JP4127274B2 - Telephone speech recognition system - Google Patents

Description

  The present invention relates to a telephone voice recognition system, and more particularly to a telephone voice recognition system that recognizes telephone voice.

  2. Description of the Related Art Conventionally, a telephone voice recognition system in which a voice recognition device is applied to telephone voice is known (for example, see Patent Document 1). This conventional telephone speech recognition system is equipped with a CPU board on which various memories are mounted, an A / D conversion board for converting analog signals into digital signals, a network termination device and a digital switch, and a telephone line connection terminal. A telephone answering board, an electrical connection member that electrically connects these boards, and a voice synthesizer that generates a signal output from the CPU board as a synthesized voice signal and supplies the synthesized voice signal to the telephone answering board. Connected to multiple sets of networks as a client, and the client automatically transmits the recognition results to the server and receives commands from the server via the network, and the client automatically performs the corresponding processing. This is what we do.

  The conventional telephone speech recognition system having such a configuration operates as follows. The analog signal sent from the telephone line connected to the telephone line connection terminal of the telephone response board is transmitted from the telephone response board to the A / D conversion board and converted into a digital signal by the A / D conversion board. The signal is transmitted to the CPU board, recognition processing is executed, and the recognition result is converted into synthesized speech by the speech synthesizer, and then transmitted from the telephone response board to the telephone line.

JP 2000-22830 A

  However, since this conventional telephone speech recognition system does not consider the quality and bandwidth of the communication line, the analog signal sent from the telephone line operates on the assumption that there is no distortion, and the distortion In other words, synthetic speech processing is performed on a certain analog signal. For this reason, the above conventional telephone speech recognition system has a problem that the voice is interrupted and the conversation cannot be smoothly performed when the quality of the communication line is poor or the bandwidth is insufficient.

  The present invention has been made in view of the above points, and by performing telephone voice supplementation by telephone voice recognition and voice synthesis, a conversation without interruption even when the quality of the communication line is poor or the bandwidth is insufficient. An object of the present invention is to provide a telephone speech recognition system that can be used.

In order to achieve the above object, the present invention receives a speech recognition signal obtained by speech recognition of a transmission voice signal from a transmission side telephone terminal at a reception side telephone terminal and synthesizes the voice to generate a transmission voice signal. In the telephone speech recognition system to be restored, the transmitting telephone terminal transmits the character information, which is a digital signal for each single sound of the transmitted voice obtained by voice recognition of the transmitted voice signal, and the analog voice for each single sound of the transmitted voice signal. A means for generating and transmitting integrated data comprising voice information as a signal, and the receiving side telephone terminal receives the integrated data and separates it into character information and voice information, and the amount of distortion of the separated voice information Separating / selecting means for selecting a speech synthesis signal obtained by speech synthesis of character information when the threshold is equal to or greater than a set threshold value, and selecting separated speech information when the amount of distortion is less than the threshold value; , Selected speech synthesis signal Or characterized in that it has a Could sound output unit audio information as the transmitted voice. In the present invention, even if the voice information is distorted, such as when the quality of the communication line is poor or the band is insufficient, the voice information can be supplemented using the voice signal that has been synthesized.

In order to achieve the above-mentioned object, the present invention receives a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission side telephone terminal by the reception side telephone terminal and synthesizes the voice to transmit voice. in telephone speech recognition system to restore the signal, the transmission side telephone terminal includes a voice input means for capturing a transmitted voice as the transmission voice signal, a digital signal for each monophonic transmitted voice from the transmitted voice signal by the voice recognition processing Integrated data obtained by integrating a voice recognition means for obtaining certain character information and an analog voice signal for each single sound of the transmitted voice signal output from the voice input means into voice information, and integrating them together with the character information output from the voice recognition means in a predetermined unit. And an integration means for generating the integrated data, and a transmission means for transmitting the integrated data to the other telephone terminal
The receiving side telephone terminal includes a receiving unit that receives the integrated data, a separating unit that separates the integrated data received by the receiving unit into voice information and character information, and a distortion analysis that measures a distortion amount of the separated voice information. Means, speech synthesis means for performing speech synthesis processing on the separated character information and outputting a speech synthesis signal, and speech when the distortion amount measured by the distortion analysis means is greater than or equal to a set threshold value. When a synthesized signal is selected and the amount of distortion is less than a threshold value, a complementing unit that selects speech information whose distortion amount has been measured, and a voice synthesis signal or voice information output from the complementing unit is pronounced as speech And an audio output means.

