JPS6038745B2 - Voice information input device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of the Invention] The present invention relates to a voice information input device for inputting information by voice by a human being.

[Background of the invention]

Conventional voice information input devices have adopted a method of displaying the recognition results of voice information input on the screen of a display device such as a CRT in the form of characters or symbols.

For this reason, there was a drawback that it was not possible to input voice information from a distance from the screen. Therefore, a method was devised to answer back the voice information recognition results in voice, which relieved the above-mentioned inconvenience, but in this case, a huge amount of synthesized voice memory for answer back was required, equal to the number of voice input information. The disadvantages are that the device becomes large and expensive. [Object of the Invention] The object of the present invention is to solve the drawbacks of the prior art, to make the device smaller and cheaper, and to provide a voice information input with a high recognition rate that answers the voice information recognition result by voice. We are in the process of providing equipment.

[Summary of the invention]

The present invention is configured to answer back a temporarily stored input voice when the recognition result of voice information input is complete, and to answer back a synthesized voice stored in advance when recognition is incomplete. It is.

[Embodiments of the invention]

Embodiments of the present invention will be described based on the drawings.

FIG. 1 shows the configuration of an embodiment of a voice input device according to the present invention. an A/D converter 12 that converts the
Registered voice memory 14 that stores registered voices in advance
and a voice recognition control circuit 13 that performs voice recognition by comparing input voice and registered voice. . On the other hand, the audio output device 2 includes an input audio memory 25 for temporarily storing input audio for audio output, a synthesized audio memory 22 for storing audio for audio output, and an input audio memory 22 for storing input audio for audio output, and a synthesized audio memory 22 for storing input audio for audio output, and a synthesized audio memory 22 for storing input audio for audio output. Voice memory 25 or synthetic voice memory 2
Audio output control circuit 21 that selects and outputs the memory contents of 2.
, a D/A converter 23 that converts the output signal of the audio output control circuit 21 into an analog signal, and an amplifier 24 that amplifies the analog signal and generates answerback audio from the speaker (or earphone) 7.

Further, the control circuit 3 is the voice discussion control circuit 1 of the voice recognition device 1.
3 to capture the voice recognition results, control the voice output control circuit 21 of the voice output device 2 to output an answer back sound from the speaker 7, display the control status and voice recognition on the display (or printer) 5, etc. This is a control computer that displays (or prints out) results, etc.

This control circuit 3 is controlled not only by voice but also by a keyboard 4. Next, a list showing an example of audio words using an embodiment of the present invention is shown below.

In the same table, the phonetic word No. 1~No. 42, that is, the range Q is the registered voice for voice recognition, and is registered in the registered voice memory 4 in advance.

To register the voice, the speaker uses the microphone 6 to register the voice word No. 1
~No. This is performed by sequentially reading out 42 yen aloud, and the old voice is stored in the registered voice memory 14 via the amplifier 11, A/D converter 12, and voice recognition control circuit 13. The range of Q in the above table is the storage content of the voice output input voice memory for making a voice answer back when voice recognition is performed. Now, the speaker uses the microphone 6 to select the audio word No. 1~No. 42 by voice, the voice recognition control circuit 13 registers the registered voice memory 4.
The same No. is selected from among the audio words stored in the . Find the audio word and enter its No. Data indicating this is output to the control circuit 3.

The control circuit 3 inputs the voice word number data and not only captures it as data and displays it on the display 5, but also issues a command to the voice output control circuit 21 to make an answer.

The voice output control circuit 21 responds to the answerback command from the control circuit 3 by audibly outputting the speaker's input word itself, which was temporarily stored in the input voice memory 25, from the speaker 7 via the O/A converter 23 and the amplifier 24. Output. That is, when the speaker inputs a voice word from the microphone 6 and recognizes that it is a voice word that has been registered in advance in the registered voice memory 14, the latter inputs the voice word from the microphone 6. The spoken word itself is outputted from the input speech memory 25 where it was temporarily stored and broadcast from the speaker 7. In short, if the input voice is completely recognized, the speaker's input voice will be heard by the speaker 7.
If the storage in the input voice memory 25 that is answered back from the input voice is erased before accepting the next input voice, it is possible to create an inexpensive device with a very small storage capacity.

Audio word No. in the table above. 51~No. 55, that is, the range 8 is the storage content of the synthesized voice memory 22 for voice output, and is recorded in the synthesized voice memory 22 in advance.

