incongruity is reduced as far as possible. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an interpreting apparatus, an interpreting method, an interpreting method, a recording medium on which an interpreting program is recorded, and an interpreter that interprets an input speech signal uttered in a first language into a second language, synthesizes and outputs the interpretation result It is about the program.

  Conventionally, an interpreting apparatus automatically recognizes a speech signal input from a microphone by speech recognition means, translates the recognition result into a desired foreign language by automatic translation means, and then translates the translation result by speech synthesis means. This is realized by the composition of synthesizing with foreign language speech.

  By the way, in such an interpreting apparatus, it is also known that the speech rate of the synthesized speech is controlled according to the input speech rate for the purpose of obtaining a natural interpretation result close to the voice quality of the speaker. (For example, refer to Patent Document 1).

  FIG. 4 is a block diagram showing the configuration of a conventional interpreting apparatus configured to control the speech speed in this way.

  In FIG. 4, sound emitted toward the microphone 401 is converted into an electric signal by the microphone 401 and amplified by the input amplifier 402. The amplified input voice signal is recognized by the voice analysis circuit 403. The result of speech recognition is supplied to an electronic translation circuit 404, where it is automatically translated into a desired foreign language. Thereafter, the speech synthesis circuit 405 synthesizes speech and outputs it as a synthesized speech signal. The output synthesized voice signal is amplified by the output amplifier 406 and outputted from the speaker 407 as synthesized voice.

  On the other hand, the measurement circuit 408 measures the speech speed of the input voice, and the result is added to the control circuit 409. The control circuit 409 controls the electronic translation circuit 404 and the carry-out speed control circuit 410 in accordance with the speech speed supplied from the measurement circuit 408, and controls the speech speed of the synthesized speech to be linked to the speech speed of the input speech. .

  As a result, the synthesized speech output from the speaker 407 matches the speech speed of the input speech input to the microphone 401, and a natural speech speed interpretation result can be obtained.

  The character display device 411 displays the translation result as characters.

  Similarly, for the purpose of obtaining a natural interpretation result close to the voice quality of the speaker, there is also known one in which the prosody of the synthesized speech is controlled based on the input speech of the speaker (for example, a patent Reference 2).

According to this, when translating from the first language to another second language, accents are taken into consideration, and more natural interpretation results can be obtained in terms of accents, and speech understanding is improved. Can do.
JP-A-57-57375 JP-A-6-332494

  However, in the above-described conventional interpreting apparatus, the speech speed of the synthesized speech is controlled in accordance with the speech rate of the input speech, or the accent of the synthesized speech is adjusted to match the accent of the input speech. Since it is only a thing to control and it is only a thing paying attention to the speech speed and accent of a synthetic speech, it was not possible to obtain a proper synthetic speech suitable for the voice quality of the speaker.

  The present invention has been made in view of such a conventional problem, and provides an interpreting apparatus, an interpreting method, and an interpreting program capable of outputting synthesized speech closer to the voice quality of a speaker. .

  The interpreting apparatus according to the present invention includes a speech recognition unit that recognizes an input speech input in a first language, a translation unit that translates the speech recognition result into a second language, and a translated second language. Voice synthesis means for voice synthesis, voice quality analysis means for analyzing voice quality of the first language, voice quality similarity measurement means for measuring the similarity between the voice quality of the first language and the voice quality of the second language, And a voice quality control means for controlling the voice quality of the second language synthesized by the voice synthesis means based on the voice quality similarity measurement result obtained by the voice quality similarity measurement means.

  With this configuration, the similarity between the voice quality of the first language and the voice quality of the second language is measured by the voice quality similarity measurement unit, and the voice synthesis unit is controlled so that the similarity approaches, so the second The voice quality of the language is similar to the voice quality of the first language, and the uncomfortable feeling can be reduced as much as possible.

  In the interpreting apparatus of the present invention, the speech recognition means recognizes the input speech signal input in the first language as a character string, a word, a word string, a sentence, or a semantic expression, and the speech synthesis means It has a configuration in which it is synthesized as a character string or a word or a word string or a sentence according to language.