  In the present invention, even if the voice information is distorted, such as when the quality of the communication line is poor or the band is insufficient, the voice information can be supplemented using the voice signal that has been synthesized.

In order to achieve the above object, the present invention receives a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission side telephone terminal by the reception side telephone terminal and synthesizes the voice to transmit the transmission voice signal. In the telephone speech recognition system that restores the speech, the transmitting telephone terminal transmits the character information, which is a digital signal for each single sound of the transmitted voice obtained by voice recognition of the transmitted voice signal, and the single voice of the transmitted voice signal . audio information is an analog audio signal, comprising means for generating and transmitting to the integrated data consisting of a digital signal der Ru voice quality information indicating the voice quality of speech waveform talker or preferences, recipient telephone terminal When the integrated data is received and separated into character information, voice information and voice quality information, and the distortion amount of the separated voice information is equal to or greater than a set threshold, the character information is approximated to the voice quality indicated by the voice quality information. Speech synthesis to obtain speech synthesis sound When the synthesized speech signal is selected and the amount of distortion is less than the threshold, separation / selection means for selecting the separated speech information, and the selected speech synthesis signal or speech information as the transmitted speech And an audio output means.

  In the present invention, since the integrated data composed of the character information of the transmitted voice obtained by voice recognition of the transmitted voice signal, the voice information of the transmitted voice signal, and the voice quality information of the received voice is transmitted and received, the communication line Even if the voice information is distorted, such as when the quality of the voice is poor or the bandwidth is insufficient, the voice information can be supplemented using the voice signal synthesized by voice synthesis, and the voice quality based on the voice quality information is further improved. Received voice can be obtained.

In order to achieve the above-mentioned object, the present invention receives a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission side telephone terminal by the reception side telephone terminal and synthesizes the voice to transmit voice. In a telephone speech recognition system that restores signals,
The transmitting-side telephone terminal includes a voice input unit that captures a transmission voice as a transmission voice signal, a voice recognition unit that obtains character information that is a digital signal for each single sound of the transmission voice from the transmission voice signal by voice recognition processing, A voice signal that is an analog voice signal for each single sound of a transmission voice signal output from the voice input means, a character information output from the voice recognition means, and a digital signal indicating a voice waveform of the sender or the desired voice quality comprising an integrated means for generating an integrated data obtained by integrating the certain voice quality information in a predetermined unit, and transmitting means for transmitting the integrated data to the destination telephone terminal,
The telephone terminal on the receiving side receives the integrated data, the separating means for separating the integrated data received by the receiving means into voice information, character information, and voice quality information, and measures the distortion amount of the separated voice information Measured by the distortion analyzing means, the speech synthesizing means for outputting the synthesized speech signal of the synthesized speech that approximates the character information to the voice quality indicated by the voice quality information, and the distortion analyzing means based on the separated character information and voice quality information A complementary means for selecting a speech synthesis signal when the amount of distortion set is equal to or greater than a set threshold value, and a complementary means for selecting speech information whose distortion amount is measured when the distortion amount is less than the threshold value; And a voice output means for generating the voice synthesis signal or voice information output from the voice as the voice.

  In the present invention, since the integrated data composed of the character information of the transmitted voice obtained by voice recognition of the transmitted voice signal, the voice information of the transmitted voice signal, and the voice quality information of the received voice is transmitted and received, the communication line Even if the voice information is distorted, such as when the quality of the voice is poor or the bandwidth is insufficient, the voice information can be supplemented using the voice signal synthesized by voice synthesis, and the voice quality based on the voice quality information is further improved. Received voice can be obtained.

