JPH04134397A - Voice recognizing device - Google Patents

Abstract

PURPOSE:To prevent the same recognition error from being repeated many times by outputting a recognition word having the highest degree of resemblance from a successive memory at the time of first recognition and outputting a recognition word having the highest degree of resemblance other than recognition words, which are already outputted as recognition results, at the time of n-th recognition. CONSTITUTION:When a control part 8 discriminates first voice recognition by a recognition end signal from a collator 4, an output switch 12 is connected to a contact (a) to output the recognition result having the highest degree of resemblance stored in a memory 1 from a recognizing device. The recognition result of first recognition is inputted to a register 10 and is compared with the recognition result having the highest degree of resemblance of second recognition, namely, contents of the memory 1 of a successive rewrite memory 6; and if they are different, the output switch 12 is connected to the contact (a). If they are equal as the comparison result, the output switch 12 is connected to a contact (b) to output the recognition result having the highest degree of resemblance different from that of first recognition. Thus, the same recognition result error is not outputted for rephrasing.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a voice recognition device that uses voice to control various devices and input data such as timer reservations for video tape recorders and the like.

Conventional technology In recent years, with the advancement of voice recognition technology, voice recognition devices have become popular for controlling equipment by voice while performing other tasks, and for inputting timer reservation data for video tape recorders due to the ease of operation. It has started to be used.

An example of the conventional speech recognition device described above will be described below with reference to the drawings.

FIG. 2 shows the configuration of a conventional speech recognition device.

In FIG. 2, 11 is a microphone that converts audio into electrical signals. 12 is an amplifier. 13 is an analog-to-digital converter (hereinafter referred to as an A-D converter). 14 is a collation device. 15 is a recognition word memory in which patterns of words to be recognized are registered. Reference numeral 16 denotes a sequentially rewritten memory for storing recognition results. 17 is a recognition button for starting recognition. 18 is a control unit that controls recognition.

Regarding the conventional speech recognition device configured as above,
The operation will be explained below.

First, when the recognition button 17 is operated to perform voice recognition, the control section 18 controls the collation device 14 to start voice recognition. The spoken voice is converted into an electrical signal by the microphone 11, and amplified by the amplifier 12 to the amplitude required for verification. The amplified analog electrical signal is converted into a digital signal by an A-D converter 13. Recognition word memory 1
5 is a device in which the pattern of the word to be recognized is registered as a digital signal, which may be registered in advance in a read-only memory, or may be rewritten sequentially by user registration at the time of use. It may be memory. For example, in order to recognize the numbers 0 to 9, the digitized sound patterns of each of the numbers 0 to 9 are stored in the recognition word memory 15. The collation device 14 collates the voice input digitized by the A/D converter 13 with the data in the recognition word memory 15, and stores the similarity result in the recognition result sequential rewriting memory 16. The recognition result of the pronounced voice with the highest degree of similarity is output.

Problems to be Solved by the Invention However, the above configuration has a problem in that errors in speech recognition cannot be corrected. It is extremely difficult to achieve a 100% recognition rate in speech recognition, and a recognition device is required with a certain error rate in mind.

Some words, such as 9 (kyu) and 10 (shuu), have extremely similar pronunciations, and depending on the user's pronunciation habits, the recognition result may differ from what has been pronounced many times. It has a certain problem.

In view of the above-mentioned problems, the present invention has a means for outputting a highly similar word different from the first recognized word as a recognition result when multiple voices are recognized, thereby preventing the same recognition error from occurring over and over again. To provide a voice recognition device that is easy to use and does not repeat the process.

Means for Solving the Problems In order to solve the above problems, the speech recognition device of the present invention determines the degree of similarity depending on the degree of matching between input speech and a recognition word memory in which a plurality of recognition words are registered. A sequential memory means for storing, a recognition result output means for outputting a recognized word with the highest similarity in the first recognition, and a recognition result output means in the nth recognition.
The recognition result output means outputs a recognition word with the highest degree of similarity other than the recognition words already output as recognition results, and the recognition result output means includes a reset means for returning to the first recognition.

Effects The present invention has the above-described configuration, and has a sequential memory means for storing the degree of similarity according to the degree of matching between the input speech and the recognition word memory in which a plurality of recognition words are registered, In the n-th recognition, the recognition words with the highest degree of similarity are sequentially output from the memory, and in the n-th recognition, the recognition words with the highest degree of similarity other than the recognition words that have already been output as recognition results can be output. Furthermore, when starting the recognition from - again, the reset means for returning to the first recognition can be used to return to the start of the recognition.

