JPH01147774A - Translation communication system - Google Patents

Abstract

PURPOSE:To eliminate a response delay, and to improve the efficiency of translation communication by setting and controlling uttering speed for a translated voice to be outputted by voice synthesis so as to be a definite value. CONSTITUTION:A regulation synthesizing part 25 provided on a translation communication terminal or a central translation system fetches a character code string to show the objective translation language information of regulation synthesis through a character code input part 25a, and controls a dictionary reading part 25c under the control of a regulation synthesis control part 25b, and obtains a phoneme parameter, etc., from a regulation synthesis dictionary 26, and generates respectively a phoneme parameter system and a rhythm parameter system necessitated for the regulation synthesis of a voice according to an input character code string. A speed setting part 25d gives a speed control parameter to regulate the uttering speed of the voice to be synthesized and outputted, to the regulation synthesis control part 25b. The uttering speed of the voice to be regulation-synthesized and outputted through a speech synthesizing part 27 can be controlled so as to be constant by this speed control parameter, and the response delay can be eliminated, and the translation communication can be made efficient.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a translation communication system for performing translation communication between a plurality of communication terminals while using mutually different language sounds.

(Prior Art) It has been a long-standing dream of humankind to have parties who speak different languages communicate with each other through translation. A system that realizes such translation communication, such as Japanese
A system for translating communication between English and English, if realized using equipment currently in practical use, would basically be constructed as shown in FIG. 2.

That is, in this system, the Japanese speech recognition section recognizes the Japanese voice input from the telephone terminal a, and the Japanese-English translation section C translates the recognized Japanese language information into English language information. Then, this English language information is synthesized and converted into English speech by an English speech synthesis section d, and this is output as speech to the telephone terminal e on the other party's side. On the other hand, the English voice recognition section f recognizes the English voice input from the other party's telephone terminal e, and the recognized English language information is translated into Japanese language information at the English-Japanese translation section g. Then, this Japanese language information is synthesized and converted into Japanese "j" speech by a Japanese speech synthesis section, and this is configured to be output as speech to the telephone terminal a.

Through such a translation communication system, the user of telephone terminal a can input Japanese by voice while listening to the call information from the other party as Japanese voice, and the user of telephone terminal e can input English by voice. It becomes possible to listen to the call information from the other party as English voice while inputting information, and the translated communication (call) is now realized.

However, various problems arise when actually performing translated communication (phone call) via such a translated communication system. For example, the series of operations of translating the recognition result of input speech, synthesizing the translated language information obtained by the translation processing using a method such as a rule synthesis method, and outputting the result takes a considerable amount of processing time. Furthermore, in order to improve the accuracy of the above-mentioned speech recognition, it is necessary for the speaker to speak slowly and clearly, and in some cases, to divide the speech into words or phrases.

However, in cases such as telephone calls, where information is communicated alternately by speaking while checking the response from the other party, it is desirable that the response time be as short as possible. However, as mentioned above, after recognizing the uttered phonetic sounds that are slowly divided into words or phrases, the recognition results are summarized into sentence units, which are translated, and then the translated results are synthesized into speech and output. If, for example, the speaking speed of the synthesized translated speech is controlled in order to convey the nuances of the speech, the synthesized output speech becomes elongated and extremely difficult to hear. In addition, there were problems such as the need to confirm the voice information that was synthesized and output, and that it took an undesired amount of time to respond to the voice information.

(Problems to be Solved by the Invention) As described above, in the conventional translation communication system, the recognition and translation results are calculated according to the speech rate of the input voice in order to convey the nuances of the speech. Since the output speed (speech rate) of the speech synthesis of the translated speech is variable, there are problems such as the translated speech becoming difficult to hear and the response tending to be delayed.

The present invention was created in consideration of these circumstances, and its purpose is to synthesize and output translated voices at a speed that is easy to hear, and to eliminate delays in response to achieve effective translated communication. Our goal is to provide a translation communication system that allows you to:

[Structure of the Invention] (Means for Solving the Problems) The present invention analyzes input speech, performs recognition processing, translates the recognized linguistic information into linguistic information of a foreign language, and uses the translated linguistic information. A translation communication system that performs translated communication between communication terminals by synthesizing and outputting speech, characterized in that the speech synthesis section is provided with means for setting and controlling the speaking rate of the translated speech synthesized and output to a constant value. That is.

(Function) According to the present invention, even when voice input is performed by slowly dividing the input voice into words or phrases in order to increase the recognition rate of the input voice, the recognition and
A speech synthesis output corresponding to the translation result is generated and output while being controlled to be constant independently of the speech rate of the input speech. In other words, regardless of the speech rhythm of the input speech, the synthesized output of translated speech is performed at a constant speech rate, for example, by omitting pause periods between words. Therefore, it effectively eliminates the difficulty of hearing the translated synthesized speech, and also eliminates the response delay and translates the communication (
This makes it possible to improve the efficiency of telephone calls.