  With this configuration, the first language can be recognized as a character string, a word, a word string, a sentence, or a semantic expression, and the second language can be synthesized as a character string, a word, a word string, or a sentence.

  In the interpreting apparatus of the present invention, the voice quality analysis means extracts the voice quality feature quantity characterizing the individuality of the input speech, and the voice quality similarity measurement means extracts the voice quality feature quantity extracted by the voice quality analysis means. Is compared with the feature quantity of the voice quality of the second language synthesized by speech synthesis means.

  With this configuration, it is possible to easily determine the similarity of voice quality based on the voice quality feature amount that characterizes the individuality of the voice.

  The interpreting device of the present invention has a configuration in which the voice quality feature amount is a spectrum envelope as a vocal tract characteristic included in the input speech signal and the speech signal of the second language that is speech-synthesized.

  With this configuration, it is possible to easily extract the feature quantity of voice quality.

  In the interpreting apparatus of the present invention, the voice quality analyzing means includes vocal tract feature quantity extracting means for extracting the vocal tract feature quantity in the input voice signal, and pitch frequency extracting means for extracting the pitch frequency in the input voice signal. A voice quality similarity metric means for comparing the vocal tract feature quantity extracted by the vocal tract feature quantity extraction means with the vocal tract feature quantity of the second language synthesized by the speech synthesis means. And a pitch frequency similarity measuring unit that compares the pitch frequency extracted by the pitch frequency extracting unit with the pitch frequency of the second language synthesized by the speech synthesizing unit.

  With this configuration, similarity is determined not only for the vocal tract feature value but also for the pitch frequency, and it is possible to capture the individual features more greatly and make it easier to understand who has occurred.

  The interpreting apparatus of the present invention further includes signal power extraction means for extracting the signal power of the input speech signal, and the second language synthesized by the speech synthesis means based on the signal power extracted by the signal power extraction means. The signal power of the audio signal is controlled.

  With this configuration, the signal power of the second language can be controlled to be similar to the signal power of the first language, and can be arbitrarily adjusted to be large when the power is large and small when the power is small.

  The interpreting apparatus of the present invention further includes an utterance speed extraction means for extracting the utterance speed of the input voice from the input voice signal, and is synthesized by the voice synthesis means based on the utterance speed extracted by the utterance speed extraction means. It has the structure which controls the utterance speed of the audio | voice signal of 2 languages.

  With this configuration, the utterance speed of the second language can be increased or decreased in accordance with the utterance speed of the first language, and one more feature of a person who speaks the first language is added. be able to.

  Further, the interpretation method of the present invention includes a speech recognition step for recognizing input speech input in a first language, a translation step for translating the result of speech recognition in the speech recognition step into a second language, A speech synthesis step for speech synthesis of the second language translated in the step; a voice quality analysis step for analyzing the voice quality of the input speech input in the first language; and a voice quality of the input speech input in the first language. And a voice quality similarity metric step for measuring the voice quality similarity of the synthesized speech translated into the second language, and a voice quality similarity metric result obtained in the voice quality similarity metric step. And a voice quality control step for controlling the voice quality of the second language.

  According to this configuration, the voice quality of the first language and the voice quality of the second language are compared with each other in the voice quality similarity metric step, and the speech synthesis method is performed so that the similarity of the voice quality of both is closer according to the result. As a result, the second language can be brought closer to the voice of the first language.

  In the interpreting method of the present invention, the speech recognition step recognizes an input speech signal input in the first language as a character string, a word, a word string, a sentence, or a semantic expression, and the speech synthesis step includes a second speech synthesis step. It has a configuration in which it is synthesized as a character string or a word or a word string or a sentence according to language.

  With this configuration, the first language can be recognized as a character string, a word, a word string, a sentence, or a semantic expression, and the second language can be synthesized as a character string, a word, a word string, or a sentence.

Furthermore, the recording medium of the present invention includes a speech recognition step for recognizing input speech input in a first language, a translation step for translating the result of speech recognition in the speech recognition step into a second language, A speech synthesis step for speech synthesis of the second language translated in the step; a voice quality analysis step for analyzing the voice quality of the input speech input in the first language; and a voice quality of the input speech input in the first language. And a voice quality similarity metric step for measuring the voice quality similarity of the synthesized speech translated into the second language, and a voice quality similarity metric result obtained in the voice quality similarity metric step. And a voice quality control step for controlling the voice quality of the second language.

