JPS59216242A - Voice recognizing response device - Google Patents

Abstract

PURPOSE:To make a natural and smooth interaction between a person and a machine possible by measuring the speaking speed of an input voice and controlling the output of a response voice in accordance with this speaking speed. CONSTITUTION:With respect to the input voice, the start and the end of a word in a voice pattern are detected by an analyzer 1, a voice pattern memory 2, and a voice section detector 3. A pattern collating circuit 4 collates the voice pattern of word data with standard patterns of plural words registered preliminarily in a word dictionary memory 5 to recognize the word. The recognition results of the input voice obtained by the pattern collating circuit 4 is given to a voice response output part 6 and a speaking speed measurer 7, and the measurer 7 uses the input voice recognition result and the time length between the start and the end to obtain a standard time length. The speaking speed is calculated on a basis of an average value of standard time lengths and variance. A voice response speed controller 9 attains information concerning this speaking speed to control variably the speed of the response voice due to a voice response output part 6.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a voice recognition response device used in an information processing system using voice input.

[Technical background of the invention and its problems]

In recent years, there has been remarkable progress in speech recognition technology and speech synthesis technology. For example, it has become possible to perform continuous speech recognition and speech recognition for unspecified speakers, and it has also become possible to perform highly accurate speech synthesis using linear predictive coding. It becomes. There is also active research and development on rule synthesis methods for converting sentences into speech.

Using such technology, attempts have been made to develop, for example, a telephone voice response service system that provides various services using public telephone lines, and an online business system for banks, etc., and their usefulness is attracting attention. ing.

However, the users of this type of system are an unspecified number of people, including people who are inexperienced, such as the elderly and children, and people who use the system several times a day. In spite of this, in the conventional device f5, the content of the voice response is uniform, and the speech rate is also constant, so that dialogue between humans and machines cannot be carried out smoothly. In other words, a problem arises in that the responses are redundant and irritating, or the responses are difficult to understand.

[Purpose of the invention]

The present invention was made in consideration of these circumstances, and its purpose is to enable natural and smooth interaction between humans and machines and to enable effective information processing through voice input. An object of the present invention is to provide a highly practical voice recognition response device.

[Invention planting]

In the present invention, when recognizing an input voice and giving a voice response, the speech rate of the input voice is measured, and the output of the response voice, such as the response voice rate and response content, is controlled according to this speech rate. It is something.

〔Effect of the invention〕

Thus, according to the present invention, since the voice response output is controlled according to the speech rate of the input voice, it is possible to give an appropriate response to the voice input person. For example, it is possible to give appropriate instructions for voice input by giving concise responses to people who use the device frequently, and polite responses to people who use it infrequently, making the dialogue more natural. This makes it possible to sufficiently improve performance and smoothness.

[Embodiments of the invention]

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

FIG. 1 is a schematic configuration diagram showing a first embodiment of the apparatus. This device uses a word as a speech recognition target, and controls the speed of the speech response according to the speech rate of the word. That is, the input speech is subjected to A/D conversion processing, spectrum analysis processing, etc. via the analyzer 1, converted into a series of characteristic parameters, and stored in the speech pattern memory 2.

The speech section detector 3 detects the start and end of a word in a speech pattern by using, for example, energy information of the characteristic parameter time series, and thereby extracts word data portions. But t'? The turn matching circuit 4 matches the audio pattern of the word data with each standard pattern of a plurality of words registered in advance in the word dictionary memory 5 to recognize the word. This /9 turn matching is performed, for example, by a similarity calculation method. This recognition result is given to the voice response output section 6.

On the other hand, the recognition result of the input speech obtained by the pattern matching circuit 4 is given to the speech rate measuring device 7. This speech rate measuring device 7 uses the recognition result Wi of the input speech and the time length Li of the word indicated as the start and end information,
The standard duration Ri of the recognized word Wi registered in advance in the word duration memory 8 is obtained, and the speech rate τ is calculated from the average value and variance thereof. This determines, for example, whether the speech rate τ of the input voice is faster or slower than its average standard speech rate. In other words, it is determined whether the voice input person is a so-called fast talker, a standard talker, or a slow talker. The voice response speed controller 9 obtains information regarding the speech rate and variably controls the speed of the response voice output by the voice response output section 6.

