JPS60104999A - Voice recognition equipment - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Industrial applications The present invention relates to a speech recognition device targeted at a specific speaker.

Conventional configuration and its problems In a word speech recognition device targeted at a specific speaker, the word speech recognition device that is targeted at a specific speaker matches the pre-stored word speech sounds with the input speech sounds, and selects the one with the highest degree of similarity as the recognition result. This method is generally used. However, in places where there is a lot of ambient noise, the noise mixes with the input voice and becomes a major cause of misrecognition. To this end, the ambient noise input is divided into each frequency band and stored, and the noise energy is subtracted from the energy of each frequency band of the input audio to obtain the audio signal, thereby improving the recognition rate in noise. There is a way.

However, in high noise, by subtracting the noise signal,
The input audio signal itself is also subject to deformation, leading to erroneous recognition. In particular, when a voice recognition device is used in a place where reliability is required, how to suppress misrecognition becomes a problem.

Purpose of the Invention In order to solve the above-mentioned problems, the present invention provides a speech recognition device that can suppress misrecognition as much as possible by first analyzing the surrounding noise situation and automatically determining whether recognition is possible. This is what we provide.

Structure of the Invention The present invention includes a storage means for storing a time series of feature vectors of each word sound to be recognized as a standard button, and detects whether or not a person has started a recognition device using a switch or the like, and starts processing. It has a detection means, a noise analysis means for determining whether the surrounding noise is at a level that allows accurate recognition, and a time monitoring means for determining whether voice is input within a predetermined time. When input, the recognition means matches the input voice with the standard bang stored in the storage means, and the result is outputted as a voice by the voice synthesis means. FIG. 1 is a block diagram showing an embodiment of the present invention. 1
is a detection means that detects whether information to start recognition processing has been received and starts the next processing. 2 is a noise analysis means that takes in ambient noise and determines whether the noise energy is at a sufficiently low level to perform recognition. If the recognition is possible, the display lamp 6 is turned on. Reference numeral 3 denotes a time monitoring means that determines whether or not a voice has been input within a certain period of time after the display lamp 6 is turned on, and if there is no input within that time, the subsequent processing is stopped. Reference numeral 4 denotes a recognition means that matches the standard bangs stored in the storage means 7 with the input voice and outputs the standard bangs with the highest degree of similarity as a recognition result.

This is a voice synthesis means that is controlled by the output of the 5id recognition means 4 and utters the recognition result.

FIG. 2 is a circuit diagram showing a specific configuration of this embodiment.

1o is a microcomputer for realizing each of the means 2.3 and 4.7 in FIG. An arithmetic control unit 1oC and an input unit 1 that analyze ambient noise and perform recognition processing of input speech.
It is configured isometrically by the 0a1 output section 10d.

8 is a microphone for inputting voice and collecting ambient noise; 9 is an input signal input from the microphone 8;
/ This is an A/D converter that performs D conversion and extracts feature parameters. 13 is a switch for starting recognition processing. It is also possible to replace this switch 13 with an infrared sensor and operate the recognition device by detecting the presence of a person when he or she is in a predetermined location. Reference numeral 12 denotes a recognition result display that outputs the recognition result in characters.

FIG. 3 is a flowchart 1 of the main part for explaining the operation of the microcomputer of this embodiment.

The operation of this embodiment with the above configuration will be explained in detail along the flowchart of FIG.

The speech recognition device according to this embodiment is started by pressing the switch 13 when a person wants to operate the recognition device. That is, first, in step 14, the switch 13 is
It monitors whether or not it is pressed, and when it is detected that it has been pressed, the output signal enters the input section 10b of the microcomputer 1o and opens a recognition window. After this step 15
Performs sound collection analysis of noise. In this case, ambient noise is input from the microphone 8, and the noise signal which is A/D converted by the A/D converter 9 is sent to the microcomputer 1.
o, and the energy level of this noise is compared with a threshold value (a noise level that is sufficient to perform the subsequent recognition processing) on the calculation side [10c, and step 1
Judging by 6.

Here, if the noise level is greater than the threshold value, it is determined that recognition is difficult, and the process returns to step 14. If it is determined in step 16 that recognition is possible, the display lamp 6 (for example, a light emitting diode) is turned on in step 17, and information to permit the person to input voice is sent. Start dividing. For the input signal from the microphone 8, a voice section is extracted in step 19, and it is determined in step 2o whether the signal is actually a voice signal, and if it is not a voice signal, the time is counted in step 21. Check whether the sum exceeds a certain time length tl, and if it does not, return to step 19 again υ and continue the same process. If it exceeds t, stop the recognition process and turn off the display lamp ei in step 24. The judgment in step 20 as to whether or not the input signal is speech depends on whether or not the word-initial stationary part of the input signal data has a formant structure, and if so, it is determined that it is speech. In addition, in order to extract a speech section from the input signal in step 19, it is necessary to have a silent section of a certain length or more before and after the beginning of the speech word, and if the length of the speech is within a certain fixed time length, A portion that satisfies the condition that the maximum energy within the section is greater than or equal to a certain threshold is regarded as a voice section.

When the voice is input, the recognition process in step 22 aligns the input voice with the standard bang stored in the storage unit 10bK to obtain the recognition result with the highest degree of similarity.

This recognition result is outputted from the microcomputer 10 to drive the speech synthesis means 6 and the speech memory 11 in step 23.
The synthesized voice is uttered. Furthermore, this recognition result is displayed on the recognition result display 12. When these series of recognition processes are completed, a control signal is sent from the output section 10d of the microcomputer 10 to the display lamp 6, and the f'LOFF
state, the recognition window closes, and recognition ends.

According to the configuration of the above-described embodiment, by not accepting recognition because recognition cannot be performed accurately when the surrounding noise level is high, it is possible to prevent erroneous recognition caused by forced recognition processing under high noise. Furthermore, by limiting the time for voice input by monitoring a timer, the present invention is useful for preventing misrecognition caused by voices of speakers other than the target speaker. It has a noise analysis means that automatically analyzes the surrounding noise state when you press , and interrupts the recognition process if the noise level is higher than a certain threshold. By having a time monitoring means that interrupts recognition even when no voice input is made, it is possible to provide a voice recognition device that can prevent factors that cause erroneous recognition and suppress erroneous recognition as much as possible.

Further, voice input can be prompted by a display lamp, and the recognition result can be confirmed with the ears by the voice synthesizing means, making it possible to provide a voice recognition device that is very easy for the speaker to use.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram showing an embodiment of the present invention, FIG. 2 is a circuit diagram showing a specific configuration, and FIG. 3 is a flowchart of main parts for explaining the operation. 1...detection means, 2...noise analysis means,
3. Time monitoring means, 4. Recognition means, 6.
...Speech synthesis means, 7...Storage means, 8...
...Microphone, 9...A/D converter, 10
・・・・・・Microcomputer, 11・・・・・・
voice memory. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure 3

Claims (1)

[Claims] A detection means for detecting information for starting a recognition process, and after the recognition process is started by the detection means,
A noise analysis means that determines whether the ambient noise level is low enough to perform accurate speech recognition and rejects recognition if the noise level is lower than a certain threshold, and the noise analysis means can recognize it. When it is determined that this is the case, the input voice is recognized using a display lamp that prompts voice input and a timer that measures the time to stop recognition if voice is not input within a certain period of time. A speech recognition device comprising a recognition means for deriving a result, and a speech synthesis means for outputting the recognition result as a voice.