  According to the present invention, even if the voice information is distorted, such as when the quality of the communication line is poor or the bandwidth is insufficient, the quality of the communication line is poor because the voice signal is supplemented with the synthesized voice signal. Even if the bandwidth is insufficient, you can have a seamless conversation.

  Further, according to the present invention, by transmitting and receiving integrated data composed of the character information of the transmitted voice obtained by voice recognition of the transmitted voice signal, the voice information of the transmitted voice signal, and the voice quality information of the received voice. When the speech information is distorted, the speech information is supplemented using the speech signal that has been synthesized, and the received speech having the voice quality based on the voice quality information is obtained. I can hear you.

  Next, the configuration of the embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 shows a system configuration diagram of an embodiment of a telephone speech recognition system according to the present invention. In the figure, the telephone speech recognition system of the present embodiment includes a telephone 1 that transmits voice, a telephone 5 that receives voice, a public network 3 that is a communication line, a telephone 1 and a telephone 5. It is composed of an exchange 2 and an exchange 4 for connecting to the public network 3.

  The telephone 1 on the voice transmission side includes a voice input unit 11, a voice recognition unit 12, a voice / character information integration unit 13, and an integrated data transmission unit 14. On the other hand, the telephone 5 on the voice receiving side includes an integrated data receiving unit 51, a voice / character information separating unit 52, a voice synthesizing unit 53, a voice distortion analyzing unit 54, a voice complementing unit 55, and a voice output unit 56. including.

  The integrated data 6 integrated by the voice / character information integration means 13 is composed of voice information 61 and character information 62 as shown in FIG. The integrated data 6 ′ received by the integrated data receiving means 51 via the public network 3 and the exchange 4 is composed of voice information 61 ′ and character information 62 as shown in FIG. .

  The voice information 61 and 61 'are, for example, single-tone analog voice signals, and the character information 62 is a digital voice signal obtained from the voice recognition result of the single sound. Since the analog voice signal may be distorted due to the quality or bandwidth of the communication line at the time of transmission, the voice information 61 on the transmission side is represented as voice information 61 'on the reception side. On the other hand, the character information 62 is a digital voice signal, and since the receiving side can normally correct errors during digital signal processing, the receiving side is represented as being able to restore the same character information 62 as the transmitting side. ing.

  Each of these means generally operates as follows. The voice input unit 11 is a unit for taking a voice from the outside into the telephone 1. The voice recognition unit 12 is a unit for converting external voice taken in by the voice input unit 11 into character information 62. The voice / character information integration unit 13 integrates the voice information 61 which is the voice itself from the outside and the character information 62 converted by the voice recognition unit 12, which is taken in by the voice input unit 11, into one piece of integrated data 6. It is means for making. The integrated data transmission unit 14 is a unit for transmitting the integrated data 6 to the exchange 2.

  On the other hand, in the telephone 5 on the voice receiving side, the integrated data receiving means 51 is means for receiving the integrated data 6 ′ from the exchange 4. The voice / character information separating unit 52 is a unit for separating the integrated data 6 ′ into the voice information 61 ′ and the character information 62. The voice synthesizing unit 53 is a unit for converting the character information 62 separated from the integrated data 6 ′ by the voice / character information separating unit 52 into synthesized speech. The voice distortion analyzing means 54 is a means for measuring the distortion of the voice information 61 ′ separated from the integrated data 6 ′ by the voice / character information separating means 52 and outputting the voice information 61 ′. The voice complementing means 55 is a means for replacing the voice information 61 ′ whose distortion has been detected by the voice distortion analyzing means 54 with the synthesized voice converted by the voice synthesizing means 53. The audio output unit 56 is a unit that performs electro-acoustic conversion on the input audio signal and outputs it as audio (sound wave).