  With this configuration, it is possible to read out this interpreting program and execute each step, and easily translate a voice uttered in the first language into a second language having a voice quality similar to that.

  Furthermore, the interpreting program of the present invention includes a speech recognition procedure for recognizing an input speech input in a first language, a translation procedure for translating the result of speech recognition in the speech recognition procedure into a second language, A speech synthesis procedure for speech synthesis of the second language translated in the procedure, a voice quality analysis procedure for analyzing the voice quality of the input speech input in the first language, and a voice quality of the input speech input in the first language And a voice quality similarity metric procedure for measuring the voice quality similarity of the synthesized speech translated into the second language, and a voice quality similarity metric result obtained by the voice quality similarity metric procedure. And a voice quality control procedure for controlling the voice quality of the second language.

  With this configuration, it is possible to cause the computer to execute the translation of the voice uttered in the first language into the second language having a voice quality similar to that.

  The interpreting apparatus according to the present invention includes a speech recognition unit that recognizes an input speech input in a first language, a translation unit that translates the speech recognition result into a second language, and a translated second language. Voice synthesis means for voice synthesis, voice quality analysis means for analyzing voice quality of the first language, voice quality similarity measurement means for measuring the similarity between the voice quality of the first language and the voice quality of the second language, Voice quality control means for controlling the voice quality of the second language synthesized by the voice synthesis means based on the voice quality similarity metric result obtained by the voice quality similarity measurement means. The voice quality of the second language is similar to the voice quality of the first language because the voice quality and the voice quality of the second language are measured by the voice quality similarity measurement means, and the speech synthesis means is controlled so that the similarity is approximated. However, it has the effect of reducing discomfort. That.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a schematic configuration of an interpreting apparatus according to Embodiment 1 of the present invention.

  As shown in FIG. 1, the interpretation apparatus according to Embodiment 1 of the present invention includes a speech recognition unit 101 that inputs an input speech signal, a translation unit 102 that receives the output of the speech recognition unit 101, and a translation unit 102. The speech synthesis unit 103 that receives the output, the voice quality analysis unit 104 that receives the input speech signal, the output of the voice quality analysis unit 104 and the speech synthesis result of the speech synthesis unit 103 are input, and the speech is generated based on these results. And a voice quality control unit 106 that controls the synthesis unit 103.

  Next, the operation of the interpreting apparatus in this embodiment will be described.

A voice uttered in a first language (for example, Japanese) is converted into a voice signal by a microphone (not shown), amplified by an input amplifier, and input to the voice recognition unit 101 as an input voice signal. When an input voice signal is input to the voice recognition unit 101, the voice recognition unit 101 recognizes the input voice signal and outputs the result in a predesignated format such as a word, a word string, a sentence, or a semantic expression. To do. The output recognition result is input to the translation unit 102, where it is translated into a second language (for example, English). Similar to the recognition result, the translation result is output in an expression format designated in advance such as a word, a word string, and a sentence. The output translation result is added to the speech synthesizer 103, where it is synthesized and output a synthesized speech signal.
The voice synthesizer 103 is a voice synthesizer capable of voice quality control, and is configured to be able to output synthesized voice signals having different voice qualities under the control of the voice quality controller 106. Specifically, for example, a method based on a vocoder which is a well-known technique (Sadago Furui “Speech Information Processing”, Morikita Publishing 1998 P40) is used, but is not limited to this method.

  Hereinafter, this will be described in more detail.

  The input voice signal is input to the voice recognition unit 101 and also to the voice quality analysis unit 104. The voice analysis unit 104 extracts, for example, a feature quantity of voice quality that characterizes the individuality of voice, such as a spectral envelope (voice tract characteristic) included in the input voice signal. As this feature amount, for example, a vector amount such as a low-order cepstrum coefficient is used.