The method for recording voice words in the synthesized voice memory 22 is voice word No. 51~No. A method of recording 55 audio words as they are, audio word No. 51~No. Each of the 55 audio words is divided into several words, and the audio output control circuit 21
Furthermore, only the phoneme segments are recorded in the synthesized speech memory 22, and the speech output control circuit 21 outputs them in the correct order. 51~No. 55 may be synthesized and output as a spoken word, and any method may be used. Now, the speaker selects audio word No. When a voice other than 1 to 42 that is not stored in the registered voice memory 14 is input, the voice recognition control circuit 13 determines that the voice is unrecognizable and outputs the unrecognizable level to the control circuit 3.

The control circuit 3 takes in the unrecognizable level and displays this on the display 5 or issues a command to the audio output control circuit 21. The voice output control circuit 21 receives a command from the control circuit 3 and reads out the phoneme data in the synthesized voice memory 22 in an orderly manner, or directly outputs the voice word number. 51~No. 55
The synthesized voice is output from the speaker 7 via the D/A converter 23 and the amplifier 24. In other words, if the voice input by the speaker cannot be recognized, the synthesized voice memory 22
The answerback voice is output as a synthesized voice by assembling the phoneme pieces within. Next, FIG. 3 shows the processing contents of the audio information input device according to the present invention.

In the figure, when the speaker's voice is input from the voice input microphone 6 at step 101, the input voice from step 101 is amplified by the amplifier 11 at step 102. Further, in step 103, the audio signal is converted into a digital signal by the A/D converter 12, and in step 104, the input audio that has been A/D converted in step 103 is temporarily stored in the input audio memory 25 in order to output the input audio. be remembered. Next, in step 105, the voice recognition control circuit 13 extracts features of the input voice to facilitate voice recognition.
In step 106, the speech recognition control circuit 13 compares the input speech with the stored contents of the registered speech memory 14, and performs recognition determination. If the stored contents of the registered voice memory 4 and the input voice match, complete recognition is assumed and the process moves to step 107. On the other hand, when the stored contents of the registered voice memory 14 and the input voice do not match, incomplete recognition is assumed and the process moves to step 113. do. In the case of complete recognition, a signal output is sent to the control data in step 112, and at the same time, in step 107
The input audio temporarily stored in the input audio memory 25 is sent to the D/A converter 23 by the audio output control circuit 21. In step 108, the digitized input audio is converted to analog audio by the D/A converter 23,
Next, in step 109, audio amplification is performed by the amplifier 24. Furthermore, in step 110, the speaker or earphone 7 outputs audio, and the audio is emitted to the speaker. Then, in step 111, the input audio data temporarily stored in the input audio memory 25 is cleared by a control signal from the audio output control circuit 21 in order to input the next input audio data. On the other hand, if it is determined that the voice recognition is incomplete in the voice recognition determination at step 106, a signal is sent to the control data at step 112, and at step 113, the voice output control circuit 21 outputs the phoneme pieces stored in the synthesized voice memory 22. After synthesis, in the next step 114, synthesized speech is output.

The following steps are performed in the same way as when complete recognition is determined in step 106, D/A conversion in step 108, and step 10.
After voice amplification in step 9 and voice output in step 110, synthesized voice is generated from the speaker or earphone 7 to the speaker. Then, the temporarily stored input audio data is cleared by the temporary memory clear in step 111, and the process advances to wait for the next audio input, that is, to the audio input in step 101. As described above, in this embodiment, when the input speech is completely recognized, the input speech is generated, and when the input speech is incompletely recognized, the phoneme segment synthesized speech is generated. [Effects of the Invention] According to the present invention, the storage capacity of the synthesized speech memory 22 for speech words is extremely small, and the device can be reduced in size and weight.

In addition, if the speaker's voice input is completely recognized, the speaker's voice itself will be answered back, and if it is incompletely recognized, a synthesized voice will be used as an answerback, so it will be answered back whether the input voice information has been completely recognized. Easily identified by sound.

[Brief explanation of drawings]

FIG. 1 is a system configuration diagram showing the configuration of a voice information input lamp according to an embodiment of the present invention, and FIG. 2 is a flowchart showing the processing contents of the voice information input device shown in FIG.

Claims (1)

[Scope of Claims] 1. In a voice information input device that answers the recognition result of voice input information by voice, the voice information is taken in and the voice input information is compared with a registered voice registered in advance to determine if they match. A voice recognition device that determines the recognition state of voice input information based on a discrepancy; a control circuit that receives the determination output of the voice recognition device and outputs an answerback command according to the recognition state; and a control circuit that temporarily stores the voice input information. In response to an answerback command from the control circuit, when the voice input information is fully recognized, the input voice is answered, and when the voice input information is incompletely recognized, the previously stored voice is answered, and the temporarily stored voice input is answered. A voice information input device comprising: a voice input device for clearing information. 2. The voice information input device according to claim 1, wherein the voice output device obtains the voice to be answered during incomplete recognition by synthesizing pre-stored phoneme piece data.