  Next, the operation of the embodiment of the present invention will be described in detail with reference to FIG. 1 and FIG. The voice on the transmitting side (transmitted voice) is acoustic-electrically converted by the voice input means 11 of the telephone set 1 and is taken in as a transmitted voice signal. The captured transmission voice signal is an analog voice signal, which is recognized by the voice recognition means 12 for each single sound by a conventionally known method, and is converted into character information which is a digital signal indicating the transmission voice as a character string. The

  The voice / character information integration unit 13 includes voice information indicating the transmitted voice signal output from the voice input unit 11 (indicated by 61 in FIG. 2A) and character information output from the voice recognition unit 12 (see FIG. 2 (A) is indicated by 62), and integrated data 6 in the format shown in FIG. The integrated data transmission unit 14 transmits the integrated data 6 output from the voice / text information integration unit 13 to the public network 3 via the exchange 2 to the telephone 5 on the receiving side.

  The exchange 2 transmits the received integrated data 6 to the telephone 5 on the receiving side via the public network 3 and the exchange 4. The integrated data receiving means 51 of the telephone 5 receives the integrated data transmitted from the exchange 4, and obtains the received integrated data 6 'having the format shown in FIG. The voice / character information separating means 52 separates the voice information 61 ′ and the character information 62 shown in FIG. 2B constituting the reception integrated data 6 ′ received by the integrated data receiving means 51 into the character information 62. Is output to the voice synthesizing means 53, and the voice information 61 ′ is output to the voice distortion analyzing means 54.

  The voice synthesizer 53 performs a known voice synthesis process on the input character information 62 to create a synthesized analog voice signal. On the other hand, the audio distortion analysis means 54 measures the distortion amount of the input audio information 61 '. This amount of distortion is measured based on, for example, the level of audio information 61 ′ that is an analog audio signal. The speech complementing means 55 replaces the distorted speech information 61 ′ with the synthesized speech signal from the speech synthesizer 53 when the measured distortion amount of the speech information 61 ′ is distorted more than a preset threshold value. If the measured distortion amount is less than a preset threshold value and can be regarded as not distorted, the received audio information 61 ′ input via the audio distortion analysis unit 54 is output as audio. It outputs to the means 56.

  The voice output unit 56 performs electro-acoustic conversion on the voice information 61 ′, which is an analog voice signal input from the voice complement unit 55, or the voice synthesis signal, which is an analog voice signal, and generates the voice as an incoming voice of the telephone 5. . Thus, according to the present embodiment, when the voice information 61 ′, which is an analog voice signal in the telephone voice signal (integrated data 6 ′) received by the telephone 5 on the reception side, is not distorted, the transmission side The synthesized speech synthesized by the speech synthesizing means 53 is produced only when the personal transmission voice of the telephone 1 is produced and the voice information 61 'is distorted, so that the quality of the communication line is poor or the band is low. Even when there is a shortage, it is possible to have an uninterrupted conversation and to prioritize the pronunciation of the individual transmitted speech of the telephone 1 on the sending side, so as much as possible to the synthesized speech by speech synthesis without taste. Can be pronounced with the voice of the speaker.

  Next, the operation of the present embodiment will be described in more detail with reference to FIG. 3 using a specific example. Assume that the voices input to the voice input means 11 are “A”, “I”, “U” as shown in FIG. 3, for example (step S1 in FIG. 3). The transmitted voices “A”, “I”, “U” are acoustic-electrically converted by the voice input means 11 of the telephone set 1 and taken in as voice information 61 of an analog voice signal.

  The captured voice information 61 (“A”, “I”, “U”) is converted into character information 62 (“A”, “I”, which is a digital signal for each tone, by a known voice recognition process by the voice recognition unit 12. "," C ") (step S2 in FIG. 3). The voice / character information integration unit 13 includes voice information 61 (“A”, “I”, “U”) captured by the voice input unit 11 and character information 62 obtained by voice recognition by the voice recognition unit 12. ("A", "I", "U") and the integrated data 6 ("A" and "A", "I" and "I", "U" in the format shown in FIG. "And" U ") are created (step S3 in FIG. 3).