  When the voice analysis unit 104 extracts a voice quality feature amount from the input voice signal, the feature amount is input to the voice quality similarity measurement unit 105. When the voice quality feature quantity of the input voice signal is input to the voice quality similarity measurement unit 105, the synthesis is performed by the voice synthesis unit 103 and output under the control of the voice quality similarity measurement unit 105 and the voice quality control unit 105. The voice synthesis unit 103 is controlled so that the voice quality of the voice signal is closest to the voice quality of the input voice signal.

  That is, the voice quality similarity measurement unit 105 compares the voice quality feature quantity of the input voice signal with the voice quality feature quantity of the voice synthesized by the voice synthesis unit 103 when the current control condition is assumed, Measure the voice quality similarity between voice and output voice. As a method for measuring the similarity of voice quality, for example, a Euclidean distance of a vector quantity such as a low-order cepstrum coefficient or an auditory weighting distance is used.

  When the voice quality similarity is measured in this manner, the result is input to the voice quality control unit 106. The voice quality control unit 106 controls the voice synthesis unit 103 so that the voice quality similarity becomes an optimum value based on the input voice quality similarity.

  When distance is used as the similarity of voice quality, it means that the smaller the distance, the better the control standard (similarly).

  According to this configuration, the voice quality similarity measuring unit 105 controls the voice synthesis unit 103 using the voice quality control unit 106 while comparing the voice quality of the input voice signal and the voice quality of the synthesized voice signal, and inputs the voice quality of the synthesized voice. It can be made closer to the voice quality.

(Embodiment 2)
FIG. 2 is a block diagram showing a schematic configuration of the interpreting apparatus according to Embodiment 2 of the present invention.

  As shown in FIG. 2, the interpreting apparatus according to Embodiment 2 of the present invention includes a speech recognition unit 201 that receives an input speech signal, a translation unit 202 that receives an output of the speech recognition unit 201, and a translation unit 202. A speech synthesizer 203 that receives the output of the input, a vocal tract feature amount extractor 204 that receives the input speech signal, a pitch frequency extractor 205 that also receives the input speech signal, and an input speech signal as input. A signal power extraction unit 206 that performs input, an audio speed extraction unit 207 that receives an input audio signal, a vocal tract feature quantity similarity measurement unit 208 that receives an output of the vocal tract extraction unit 204, and a pitch frequency extraction unit 205. , A pitch frequency similarity metric unit 209, a vocal tract feature quantity similarity metric unit 208, a pitch frequency similarity metric unit 209, a signal power extraction unit 206, and a voice speed extraction unit 20. As input the output of each, and a voice control unit 210 for controlling the speech synthesizing portion 203.

  Next, the operation of the interpreting apparatus in this embodiment will be described.

  The voice uttered in the first language is converted into a voice signal by a microphone (not shown), amplified by an input amplifier, and input to the voice recognition unit 201 as an input voice signal. When an input speech signal is input to the speech recognition unit 201, the speech recognition unit 201 recognizes the input speech signal that has been input and outputs the result in a predesignated format such as a word, word string, sentence, or semantic expression. To do. The output recognition result is input to the translation unit 202, where it is translated into the second language. Similar to the recognition result, the translation result is output in a pre-designated expression format such as a word, a word string, or a sentence. The output translation result is added to the speech synthesizer 203, where the speech is synthesized and a synthesized speech signal is output.

  The voice synthesizer 203 is a voice synthesizer capable of voice quality control, and is configured to output synthesized voice signals having different voice qualities under the control of the voice quality controller 210. The voice quality control unit 210 is configured to receive the outputs of the vocal tract feature quantity similarity measurement unit 208, the pitch frequency similarity measurement unit 209, the signal power extraction unit 206, and the utterance speed extraction unit 207. . Therefore, the voice quality control unit 210 is controlled by these outputs, and controls the speech synthesis unit 203 according to the outputs.

  This will be described in more detail below.

  First, the vocal tract feature value extraction unit 204 extracts, for example, a feature value of voice quality that characterizes the individuality of speech, such as a spectrum envelope (voice tract characteristic) included in the input speech signal. As this feature amount, for example, a vector amount such as a low-order cepstrum coefficient is used. When the feature quantity of the voice quality is extracted, the feature quantity is supplied to the vocal tract feature quantity similarity measurement unit 208. Then, it is compared with the voice quality feature amount of the synthesized speech synthesized by the speech synthesis unit 203.