(Example) Hereinafter, an example system of the present invention will be described with reference to the drawings.

FIG. 3 is a schematic configuration diagram of the embodiment system, which is constructed by connecting translation communication terminals 1a, -In and a central translation system 2 to a communication line.

The communication lines to which these translation communication terminals 1a, ~In and the central translation system 2 are connected are communication networks built for each communication company, communication networks built for each predetermined region (country), and these networks. It is realized as a network that interconnects communication networks.

The translation communication terminals 1a, ~in basically constitute a translation communication system having the configuration shown in FIG. 2 via the above-mentioned communication line, and perform voice translation communication between the communication terminals. The form in which the processing function is provided differs depending on the specifications of various translation communication systems. In other words, these translation communication terminals 1a to 1n may be translation communication terminals that are equipped with all of the voice translation communication functions, or translation communication terminals that have a simple configuration that only includes a part of the voice translation communication function, for example, only a voice codec. Consists of communication terminals, etc.

In addition, the central translation system 2 equipped with a predetermined voice translation communication function supplements the processing functions that are insufficient in the translation communication terminal alone, depending on the configuration of the translation communication terminal that performs the translation communication and the form of the translation communication. It relays translation communications between terminals.

Specifically, in the case where the translation communication terminal is only equipped with an audio codec, the central translation system 2 recognizes the audio signal given from the translation communication terminal, processes the translation, and synthesizes and outputs the translated speech. This supplements the processing functions that the above-mentioned translation and communication device does not have. By relaying such a central translation system 2, translation communication can be performed between translation communication terminals that are not equipped with all processing functions.

Figure 4 is a diagram showing a basic configuration example of a translation communication terminal that includes all the processing functions necessary for voice translation communication.
11 is a control section, 12 is a key input section, and 13 is a display. Prior to speech translation communication, when a predetermined key input is made from the key input unit 12, the input information is transmitted to the control unit 11.
The signal is then sent out to the line via the network termination device 14. Depending on this communication mode, settings such as the configuration of the translation communication terminal, the type of information to be communicated (direct voice communication or translated communication), and the form of translation (specifying the translation language and translation method) are made.
The information is notified to the central translation system 2, and connection control of the communication line is performed. At this time, necessary message information and the like are displayed and outputted via the display 13.

Now, the voice input through the microphone 15 is A/
The signal is taken in through the D converter 16 and stored in the data memory 17, and is subjected to acoustic analysis such as filtering and tarring in the audio analysis section 18. The segment conversion unit 19 refers to the standard pattern memory 20 and converts the acoustic analysis results into
For example, segment information for speech recognition in units of phonemes, syllables, or vCv is required. The speech recognition unit 21 performs recognition processing on the input speech described above with reference to the recognition dictionary 22 in accordance with this segment information. This speech recognition process is
This is done using P matching, transition networks, etc.

At this time, you will be prompted to re-enter the voice if necessary.

The recognition results (linguistic information) obtained in this way are divided, for example, into units of phrases, and then stored in the data memory 1.
7 as appropriate.

The translation unit 23 refers to the translation dictionary 24 and translates the linguistic information recognized as two consecutive words.

This translation process includes not only translation between predetermined languages, such as Japanese-English translation and English-Japanese translation, but also translation between intermediate languages that are commonly set in this translation communication system. However, in boat fishing, the form of translation processing is set for each translation communication terminal. The language information translated in this way is sent to the communication line via the network termination device 14.

On the other hand, for the linguistic information received from the communication line via the network termination device 14, the rule synthesis section 25 refers to the rule synthesis dictionary 2B to generate a phonological/prosodic parameter sequence for the linguistic information. The speech synthesis unit 27 synthesizes such phonemes and
A speech signal is generated by rule synthesis according to the prosodic parameter series, and is outputted via the D/A converter 28. In this way, the speaker 2 uses the regularly synthesized audio signals.
9 will be driven and synthesized speech will be emitted.

Incidentally, the program memory 30 stores control programs and the like necessary for controlling the operations of each of the above-mentioned sections, and provides them to the control section 11.

Therefore, each translation communication terminal that performs translation communication is connected to a fourth
In the case of the configuration shown in the figure, for example, a Japanese input voice is translated into English information and communicated to the other translation communication terminal, and the translation communication terminal synthesizes the English speech and outputs it. Furthermore, the information voice input in English from the other translation communication terminal is translated into Japanese information, sent to the communication line, and given to the one translation communication terminal mentioned above.