  The integrated data transmission means 14 transmits the integrated data 6 (“A” and “A”, “I” and “I”, “U” and “U”) to the exchange 2 to the telephone 5 of the transmission partner. The exchange 2 transmits the received integrated data 6 ′ (“A” and “A”, “?” And “I”, “U” and “C”) to the exchange 4 via the public network 3. At this time, due to the poor quality of the communication line or the lack of bandwidth, the integrated data 6 is distorted when the voice information “I” is transmitted to the exchange 2, and the voice information “I” is transmitted to the exchange 4. It is assumed that “U” is distorted. Thereby, the integrated data receiving means 51 of the telephone 5 receives the integrated data 6 ′ (“a” and “a”, “?” And “b”, “?” And “c”) (step of FIG. 3). S4).

  The reception integrated data 6 ′ (“a” and “a”, “?” And “b”, “?” And “c”) is converted into voice information 61 ′ (“a”) by the voice / character information separation means 52. , “?”, “?”) And character information 62 (“A”, “I”, “U”). Character information 62 (“A”, “I”, “U”), which is a separated digital signal, is supplied to the voice synthesis means 53 and converted into an analog voice signal synthesized by a known voice synthesis process. (Step S5 in FIG. 3).

  On the other hand, the audio information 61 ′ (“A”, “?”, “?”), Which is a separated analog audio signal, is supplied to the audio distortion analysis means 54 to measure the distortion amount (step S6 in FIG. 3). ). Here, of the audio information 61 ′, the first “a” audio information has a measured distortion amount less than the threshold, but the measured distortion amounts of the second and third audio information are greater than or equal to the threshold, and the audio is Since it is determined that the sound is distorted, the sound complementing means 55 takes out the sound information of the first sound “A” that is not distorted from the sound distortion analyzing means 54 and outputs it as it is. The voice information is replaced with voice signals “a” and “c” synthesized by voice input from the voice synthesizer 53 (step S7 in FIG. 3). The audio output means 56 performs electro-acoustic conversion on the audio signals (“A”, “I”, “U”) output from the audio complementing means 55 and generates them as audio (step S8 in FIG. 3).

  Therefore, according to the present embodiment, the voice is supplemented by using the voice synthesized voice signal even when the voice is distorted, such as when the quality of the communication line is poor or the bandwidth is insufficient. Therefore, you can have an uninterrupted conversation.

  Next, another embodiment of the present invention will be described in detail with reference to the drawings. The system configuration of the present embodiment is almost the same as that of FIG. 1, but in this embodiment, the transmission side telephone 1 transmits the integrated data 7 in the format shown in FIG. 5 is characterized in that the integrated data 7 'in the format shown in FIG.

  Here, as shown in FIG. 4A, the integrated data 7 includes the voice quality indicating the voice quality of the received voice (for example, the voice quality of the sender) in addition to the voice information 61 and the character information 62 of the integrated data 6. Information 71 is added. This voice quality information 71 is a voice recognition result for each single sound by a known method based on the analog voice transmission signal of the speaker captured by the voice input unit 11 in the voice recognition unit 12 shown in FIG. It is a digital signal relating to a voice waveform indicating voice quality, generated together with the voice information 61.

  The voice / character information integration unit 13 in FIG. 1 includes voice information (indicated by 61 in FIG. 4A) indicating the transmitted voice signal output from the voice input unit 11 and the character output from the voice recognition unit 12. Information (indicated by 62 in FIG. 4 (A)) and voice quality information (indicated by 71 in FIG. 4 (A)) are integrated for each single tone to create integrated data 7 in the format shown in FIG. 4 (A). To do. The integrated data transmitting unit 14 transmits the integrated data 7 output from the voice / text information integrating unit 13 to the public network 3 via the exchange 2 to the telephone 5 on the receiving side.

  The exchange 2 transmits the received integrated data 7 to the telephone 5 on the receiving side via the public network 3 and the exchange 4. The integrated data receiving means 51 of the telephone 5 receives the integrated data transmitted from the exchange 4, and obtains the received integrated data 7 'having the format shown in FIG. The voice / character information separation means 52 separates the voice information 61 ′, the character information 62, and the voice quality information 71 shown in FIG. 4B constituting the reception integrated data 7 ′ received by the integrated data receiving means 51. The character information 62 and the voice quality information 71 are output to the voice synthesis means 53, and the voice information 61 'is output to the voice distortion analysis means 54.

