JPH02193196A - Voice recognizing device - Google Patents

Abstract

PURPOSE:To contrive the reduction of troublesomeness at the time of registration by executing together a voice pattern input of a recognition word and a key input of recognition result display use data at the time of registration. CONSTITUTION:A voice inputted from a microphone 11 is amplified by an amplifier 12, sent to a recognition use voice analyzing part 13 and a synthesis use voice analyzing part 14, and analyzed to a recognition use parameter and a synthesis use parameter, respectively. Subsequently, the recognition use parameter is stored in a standard voice pattern memory 15, the synthesis use parameter is stored in a synthesis use pattern memory 17, and until the registration of all recognition words is ended, this processing is repeated. Subsequently, a synthetic pattern is sent to a voice synthesizing part 20 in order from the head of the synthesis use pattern memory 17, and the recognition word is synthesized and outputted. A user confirms this composite tone, and thereafter, inputs recognition result display use data from a key input part 18 and confirms it in a display part 19. In such a way, a voice input and a key input can be executed together, and troublesomeness at the time of registration is reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a speech recognition device for a specific speaker, which is equipped with a registration means with excellent operability.

(c) In conventional speaker-specific speech recognition devices in which the user registers recognition words in advance, it is necessary for the user to input letters, numbers, etc. that correspond to the recognition words as recognition results in advance. However, since the input speech pattern and the data for displaying recognition results are a set of data that are associated with each other, a method is used in which the data for display is entered manually after registering the speech of the recognition word for each word. Ta. However, with this method, it is necessary to alternately utter the voice and press the keys manually, which makes the operation cumbersome and increases the burden on the user.

(C) The problem to be solved by the invention If only the voice patterns of the recognition words are registered at once, and then the data for displaying the recognition results is input manually, voice input and keyboard input can be reduced. This can be done separately, making the operation easier to understand, but as the number of words increases, it becomes difficult to remember which word was uttered in which order, which is a problem.

(d) Means for Solving the Problems In the speech recognition device according to the present invention, when registering a recognition word, #
,'l A means for analyzing and storing the data as a speech pattern for a product and at the same time storing the analyzed data as a speech pattern for synthesis, a means for synthesizing the speech pattern for synthesis, and a recognition result input after speech synthesis. By providing a means for storing the display data in a position corresponding to the speech pattern for recognition, registering the recognition words all at once, and then synthesizing the speech patterns for synthesis in order from the beginning. It becomes possible to input display data of recognition results corresponding to recognition words.

(e) Effects According to the present invention, there is no need to remember which recognition word was registered in which position when registering a voice, and the input of voice and the key input of data for displaying recognition results can be performed separately. , can be done all at once, which reduces the hassle of registration.

(f) Examples FIG. 1 shows an embodiment of the speech recognition device of the present invention.

Referring to the same figure, the flow of processing will be shown below when voice registration of recognized words and key manual input of recognition result display data are performed simultaneously at the time of voice registration.

The voice input from the microphone 11 is amplified by the amplifier 12 to an extent that the amplitude is not saturated, and is sent to the voice analysis section 13 for recognition and the voice analysis section 14 for synthesis, where it is converted into recognition parameters and synthesis parameters, respectively. be analyzed. The recognition parameters are stored in the standard speech pattern memory 15, and the synthesis parameters are stored in the synthesis pattern memory 17. This process is repeated until registration of all recognition words is completed.

Next, the synthesis patterns are sequentially sent to the speech synthesis section 20 from the beginning of the synthesis pattern memory 17, and the recognized words are synthesized and output. After confirming this synthesized voice, the user inputs recognition result display data from the key input section 18 and confirms it on the display section 19.

As described above, this process is repeated until the input of recognition result display data for all recognized words is completed.

At the time of recognition, the voice input from the microphone 11 is amplified by the amplifier 12 to such an extent that the amplitude is not saturated, and analyzed by the voice analysis section 13 for recognition to create an input voice pattern. A matching section 16 performs pattern matching between the input speech pattern and the standard speech pattern in the standard speech pattern memory 15, calculates the standard speech pattern with the smallest distance, and determines a recognized word. Then, display section 1
At 9, recognition result display data for the recognized word is displayed.

(G) Effects of the Invention As explained above, according to the present invention, inputting the speech pattern of the recognized word at the time of registration and key human input of the recognition result display data can be performed at the same time. The hassle is reduced.

Furthermore, since it is possible to confirm that the human voice has been correctly registered by checking the synthesized voice, updating the registered voice can be omitted and the burden on the user can be reduced.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of a speech recognition device according to the present invention. 11...Microphone, 16...Pattern matching section, 12...Amplifier, 17...Speech pattern memory for synthesis, 13...Speech analysis section for recognition, 18...Data input section for display , 14...Speech analysis unit for synthesis, 19.
...Display section, 15...Standard audio pattern memory, 20.
...Speech synthesis section.

Claims (1)

[Claims] (1) In a specific speaker speech recognition device equipped with a speech input means and a speech analysis means, a memory for storing speech patterns for recognition at the time of speech registration, and speech data for synthesizing input speech corresponding to each speech pattern are stored. a storage means comprising a memory for storing and a memory for storing display data of the recognition result;
Means for analyzing input speech into speech patterns for recognition at the time of speech registration and simultaneously analyzing and storing speech data for synthesis, means for speech-synthesizing the stored speech data for synthesis, and displaying input recognition results after speech synthesis. A method is provided to store the data for recognition in a location that corresponds to the voice pattern for recognition, and the input voice is recorded when registering recognition words, and when all recognition words have been registered, they are played back in order from the beginning. A speech recognition device characterized by inputting data for displaying recognition results.