JPH0721759B2 - Speech recognition response device - Google Patents

Description

TECHNICAL FIELD OF THE INVENTION The present invention relates to a voice recognition response device used in an information processing system by voice input.

[Technical background of the invention and its problems]

Recently, the development of speech recognition technology and speech synthesis technology has been remarkable. For example, continuous speech recognition and speech recognition for unspecified speakers are possible, and highly accurate speech synthesis using the linear predictive coding method is possible. Has become. In addition, research and development have been actively conducted on a rule synthesis method for converting sentences into speech.

Thus, using such a technology, for example, a telephone voice response service system for performing various services using a public telephone line, and an online business system in a bank or the like are being developed, and its usefulness is drawing attention. ing.
However, the number of users of this type of system is unspecified, and for example, there are unfamiliar people such as old people and children, or people who use the system many times a day. Despite this,
In the conventional device, the content of the voice response is uniform and the utterance speed is also constant, so that the dialogue between the human and the machine is not smooth. In other words, the response is redundant and frustrating, or the response is difficult to understand.

[Object of the Invention]

The present invention has been made in consideration of such circumstances, and an object thereof is to enable natural and smooth dialogue between a human and a machine to enable effective information processing by voice input. It is to provide a highly practical voice recognition response device.

[Outline of Invention]

The present invention, in a voice recognition response device for recognizing an input voice and outputting a response voice in response to the recognition result, a detecting means for detecting a time length of the input voice in word units, and a recognizing means for recognizing the word of the input voice. And a storage unit that pre-registers information about a standard time length of the input voice in word units, a time length detected by the detection unit and a standard time registered in the storage unit regarding a word detected by the recognition unit. It is characterized by further comprising: measuring means for measuring the utterance speed of the input voice using the length information; and control means for controlling the response voice speed according to the utterance speed measured by the measuring means.

〔The invention's effect〕

According to the present invention, since the speed of the response voice is controlled according to the utterance speed of the input voice, it is possible to appropriately give the response voice to the voice input person. For example, when the voice-inputting person speaks quickly, the response is made in a fast-paced manner, and when the speech is slow, the response is made at a slow speed, which enhances the naturalness and smoothness of the dialogue between human and machine. Can be planned.

Furthermore, in the present invention, in particular, the speech length of the input voice is measured using the time length of the input voice in word units and the standard time length registered in word units. Even if the speech rate is different, the speech rate can be accurately measured. Therefore, the output response speed can be controlled more appropriately according to the speech speed of the input voice.

Example of Invention

 Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram showing the device of the first embodiment. In this device, a speech recognition target is a word, and the speed of a voice response is controlled according to the speech speed of the word. That is, the input voice is subjected to A / D conversion processing, spectrum analysis processing, and the like via the analyzer 1, converted into a series of characteristic parameters thereof, and stored in the voice pattern memory 2. The voice section detector 3 detects the start and end of a word in a voice pattern by using, for example, energy information of the characteristic parameter time series, and the word data portion is cut out by this. Then, the pattern matching circuit 4 matches the voice pattern of the word data with each standard pattern of a plurality of words registered in the word dictionary memory 5 in advance to recognize the word. This pattern matching is performed by, for example, a similarity calculation method. The recognition result is given to the voice response output unit 6.

On the other hand, the recognition result of the input voice obtained by the pattern matching circuit 4 is given to the speech rate measuring device 7. The speech rate measuring device 7 uses the recognition result Wi of the input voice and the time length Li of the word shown as the information of the start end and the end, and uses the recognition word previously registered in the word duration memory 8 for the recognition word.
The standard time length Ri of Wi is obtained, and the utterance speed v is calculated from the average value and the variance thereof. Thus, for example, it is determined whether the speech speed v of the input voice is faster or slower than the average standard speech speed. In other words, this determines whether the voice input person is a so-called quick mouth, standard voice, or late voice. The voice response speed controller 9 variably controls the speed of the response voice by the voice response output unit 6 based on the information about the speech rate.

In other words, the utterance speed of the input voice is not determined by the movement or pitch of the vowel as the entire input voice, but by the speed of the meaningful part that depends on the linguistic content, that is, the word part. Has been done. As a result, for example, when only an important portion of a word is slowly emphasized and uttered, the voice (word) responding to the utterance is controlled to be slowly uttered.

As a result, the voice response output unit 6 variably controls the voice output speed of the response sentence determined in accordance with the recognition result of the input voice in accordance with the utterance speed of the input voice to provide a voice response. . At this time, when the response voice is synthesized and output by the rule synthesis method, the response voice speed is variably controlled by controlling the changing speed of various parameters for the rule synthesis. Further, in the case of performing the voice synthesis of the recording edit type, the response voice speed is controlled by selecting a prerecorded sentence or voice unit having a different utterance speed.

Thus, according to the present device configured as described above, since the voice response is performed at the speaking speed according to the speaking speed of the voice input person, the so-called impatient and fast-paced type can be used for the so-called impatient and quick-mouthed person, and the leisurely type can also be used. With this, it becomes possible to make a voice response to a slow-moving person at a slow speed, and it becomes possible to enhance the naturalness of the dialogue between the human and the machine and smooth the dialogue. As a result, it is possible to comprehensively improve the information processing efficiency by the voice recognition response.

As described above, according to the present invention, the voice utterance speed that well reflects the character of the voice input person is detected, and the speed of the voice response is controlled accordingly, so that the naturalness of the dialogue with the voice input person can be improved. Can be increased. As a result, practically great effects such as elimination of problems such as irritation to the voice input person can be obtained.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a voice recognition response device according to an embodiment of the present invention. 1 …… Analyzer 2 …… Voice pattern memory 3 …… Voice section detector 4 …… Pattern matching circuit 5 …… Word dictionary memory 6 …… Voice response output unit 7 …… Speech rate measuring instrument 8 …… Word duration Long memory 9 ... Voice response speed controller

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Teruhiko Ukita 1 Komukai Toshiba-cho, Kouki-ku, Kawasaki-shi, Kanagawa Higashi Koshibaura Electric Co., Ltd. (56) Reference JP-A-59-153238 (JP, A) ) JP-A-57-57375 (JP, A)

Claims (1)

[Claims] 1. A voice recognition response device for recognizing an input voice and outputting a response voice in response to the recognition result, comprising: a detecting means for extracting a word portion of the input voice and detecting a time length of the extracted word unit. A recognition unit that recognizes a word of the input voice, a storage unit that pre-registers information regarding a standard time length of the input voice in word units, and a time that is detected by the detection unit regarding a word recognized by the recognition unit. Measuring means for measuring the utterance speed of the input voice using the length and information of the standard time length registered in the storage means, and outputting the response voice in accordance with the utterance speed measured by the measuring means. A voice recognition response device, comprising: a control unit for controlling a speed.