  The pitch frequency extraction unit 205 extracts the pitch frequency of the input audio signal or the transition pattern thereof. The extracted pitch frequency or its transition pattern is input to the pitch frequency similarity metric unit 209, and the pitch frequency similarity metric unit 209 synthesizes the pitch frequency of the synthesized speech synthesized by the speech synthesizer 203, or its Compared to transition pattern.

  The results obtained by comparison by the vocal tract feature quantity similarity measurement unit 208, the results obtained by comparison by the pitch frequency similarity measurement unit 209, and the signal power extraction unit 206 and the utterance speed extraction unit 207, respectively. The input voice signal power and utterance speed are input to the voice quality control unit 210, respectively. As a result, the speech synthesizer 203 is controlled as follows.

  That is, first, the speech synthesis unit 203 is configured such that the voice quality of the synthesized sound becomes a voice quality close to that of the input voice by operating the vocal tract feature quantity similarity measurement unit 208, the pitch frequency similarity measurement unit 209, and the voice quality control unit 210, respectively. Is controlled.

  The vocal tract feature quantity similarity measurement unit 208 compares the voice quality feature quantity of the input voice signal with the voice quality feature quantity of the voice synthesized by the voice synthesis unit 203 when the current control condition is assumed. Measure the voice quality similarity between voice and output voice. As a method for measuring the similarity of voice quality, the Euclidean distance or the auditory weighting distance of the vector amount may be used. The similarity of the measured voice quality is added to the voice quality control unit 210, and the voice synthesis unit 203 is controlled based on the similarity. Therefore, the voice quality of the synthesized sound synthesized by the voice synthesizer 203 is similar to the voice quality of the input voice. When distance is used as the similarity, the smaller the distance value, the better the control standard (similarly).

The pitch frequency similarity measuring unit 209 measures the similarity between the pitch frequency pattern of the input voice signal and the pitch frequency pattern of the voice synthesized by the voice synthesizing unit 203 when the current control condition is assumed. The similarity of the pitch frequency pattern is also supplied to the voice quality control unit 210, and the voice synthesis unit 203 is controlled based on the similarity. Therefore, the pitch frequency pattern of the synthesized sound synthesized by the speech synthesis unit 203 is a pitch frequency pattern close to the input voice.

  In addition, when controlling the voice quality, the power of the input voice signal and the utterance speed of the input voice are also referred to. The power of the synthesized voice and the utterance speed are also linked to the input voice and the utterance speed. become.

  As described above, according to this configuration, the voice quality feature extraction unit 204, the pitch frequency extraction unit 205, the signal power extraction unit 206, and the utterance speed extraction unit 207 respectively extract voice quality features, and the vocal tract feature amount similarity metric. The vocal tract feature amount similarity and the pitch frequency similarity are calculated using the unit 208 and the pitch frequency similarity measuring unit 209, and are input to the voice quality control unit 210. The vocal tract feature value, pitch frequency, signal power, and utterance speed can be made closer to the input voice, and a synthesized voice with a voice quality closer to the input voice can be obtained. .

(Embodiment 3)
FIG. 3 is a flowchart for explaining an interpretation method as the third embodiment of the present invention.

  In FIG. 3, step S301 is a voice input step for inputting a voice signal uttered in the first language. The input speech signal of the first language is speech-recognized in the next speech recognition step S302, and the result is output in a predesignated format such as a word, a word string, a sentence, or a semantic expression. In the translation step S303, the recognition result is translated into the second language, and the translation result is output in a previously designated expression format such as a word, a word string, or a sentence. In speech synthesis step S304, the translation result is input and a synthesized speech signal in the second language is output. In the voice quality analysis step S305, for example, a voice quality feature amount characterizing voice personality such as a spectrum envelope (voice tract characteristic) included in the input voice is extracted. As this feature amount, for example, a vector amount such as a low-order cepstrum coefficient may be used.

  The voice quality similarity measurement step S306 and the next voice quality control step S307 perform processing in conjunction with each other, thereby controlling the voice quality of the synthesized sound so that the voice quality of the synthesized sound is close to that of the input voice.