Embodiment Hereinafter, a speech recognition device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a speech recognition device according to an embodiment of the present invention.

In FIG. 1, 1 is a microphone that converts audio into electrical signals. 2 is an amplifier. 3 is an analog-to-digital converter (hereinafter referred to as an A-D converter). 4
is a matcher. 5 is a recognition word memory in which patterns of words to be recognized are registered. 6 is a sequentially rewritten memory that stores recognition results. 7 is a recognition button for starting recognition. 8 is a control unit that controls recognition. 9 is a switch that outputs the result with the highest recognition similarity. 10 is a register that temporarily holds the result with the highest degree of recognition similarity.

11 is a comparator. 12 is a recognition result output changeover switch.

The operation of the speech recognition device configured as described above will be explained below.

First, when the recognition button 7 is operated to perform voice recognition, the control section 8 controls the collation device 4 to start voice recognition. The voice produced is converted into an electrical signal by a microphone 1 and amplified by an amplifier 2 to the amplitude necessary for verification. The amplified analog electrical signal is converted into a digital signal by an A-D converter 3. The recognition word memory 5 registers the pattern of the word to be recognized as a digital signal, and may be a read-only memory that is pre-registered, or can be registered by the user at the time of use. It may be a sequentially rewritten memory. For example, when the numbers 0 to 9 are to be recognized, the recognition word memory 5 stores digitized speech patterns for each of the numbers 0 to 9. The collation device 4 collates the voice input digitized by the A-D converter 3 with the data in the recognition word memory 5, and stores the similarity results in the recognition result memory 6.
Store in. The data are encoded and stored in order from memory 1 of the rewriting memory 6 in order of highest similarity.

The control unit 8 determines whether this is the first speech recognition based on the recognition end signal from the collation device 4, and if it is the first recognition, sets the output switch 12 to a, and selects the recognition with the highest degree of similarity stored in the memory 1. Output the results from the recognition device. Also,
The switch 9 is closed, the data of the memory 1 is stored in the register 10, and the switch 9 is opened again.

Here, when voice recognition is performed again and the second recognition result is registered in the sequential rewriting memory 6 in the same way as in the above process, the control unit 8 detects that it is the second recognition and starts the comparator. 11 is activated. The first recognition result is in register 10
If the word is different from the recognized word with the highest degree of similarity in the second recognition, that is, the content of memory 1 of the sequential rewriting memory 6, the output switch 12 is turned to a. If the comparison results are the same, the output switch 12
By changing b to b, it is possible to output a recognition result with a high degree of similarity that is different from the first recognition result. This operation can be repeated n times by increasing the number of registers 10 and registering a plurality of past recognition results. Further, in order to return to the initial stage, it is possible to set a limit on the number of times in the control section 8 or by pressing the switch 7.

Effects of the Invention As described above, according to the present invention, there is provided a sequential memory means for storing the degree of similarity according to the degree of matching between input speech and a recognition word memory in which a plurality of recognition words are registered; a recognition result output means for outputting a recognized word with the highest degree of similarity for the first recognition; and a recognition result output means for outputting a recognized word of the highest degree of similarity for the n-th recognition, other than the recognition word that has already been output as a recognition result. By having a configuration including a reset means for returning to the first recognition, it is possible to obtain an excellent effect that an error in the same recognition result will not be output when rewording is performed.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a timer reservation device according to an embodiment of the present invention, and FIG. 2 is a block diagram of a conventional timer reservation device. 1...Microphone, 2...Amplifier, 3...A-D converter, 4...Verifier, 5...Recognized word Memory 6... Recognition result memory 7... Recognition switch, 8...
...Control unit, 9...Switch, 10...
...Register, 11...Comparator, 12...
... Out of stock.

Claims (1)

[Claims] A speech recognition device for recognizing speech of a plurality of words, comprising a sequential memory means for storing a degree of similarity according to a degree of matching between an input speech and a recognition word memory in which a plurality of recognized words are registered; , a recognition result output means that outputs the recognized word with the highest degree of similarity in the first recognition, and a recognition result output means which outputs the recognized word with the highest degree of similarity in the n-th recognition, other than the recognition word that has already been output as the recognition result. output means;
A speech recognition device comprising a reset means for returning to the first recognition.