Then, it is synthesized into Japanese speech and output, and speech translation communication between Japanese and English is performed here.

By the way, although one translation communication terminal is configured as shown in FIG. 4, the other translation communication terminal is an A/D converter! 6 and a D/A switch 28, and this voice codec is connected to a communication line via a network termination device 14. In such a case, the aforementioned central translation system 2 is activated to supplement the translation communication function that is lacking in the translation communication terminal configured as shown in FIG. That is, this central translation system 2 includes the above-mentioned speech analysis section 18. Segment converter 19. Standard pattern memory 20. speech recognition unit 21;
Recognition dictionary 22. Translation Department 23. Translation dictionary 24. Rule synthesis section 25. Rule synthesis dictionary 26. Speech synthesis section 27. It is configured to include a data memory 17 and a program memory 30. In other words, the central translation system 2 is configured to include all the processing functions necessary for the translation communication function except for the voice codec.

However, since the central translation system 2 is notified of the type (configuration) of the translation communication terminal that performs the voice translation communication and the mode of the translation communication by the above-mentioned communication mode prior to the start of the voice communication, the central translation system 2 uses the above information according to this information. It determines the processing functions that are insufficient to perform translation communication between translation communication terminals. In order to compensate for the insufficient processing functions, the processing mode is changed to selectively activate each of the above-mentioned units to relay speech translation communication between the translation communication terminals.

For example, if the translation communication terminal is configured with only a voice codec, the central translation system 2 inputs the voice information provided from the translation communication terminal, analyzes it, performs voice recognition, and then sends the message to the communication partner. Translated into linguistic information. This translated language information is then relayed and output to the other party's translation communication terminal configured as shown in FIG. 4, for example. Conversely, when translation language information is given from the other party's translation communication terminal, the central translation system 2 synthesizes this language information according to rules, converts it into speech information, and outputs it to the translation communication terminal equipped only with the above-mentioned speech codec. ing.

By relaying the central translation system 2 in this way, even if the translation communication terminal is equipped with only an audio codec, the voice translation communication function that is not provided therein can be supplemented by the central translation system 2. , it becomes possible to participate in the above-mentioned voice translation communication.

This means that speech translation communication between different languages can be carried out via an intermediate language, which is extremely flexible.
Moreover, it is shown that it is possible to construct a highly expandable translation communication system.

Now, the features of the present invention in the translation communication system configured as described above are as shown in FIG. , is provided with a speed setting function that controls and constantizes the speaking speed of the translated speech that is synthesized according to the rules and outputted.

That is, the rule synthesis unit 25 inputs a character code string indicating the translation language information to be subjected to rule synthesis to the character code string input unit 25a.
The dictionary reading unit 25c is controlled under the control of the rule synthesis control unit 25b to obtain phoneme parameters etc. from the rule synthesis dictionary 26, and the phoneme parameter series and prosodic parameter series necessary for rule synthesis of speech are obtained. are generated according to the input character code string. Speed setting section 25
d provides a speed control parameter that defines the speech speed of the voice to be synthesized and output to the rule synthesis control section 25b. With this speed control parameter, the speech synthesis section 2
7, the speech rate of the voice that is synthesized and output according to the rules is controlled to be constant.

That is, each element constituting the rule synthesis section 25 is composed of, for example, a program memory or dedicated hardware. In response to a request from the rule synthesis control section 25b, the dictionary reading section 25e reads, for example, synthesis (phoneme) parameters on a syllable basis or a word basis from the rule synthesis dictionary 26. The synthesis parameters used here consist of time series data such as PARCOR parameters and formant parameters, for example.
It consists of a preset number of data to generate synthetic speech at a constant speaking rate.

The character code string input unit 25a reads a character code string given from the control unit 11, that is, a character code string that is translation result data, for example, formed by inserting word delimiter codes in each sentence, This is divided into word units and sequentially provided to the rule synthesis control section 25b. The rule synthesis control section 25b issues a request to the dictionary reading section 25c to read out the synthesis parameters in accordance with the word-by-word character code string thus provided. In response to this request, intonation and accent information is added to the synthesis parameters read from the rule synthesis dictionary 2G, and this is output to the speech synthesis section 27 as time series data.

Incidentally, data regarding these information additions are stored as processing operation rules inside the rule synthesis control section 25b.

The rule synthesis control unit 25b linguistically analyzes the input sentence indicated by the input character code string, and adds accent information and the like using the above-described rules as appropriate according to the analysis result.