  The voice synthesizer 53 generates a voice synthesized sound of the character indicated by the character information 62 by a known voice synthesis process based on the input character information 62 and voice quality information 71. At this time, the voice synthesized sound is converted into voice quality. It is generated so as to be a voice synthesized sound that approximates the voice quality indicated by the information 71. On the other hand, the audio distortion analysis means 54 of FIG. 1 measures the distortion amount of the input audio information 61 '.

  When the measured distortion amount of the voice information 61 ′ is distorted by a predetermined threshold value or more, the voice complementing unit 55 converts the distorted voice information 61 ′ into the synthesized voice synthesis signal from the voice synthesis unit 53. When the measured distortion amount is less than a preset threshold value and can be regarded as not distorted, the received audio information 61 ′ is output to the audio output unit 56. The voice output unit 56 performs electro-acoustic conversion on the voice information 61 ′, which is an analog voice signal input from the voice complement unit 55, or the voice synthesis signal, which is an analog voice signal, and generates the voice as an incoming voice of the telephone 5. .

  As described above, according to the present embodiment, when the voice information 61 ′, which is an analog voice signal in the telephone voice signal (integrated data 7 ′) received by the receiver-side telephone 5, is not distorted, The synthesized speech synthesized by the speech synthesizing means 53 is produced only when the personal transmission voice of the telephone 1 is produced and the voice information 61 'is distorted. Even when there is a shortage of speech, the conversation can be performed without interruption, and the synthesized speech sound is generated with a voice quality similar to the voice of the sender, so that if the speech 61 'is distorted, the synthesized speech sound Has a new effect that the voice of the sender can be replaced without any sense of incongruity and can be heard by the receiver.

  Note that the present invention is not limited to the above embodiment. For example, the voice quality information 71 defines the voice quality of the received voice, not the voice quality of the sender as in the above embodiment. It is also conceivable that the voice quality information of the sender or receiver is set in advance, and the voice synthesized sound generated by the telephone 5 on the receiving side is set as the voice synthesized voice of the desired voice quality. Further, the present invention can be applied not only to a telephone line but also to a use such as a mobile phone and a wireless environment, and further to a use such as a conversation under a noisy environment. That is, the present invention can be applied not only to fixed telephones but also to general telephone terminals such as mobile phones and simplified mobile phones (PHS).

It is a system configuration figure of one embodiment of the invention. It is a figure which shows an example of the format of the integrated data transmitted and received by one embodiment of this invention. 1 is a system configuration diagram illustrating a specific example of an embodiment of the present invention. It is a figure which shows an example of the format of the integrated data transmitted and received by other embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Transmission side telephone 2, 4 Switch 3 Public network 5 Reception side telephone 6, 7 Integrated data 6 ', 7' Integrated data received 11 Voice input means 12 Voice recognition means 13 Voice / character information integration means 14 Integration Data transmitting means 51 Integrated data receiving means 52 Voice / character information separating means 53 Voice synthesizing means 54 Voice distortion analyzing means 55 Voice complementing means 56 Voice output means 61, 61 'Voice information 62 Character information 71 Voice quality information

Claims (4)

