JPH043099A - Voice recognition processor - Google Patents

Abstract

PURPOSE:To eliminate the need for the registration of the recognized words by a subscriber at the time of the recognition of a specific speaker by specifying the voice pattern group to which the subscriber's voice patterns belong from previously registered voice pattern groups at the time of registration and recognizing the specific speaker by the specified voice pattern group at the time of the recognition of the subscriber. CONSTITUTION:The voice recognition data 10 from the subscriber is related to the most resembling voice recognition data 13 in the voice recognition data 12 of the mother subscriber group consisting of the voice patterns of the plural recognition words which are respectively previously registered. The specific speaker is recognized by the voice recognition data 13 of the related mother subscriber group at the time the subscriber utilizes a voice recognition service. The subscriber is able to utilize the specific speaker recognition which can recognize the many words with a high recognition ability in this way without registering the recognition words by the subscriber.

Description

TECHNICAL FIELD The present invention relates to a speech recognition processing device, and more particularly to speech recognition processing of speech from a telephone connected to a telephone line.

Conventional technology Conventionally, in this type of speech recognition processing, either speaker-independent recognition with a very limited number of words or specific speaker recognition was performed in which all recognized words for each subscriber were registered in advance. .

Such conventional speech recognition processing has the disadvantage that the number of recognized words is extremely limited in the case of speaker-independent recognition where the number of words is extremely limited.

In addition, in the case of specific speaker recognition in which all recognized words are registered for each subscriber in advance, the troublesome task of having all subscribers (users using voice recognition and processing) register all recognized words is eliminated. There is a drawback to being forced.

Purpose of the Invention The present invention has been made to eliminate the drawbacks of the conventional methods as described above, and is capable of recognizing a specific speaker that can identify many words with high discrimination ability without registering recognized words by the subscriber. The purpose is to provide a speech recognition processing device that can be used.

Structure of the Invention The speech recognition 1 processing device according to the present invention includes a storage means for storing a plurality of speech pattern groups each consisting of speech patterns of a plurality of recognition words registered in advance, and a storage means for storing a plurality of speech pattern groups each consisting of speech patterns of a plurality of recognition words registered in advance; specifying means for specifying a voice pattern group to which the user belongs from among the plurality of voice pattern groups stored in the storage means, and recognizing a specific speaker using the voice pattern group specified by the specifying means when identifying the subscriber. It is characterized by having a recognition processing means for performing.

Embodiment Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, and FIG. 2 is a conceptual diagram showing speech recognition data according to the embodiment of the present invention. In these figures, when a subscriber performs subscriber identification and registration using a voice registration telephone 8 and a voice registration support personal computer 9, the voice registration telephone 8 and voice registration support personal computer 9 are connected via a telephone network 6. Then, the voice recognition processing device 1 receives subscriber's voice recognition data 10 based on a small number of recognition words manually input from the voice registration telephone 8 and the voice registration support personal computer 9, and the subscriber's voice recognition data 10 registered in advance in a memory (not shown). The voice recognition data 12 of each of the plurality of recognized words of the mother subscriber group are compared, and the voice recognition section 2, voice transmission section 3, and terminal control section 4 perform subscriber identification registration.

Through this subscriber identification registration, the subscriber's voice recognition data 10 from the voice registration telephone 8 and voice registration support personal computer 9 is associated with the voice recognition data 13 closest to the voice recognition data 12 of the mother subscriber group. It will be done.

At this time, through the subscriber identification registration, the voice recognition data 10 created by the subscriber obtained through the registration process, the voice recognition data 13 of the mother subscriber group associated with this voice recognition data 10, and the subscriber uniquely registered subscriber registered voice recognition data 11 is created.

After this, the subscriber or telephone 7-1. When receiving the voice recognition service by 7-2, the voice recognition data 10 created by the subscriber, the voice recognition data 13 of the mother subscriber group associated with this voice recognition data 10, and the subscriber registered voice recognition data 11 are used. Specific speaker recognition is performed by the speech recognition processing device 1.

In this specific speaker recognition, calls made from a specific telephone are considered to be used by a specific subscriber.
The voice recognition processing device 1 checks whether the call is from a specific telephone, and then checks whether the subscriber registered voice recognition data 11 from the subscriber is registered, thereby identifying the subscriber. Ru.