The control of the speech rate of the voice synthesized and outputted here is as follows:
This is done by referring to the speed control parameters stored in the speed setting section 25d and controlling the data output to the speech synthesis section 27. This speed control parameter is
For example, it consists of data on the ratio of standard synthesized speech to speech rate stored in the rule synthesis dictionary 2B, and is configured to be able to be set for each translation communication terminal via the control section 11, for example. The speed control parameters may be set, for example, by selectively switching between several set values using a switch, or by directly inputting the ratio data from the key input unit 12. good.

For example, if the value of such a speed control parameter is expressed as (5/4), 4 out of every 5 series of data are thinned out from the time series of the composite parameter, and the time series data is compressed. conduct. By supplying the time-series data compressed in this manner to the speech synthesis section 27, the speech rate can be increased to (5/4) times the standard speech rate. At this time, it is preferable that the thinned out portion of the data be interpolated by computer processing to prevent the synthesized speech from becoming unnatural.

In addition, when a speed control parameter indicated by the value (415) is given, the time series of the four series of synthesis parameters is appropriately interpolated and expanded to five series, and this time series data is used in the speech synthesis section. 27.

According to such speech rate control, for example, the pitch frequency of the synthesized speech is determined by the speech synthesis unit 27, so the speech rate can be variably set without causing a change in the speech pitch. .

Thus, according to such a rule synthesis section 25, according to the character code string indicating the translation result, time series data of speech synthesis parameters is generated according to the set speed control parameter and output to the speech synthesis section 27. It becomes possible to obtain synthesized speech (translated speech) at a constant speech speed regardless of the speech speed of the input speech from which the translation result was obtained.

Therefore, in order to improve the recognition rate for input speech during translation communication, for example, even if the speech is slowly divided into words and input, regardless of the speech speed, highly natural translation synthesized speech is produced at a constant speech speed. It becomes possible to obtain. As a result, it is possible to effectively eliminate problems such as the translated synthesized speech becoming difficult to hear or the response to the synthesized speech tending to be delayed. Therefore, effects such as being able to effectively carry out highly natural translation communication (phone call) can be achieved.

Further, since there is no need to control the speech rate of the synthesized and output voice according to the speech rate of the input voice in order to convey the nuance of the input voice, the processing is greatly simplified. Specifically, the input language information indicated by the recognition result of the input speech is
Each translation processing unit can be processed all at once, and the translation information obtained from the translation processing can be directly submitted to the speech synthesis processing, so for example, it can be used to divide sections (pause sections) between words of input speech, etc. There are great practical effects such as no need to pay attention to the above, and the processing procedure can be greatly simplified.

Note that the present invention is not limited to the embodiment system described above. For example, it may be possible to perform translation into a language other than the exemplified language, or it may be possible to perform simultaneous translation communication between three or more languages. Furthermore, as for input speech recognition processing, translation processing, and speech synthesis methods, various methods that have been proposed in the past may be adopted depending on the system specifications. It goes without saying that the speech rate control method can be modified depending on the speech synthesis method. In addition, the present invention can be modified and implemented without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, the speech rate of the synthesized output of translated linguistic information is controlled to be constant regardless of the speech rate of the input speech in consideration of the speech recognition rate. It is possible to realize a highly practical translation communication system that eliminates the difficulty of hearing and also eliminates the delay in response to synthesized output speech, making efficient translation communication possible.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of the configuration of a rule synthesis section of translated language information in a translation communication system according to an embodiment of the present invention, and FIG.
The figure is a configuration diagram showing the basic concept of translation communication, Figure 3 is an overall configuration diagram of the translation communication system, and Figure 4 is a configuration example of a translation communication terminal equipped with all translation communication functions in the embodiment system. FIG. la, ~in... Translation communication system, 2... Central translation system, 11... Control unit, 14... Network termination device, 16... A/D converter, 18... Voice analysis Part 1
9... Segment conversion unit, 21... Speech recognition unit, 2
3... Translation section, 25... Rule synthesis section, 25a...
Character code string input section, 25b... rule synthesis control section, 2
5c...Dictionary reading section, 25d...Speed setting section, 26
... Rule synthesis dictionary, 27... Speech synthesis section, 28...
・D/A converter. Applicant's agent Patent attorney Takehiko Suzue Figure 1 Figure 20 Figure 30 Figure 4

Claims (1)

[Claims] It comprises means for analyzing and recognizing input speech, means for translating the recognized linguistic information into linguistic information of another language, and speech synthesis means for synthesizing and outputting the translated linguistic information into speech. A translation communication system for performing translation communication between communication terminals, characterized in that the speech synthesis means includes means for setting and controlling the speaking rate of the translated speech outputted through speech synthesis to a constant value. .