As a result, the voice response output unit 6 outputs a voice response by variably controlling the voice output speed of the response sentence determined according to the recognition result of the input voice according to the speaking speed of the input voice. .

At this time, when the response voice is synthesized and output using the rule synthesis method, the response voice speed is variably controlled by controlling the rate of change of various parameters for the rule synthesis. Furthermore, when voice synthesis in the form of recording and editing is performed, the response voice speed is controlled by selecting pre-recorded sentences or voice segments having different speaking speeds.

Thus, according to the present device configured in this way, the voice response is performed at a speaking speed that corresponds to the speaking speed of the person inputting the voice, so it is possible to respond to the so-called impatient and fast-talking person in a fast-talking manner, and to respond to a person who is an impatient person who speaks quickly, or to a leisurely person. This makes it possible to respond by voice at a slower speed to people who are late, thereby increasing the naturalness of dialogue between humans and machines and making it smoother. As a result, it is possible to improve the information processing efficiency through voice recognition responses.

Figure 2 is a schematic configuration diagram showing the second embodiment of the present invention, in which the pitch frequency of the input voice is determined, the speaking rate is detected from the change, and the content of the response sentence itself is changed. This is what I did. That is, the input speech is analyzed through the analyzer 11, and the speech pattern is stored in the speech pattern memory 12. The speech recognition unit 13 recognizes this speech pattern by referring to a speech dictionary registered in the dictionary memory 14.

On the other hand, the pitch frequency component of the input voice is detected through the pitch extractor 15.

Detection of this pitch frequency component is performed independently of the recognition process of the input speech using, for example, the cepstral method, the modified correlation method, the ADIVI method, or the like. The speech rate τ of the input voice is determined by the speech rate measuring device 16 from the change in the time-series pattern of the pitch frequency. The voice response control unit 17 inputs this speech rate information and the recognition result of the input voice, and determines a response sentence with corresponding content, which is output as voice via the voice response output unit 18. Ru.

Depending on the speech recognition result and the speaking speed, for example,
Even if the responses express the same meaning, such as "Thank you,""Thankyou,""Thank you, thank you," or "Thank you, thank you very much." Output. In other words, a voice response is made with the content and speed depending on the voice input person.

Therefore, when giving the next voice input instruction as a voice response, it becomes possible to politely repeat the instruction to an unfamiliar person by giving that instruction to Kanshu9.
It becomes possible to enhance the naturalness of dialogue and improve processing efficiency.

As described above, according to the present invention, since the speech rate that well reflects the personality of the person inputting voice is detected and the voice response is controlled accordingly, the naturalness of the dialogue with the person inputting voice can be enhanced. I can do it.

As a result, great practical effects can be achieved, such as eliminating problems such as irritating the voice inputter.

Note that the present invention is not limited to the above embodiments. For example, the utterance method n'[ of the input speech can be measured using vowel similarity or temporal changes in the spectrum, or when the continuous occurrence of numbers is to be recognized, the pause duration can be used. You can also do that. In addition, the voice response may be controlled by controlling the content sentence as well as the speech rate. Furthermore, various conventionally known methods for speech recognition processing and speech synthesis may be employed as appropriate. In short, the present invention can be modified and implemented without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of a device according to a first embodiment of the present invention, and FIG. 2 is a schematic diagram of a device according to a second embodiment of the present invention. 1.11...Analyzer, 2.12...Level 1 pattern memory, 3...Speech section detector, 4...Tsuya turn matching circuit, 5...Word sound material memory, 6,18 ...Voice response output unit, 7, 16...Speech rate measuring device, 8...
・Word duration length memory, 9...Voice response speed! Controller, 13... Voice recognition unit, 14... Dictionary memo 1),
15... Pitch extractor, 17... Voice response control unit.

Claims (2)

[Claims] (1) In a voice recognition response device that recognizes input voice and outputs a voice response to the recognition result, the speech rate of the input voice is measured and the voice response output is controlled according to this speech rate. Characteristic voice recognition response device. (2) Control of the voice response output varies the voice response speed,
The voice recognition response device according to claim 1, wherein the voice recognition response device is configured to change the content of the voice response sentence.