In a telephone voice recognition system for receiving a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission-side telephone terminal and restoring the transmission voice signal by synthesizing the voice recognition signal at the reception-side telephone terminal,
The transmitting-side telephone terminal is characterized by character information which is a digital signal for each single sound of the transmitted voice obtained by performing voice recognition on the transmitted voice signal and voice information which is an analog voice signal for each single sound of the transmitted voice signal. Means for generating and transmitting integrated data comprising:
The receiving-side telephone terminal receives the integrated data and separates the character information and the voice information. When the distortion amount of the separated voice information is equal to or greater than a set threshold, the character information is voiced. A speech synthesis signal obtained by synthesis is selected, and when the amount of distortion is less than the threshold, separation / selection means for selecting the separated speech information, and the selected speech synthesis signal or the speech information is transmitted. A telephone voice recognition system comprising voice output means for generating voice as spoken voice. In a telephone voice recognition system for receiving a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission-side telephone terminal and restoring the transmission voice signal by synthesizing the voice recognition signal at the reception-side telephone terminal,
The sender telephone terminal is
Voice input means for capturing the transmitted voice as a transmitted voice signal;
Speech recognition means for obtaining character information which is a digital signal for each single sound of the transmitted speech by speech recognition processing from the transmitted speech signal;
Integration means for generating analog data for each single tone of the transmitted voice signal output from the voice input means as voice information, and generating integrated data integrated with the character information output from the voice recognition means in a predetermined unit; ,
Transmitting means for transmitting the integrated data to the other telephone terminal;
The receiving telephone terminal is
Receiving means for receiving the integrated data;
Separating means for separating the integrated data received by the receiving means into the voice information and the character information;
Distortion analysis means for measuring the amount of distortion of the separated voice information;
Speech synthesis means for performing speech synthesis processing on the separated character information and outputting a speech synthesis signal;
The speech synthesis signal is selected when the distortion amount measured by the distortion analysis means is greater than or equal to a set threshold value, and the distortion amount is measured when the distortion amount is less than the threshold value. Complementing means for selecting the audio information;
A telephone speech recognition system, comprising: a voice output unit that generates the voice synthesis signal or the voice information output from the complementing unit as voice. In a telephone voice recognition system for receiving a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission-side telephone terminal and restoring the transmission voice signal by synthesizing the voice recognition signal at the reception-side telephone terminal,
The sender telephone terminal, speech the transmitter and the character information is an audio signal and a digital signal for each monophonic transmission voice obtained by speech recognition, which is an analog audio signal for each single tone of the transmitted voice signal a information and, means for generating and sends the integrated data consisting of a digital signal der Ru voice quality information indicating the voice quality of speech waveform talker or preferences,
The receiving-side telephone terminal receives the integrated data and separates the character information, the voice information, and the voice quality information, and when the amount of distortion of the separated voice information is equal to or greater than a set threshold value, A speech synthesis signal obtained by speech synthesis processing that obtains a speech synthesis sound that approximates the character information to the voice quality indicated by the voice quality information is selected, and when the distortion amount is less than the threshold value, the separated speech information A telephone speech recognition system, comprising: separation / selection means for selecting a voice, and voice output means for generating the selected voice synthesis signal or the voice information as a transmission voice. In a telephone voice recognition system for receiving a voice recognition signal obtained by voice recognition of a transmission voice signal from a transmission-side telephone terminal and restoring the transmission voice signal by synthesizing the voice recognition signal at the reception-side telephone terminal,
The sender telephone terminal is
Voice input means for capturing the transmitted voice as a transmitted voice signal;
Speech recognition means for obtaining character information which is a digital signal for each single sound of the transmitted speech by speech recognition processing from the transmitted speech signal;
Voice information which is an analog voice signal for each single sound of the transmitted voice signal output from the voice input means, the character information output from the voice recognition means, and a voice waveform of a sender or favorite voice quality and integration means for generating integrated data and voice quality information that integrates a predetermined unit is a digital signal indicating,
Transmitting means for transmitting the integrated data to the other telephone terminal;
The receiving telephone terminal is
Receiving means for receiving the integrated data;
Separating means for separating the integrated data received by the receiving means into the voice information, the character information, and the voice quality information;
Distortion analysis means for measuring the amount of distortion of the separated voice information;
Based on the separated character information and voice quality information, voice synthesis means for outputting a voice synthesis signal of a voice synthesized sound in which the character information is approximated to the voice quality indicated by the voice quality information;
The speech synthesis signal is selected when the distortion amount measured by the distortion analysis means is greater than or equal to a set threshold value, and the distortion amount is measured when the distortion amount is less than the threshold value. Complementing means for selecting the audio information;
A telephone speech recognition system, comprising: a voice output unit that generates the voice synthesis signal or the voice information output from the complementing unit as voice.

Images