When a subscriber is identified through these checks, specific speaker recognition processing is performed using the voice recognition data 13 of the mother subscriber group associated with the voice recognition data 10 of this subscriber.

FIG. 3 is a diagram showing an example of the voice recognition data 12 of the mother subscriber group shown in FIG. In the figure, the voice recognition data 12 of the mother population group includes the recognized words "yes", "no", . . . "Tokyo", "Osaka", . . . for mother population patterns 1 to n, respectively. ... "Director", "Section Manager",
The voice patterns of ... have been registered.

The operation of one embodiment of the present invention will be explained using these FIGS. 1 to 3.

When a subscriber performs subscriber identification and registration using a voice registration telephone 8 and a voice registration support computer 9, when the voice registration telephone 8 and voice registration support computer 9 are connected via the telephone network 6, voice recognition is performed. The processing device 1 asks the subscriber to say some of the recognition words registered as the voice recognition data 12 of the mother subscriber group, such as "yes,""no,""Tokyo," and "manager." The speech pattern of the recognized word is collected.

The speech recognition processing device 1 performs the subsequent recognition processing, that is, the association with the mother subscriber group patterns 1 to n by comparing the speech patterns with the mother subscriber group patterns 1 to n.

- In terms of numbers, ask students to say recognition and idioms three times, take the average speech pattern from the first and second times, and judge whether or not the average speech pattern can be recognized from the third time.

At this time, in addition to the voice pattern collection described above, recognition words for identifying the subscriber, such as "Ore" and "My name", are registered as the subscriber's registered sound recognition data 11.

After this, the subscriber or telephone 7-1. When receiving the sound width recognition service by 7-2, the voice recognition data W1 is checked to see if the call is from a specific telephone;
-1. The subscriber is identified by checking whether the recognition word received via 7-2 and the telephone network 6 is subscriber registered voice recognition data 11 or 100.

After the subscriber is identified through these checks, the telephone 7-1. 7-2 and the telephone network 6, the voice recognition processing device 1 converts the input voice pattern, the registered voice pattern, and the voice of the related subscriber group patterns 1 to n. A comparison search with a pattern is performed, and specific speaker recognition is performed based on the comparison search.

In this way, at the time of subscriber identification registration, the voice recognition data 0 from the subscriber is converted into the voice recognition data 12 of the mother group consisting of the voice patterns of a plurality of recognized words, each registered in advance.
When a subscriber uses the voice recognition service, specific speaker recognition is performed using the voice identification data 13 of the mother subscriber group that is related to the nearest voice recognition group 13. By doing this, specific speaker recognition that can identify many words with high discrimination ability can be used without registering recognized words by the subscriber.

DETAILED DESCRIPTION OF THE INVENTION According to the present invention, when a subscriber is registered, the voice pattern group to which the subscriber's voice pattern belongs is divided into a plurality of voice pattern groups each consisting of fame patterns of a plurality of recognition words registered in advance. By identifying a particular speaker from among them and performing specific speaker recognition using the identified W f4 pattern group when identifying a subscriber, specific speaker recognition that can identify many words with high discrimination ability can be performed. This has the advantage that it can be used without any effort to register recognition words.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a conceptual diagram showing voice recognition data according to an embodiment of the present invention, and FIG. 3 is the voice of the mother group of subscribers in FIG. It is a figure showing an example of recognition data. Explanation of symbols of main parts 1...Speech recognition processing device 2...Speech recognition section 3...Speech transmission section 4...Terminal control section 8...Sound registration Telephone 9...Voice registration support computer 10...Voice recognition data created before joining 11...
・Subscriber registration voice recognition data 12...Speech recognition data of mother subscriber group 13...Speech recognition data of associated mother subscriber group Figure 2, Figure 3

Claims (1)

[Claims] (1) storage means for storing a plurality of voice pattern groups each consisting of voice patterns of a plurality of recognition words registered in advance;
specifying means for specifying a voice pattern group to which the voice pattern of the subscriber belongs from among the plurality of voice pattern groups stored in the storage means when registering a subscriber; and specifying by the specifying means when identifying the subscriber; and recognition processing means for performing specific speaker recognition using the voice pattern group.