  That is, in the voice quality similarity metric step S306, by comparing the voice quality feature quantity of the input voice signal with the voice quality feature quantity of the voice synthesized by the voice synthesis step S304 when the current control condition is assumed, Measure the voice quality similarity between input and output speech. As a method for measuring the similarity of voice quality, a vector quantity of Euclidean distance or auditory weighting distance may be used. When distance is used as the similarity, the smaller the distance value, the better the control standard (similarly). In the voice quality control step S307, the synthesized voice is controlled so that the result of the voice quality similarity metric becomes an optimum value.

  According to this method, it is possible to control the voice quality of the synthesized voice while comparing the voice quality of the input voice and the voice quality of the synthesized voice using the voice quality similarity measurement step S306 and the voice quality control step S307. Synthetic voices with similar voice quality can be output.

  In this embodiment, when an interpreting program including these steps is recorded on a recording medium, the recording medium is mounted on a computer or the like, and the program including these steps is read using the computer. The speech uttered in the first language can be translated into the second language and synthesized as speech.

  In this embodiment, when an interpreter program including these steps is distributed or moved to a computer or the like via a communication medium such as the Internet, the distributed or moved computer executes the program as it is. The voice uttered in the first language can be arbitrarily translated into the second language and synthesized as a voice.

  According to the present embodiment, the voice quality of the translated second language can be brought close to the voice quality of the first language. Is less likely to cause a sense of incongruity, such as being translated by a woman's voice, but enables translation with less sense of incongruity.

  In each of the above embodiments, the first language means the language to be translated, and the second language means the translated language. That is, if Japanese is translated into English as in the above embodiment, Japanese is the first language and English is the second language. If the Japanese dialect is translated into the standard language even in the same Japanese language, the Osaka dialect is the first language and the standard language is the second language. That is, the first and second languages include not only so-called national languages but also dialects, local languages, and all other languages with different expressions.

  The interpreting apparatus according to the present invention includes a speech recognition unit that recognizes an input speech input in a first language, a translation unit that translates the speech recognition result into a second language, and a translated second language. Voice synthesis means for voice synthesis, voice quality analysis means for analyzing voice quality of the first language, voice quality similarity measurement means for measuring the similarity between the voice quality of the first language and the voice quality of the second language, Voice quality control means for controlling the voice quality of the second language synthesized by the voice synthesis means based on the voice quality similarity metric result obtained by the voice quality similarity measurement means. The voice quality of the second language is similar to the voice quality of the first language because the voice quality and the voice quality of the second language are measured by the voice quality similarity measurement means, and the speech synthesis means is controlled so that the similarity is approximated. And speech synthesis that minimizes the sense of discomfort It is useful in various types of equipment to perform.

The block diagram which shows the structure of the interpretation apparatus in Embodiment 1 of this invention. The block diagram which shows the structure of the interpretation apparatus in Embodiment 2 of this invention. The flowchart for demonstrating the interpretation method in Embodiment 3 of this invention. Block diagram showing the configuration of a conventional interpreter

Explanation of symbols

101, 201 Speech recognition unit 102, 202 Translation unit 103, 203 Speech synthesis unit 104 Voice quality analysis unit 105 Voice quality similarity measurement unit 106, 210 Voice quality control unit 204 Vocal tract feature value extraction unit 205 Pitch frequency extraction unit 206 Signal power extraction unit 207 Speech rate extraction unit 208 Vocal tract feature value similarity measurement unit 209 Pitch frequency similarity measurement unit

Claims (11)

Speech recognition means for recognizing input speech input in a first language, translation means for translating the speech recognition result into a second language, and speech for speech synthesis of the translated second language Synthesis means, voice quality analysis means for analyzing voice quality of the first language, voice quality similarity measurement means for measuring the similarity between the voice quality of the first language and the voice quality of the second language, and the voice quality An interpreting apparatus comprising: voice quality control means for controlling the voice quality of the second language synthesized by the speech synthesis means based on the voice quality similarity metric obtained by the similarity metric means. The speech recognition means recognizes an input speech signal input in the first language as a character string, a word, a word string, a sentence, or a semantic expression, and the speech synthesis means recognizes a character string in the second language. 2. The interpreter according to claim 1, wherein the interpreter is synthesized as a word, a word string, or a sentence. The voice quality analysis unit extracts a voice quality feature amount characterizing the individuality of the input speech, and the voice quality similarity measurement unit extracts the voice quality feature amount extracted by the voice quality analysis unit as the voice synthesis unit. The interpreting apparatus according to claim 1, wherein the interpreting apparatus compares with the feature quantity of the voice quality of the second language synthesized by the voice. 4. The interpreting apparatus according to claim 3, wherein the feature quantity of the voice quality is a spectrum envelope as a vocal tract characteristic included in the input speech signal and the speech signal of the second language synthesized by speech. . The voice quality analysis means comprises: a vocal tract feature quantity extraction means for extracting a vocal tract feature quantity in the input voice signal; and a pitch frequency extraction means for extracting a pitch frequency in the input voice signal, the voice quality similarity Vocal tract feature quantity similarity metric for comparing the vocal tract feature quantity extracted by the vocal tract feature quantity extraction means with the vocal tract feature quantity of the second language synthesized by the speech synthesis means. And a pitch frequency similarity measuring means for comparing the pitch frequency extracted by the pitch frequency extracting means with the pitch frequency of the second language synthesized by the speech synthesizing means. The interpreting device according to any one of claims 1 to 4. Signal power extraction means for extracting the signal power of the input voice signal is further provided, and the second language voice signal synthesized by the voice synthesis means based on the signal power extracted by the signal power extraction means. 6. The interpreting apparatus according to claim 1, wherein the signal power is controlled. A voice rate extraction unit for extracting a voice rate of the input voice from the input voice signal, and the second language synthesized by the voice synthesis unit based on the voice rate extracted by the voice rate extraction unit; 6. The interpreting apparatus according to claim 1, wherein a speech rate of the voice signal is controlled. A speech recognition step for recognizing input speech input in a first language, a translation step for translating the speech recognition result in the speech recognition step into a second language, and the first translated in the translation step. A speech synthesis step for speech synthesis of two languages; a voice quality analysis step for analyzing voice quality of the input speech input in the first language; a voice quality of the input speech input in the first language; and the second A voice quality similarity metric step for measuring the voice quality similarity of the synthesized speech translated into the language, and a voice quality similarity metric result obtained in the voice quality similarity metric step, and synthesized in the voice synthesis step And a voice quality control step for controlling the voice quality of the second language. The speech recognition step recognizes an input speech signal input in the first language as a character string, a word, a word string, a sentence, or a semantic expression, and the speech synthesis step includes a character string in the second language, 9. The interpretation method according to claim 8, wherein the interpretation is performed as a word, a word string, or a sentence. A speech recognition step for recognizing input speech input in a first language, a translation step for translating the speech recognition result in the speech recognition step into a second language, and the first translated in the translation step. A speech synthesis step for speech synthesis of two languages; a voice quality analysis step for analyzing voice quality of the input speech input in the first language; a voice quality of the input speech input in the first language; and the second A voice quality similarity metric step for measuring the voice quality similarity of the synthesized speech translated into the language, and a voice quality similarity metric result obtained in the voice quality similarity metric step, and synthesized in the voice synthesis step A computer-readable recording medium recording an interpreter program for causing a computer to execute a voice quality control step for controlling the voice quality of the second language. A speech recognition procedure for recognizing input speech input in a first language; a translation procedure for translating a speech recognition result in the speech recognition procedure into a second language; and the first translated in the translation procedure. A speech synthesis procedure for speech synthesis of two languages, a voice quality analysis procedure for analyzing voice quality of input speech input in the first language, a voice quality of input speech input in the first language, and the second A voice quality similarity metric procedure for measuring the similarity of voice quality of synthesized speech translated into the language of the voice and a voice quality similarity metric result obtained by the voice quality similarity metric procedure are synthesized by the voice synthesis procedure. An interpreting program for causing a computer to execute a voice quality control procedure for controlling the voice quality of the second language.

Images