JPH02103599A - Voice recognizing device - Google Patents

Abstract

PURPOSE:To stabilize utterance and to improve a recognition rate by stopping the output of a voice response section when the voice recognized during voice response is detected in a voice section detecting section. CONSTITUTION:The voice inputted from a microphone 1 is inputted to a characteristic extracting section 2 and the voice section detecting section 3. The characteristic quantity is extracted from the voice inputted in the characteristic extracting section 2 and the voice section is detected from the inputted voice in the voice section detecting section 3. The characteristic quantity in the voice section is compared with a voice recognition dictionary 5 in a recognition section 4 and the most analogous word of the dictionary is determined as a correct answer. On the other hand, the voice is outputted from the voice response data by the command of a voice control section 6 in the voice response section 8. This device is so constituted as to stop the output of the voice response section 8 when the voice recognized during the voice response is detected in the voice section detecting section 3. The high recognition rate is obtd. in this way even if the user begins to start utterance during the response voice output.

Description

[Detailed description of the invention] division of powers The present invention relates to a speech recognition device.

Voice response is most commonly used as a means for the user to confirm the recognition results of the voice recognition device. Additionally, voice responses are used as guidance for systems using recognition devices. Their usage is as follows. For example, in the case of guidance, the user will utter a voice in accordance with the guidance after the output of the voice response unit is finished. In the case of a voice response as an output of a voice recognition result, the next utterance is made after confirming the output of the voice response section.

However, as the user gets used to it, he or she will start speaking without listening to the output of the voice response section until the end. This is because if the output of the voice response unit is listened to until the end, the user must wait for the end of the response voice, which reduces the number of times the voice can be uttered and slows down data input by voice. In this case, the recognition section can operate independently of the voice response section, so recognition can begin.

However, the user must speak while listening to the output of the voice response section. The output of the voice response is Hera 1-Seno 1~
Since it is output from a device such as a computer or a handset, it is output close to the user's ear. In general, the volume of a speaker's voice and the manner in which he or she speaks changes depending on whether the speaker speaks in a quiet place or in a noisy place. When speaking while listening to the output of the voice response unit, the environment is similar to when the surroundings are noisy, and the disadvantage is that the speech becomes unstable and the speaking sound If becomes louder, resulting in a poor recognition rate. . Furthermore, it was unpleasant for the user to have to listen to the output of the voice response unit while he or she was speaking.

The present invention was made in view of the above-mentioned circumstances.
In particular, the purpose of this invention is to provide a voice recognition device that can obtain a high recognition rate even if the user starts speaking while outputting a response voice.

In order to achieve the above object, the present invention provides a microphone that picks up voices, a feature-animal extractor that extracts voice features, a voice section detection means that detects voice sections, and a voice recognition system. a dictionary, a recognition unit that compares the input speech with a speech recognition 3il dictionary and outputs the dictionary that is most similar as the correct answer;
In a voice recognition device that includes a voice response section, a voice response data section, a voice output section, a response control section that controls the operation of the voice response section, and a timer section, the voice to be recognized during the voice response is a voice. It is characterized in that the output of the voice response section is stopped when the section detection section detects the detection. Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, in which 1 is a microphone, 2 is a feature extraction section, 3 is a speech section detection section, 4 is a recognition section, 5 is a speech dictionary, 6 is a response control section, 7 is a voice response data section, 8 is a voice response section, 9 is a voice output section, 10 is a speaker, and the voice input from the microphone 1 is passed through a feature extraction section 2 and a voice section 171j inspection, '4
3 Part 3 is input.

The feature extractor 2 extracts feature amounts from the input voice. The voice section detecting section 3 detects a voice section from the input voice. The feature Fat within the speech section is compared with the speech recognition dictionary 5 in the recognition unit 4, and the most similar word in the dictionary is determined as the correct answer. on the other hand.

The voice response section 8 outputs voice from the voice response data 7 in response to a command from the response control section 6 .

Next, the relationship between the response output operation and the recognition operation will be explained, but here, the case of recognition operation with guidance will be explained. At the beginning of the operation, the response control section 6 gives a command to output guidance to the voice response section 8 (command signal b to the voice response section). In response to this command, the voice response unit 8 outputs guidance as voice. At the same time, the recognition unit 4 starts detecting voice sections (voice section detection 4
No. 8). Here, when the user utters a voice after outputting a voice response, it is sufficient to perform a normal recognition operation. However, if the user speaks during the voice response operation, the following operation is performed. Since the speech section detection unit 3 is already capable of detecting speech sections, the speech section of the uttered voice is detected. The response control section 6 detects that the voice section has been detected and stops the operation of the voice response section 8. This allows the speaker to speak normally without hearing the voice response output. When the next recognition is performed, the recognition result is generally output from the voice response unit 8 in order to notify the user of the recognition result. Here, the user may be able to tell whether the recognition result is correct or incorrect without listening to the entire response output. In that case, make the next utterance as soon as you know. This is because data input by voice becomes faster.

Even in this state, it is possible to detect a voice section if the recognition result is immediately recognized, so if a voice section is detected, the response control unit 6 stops outputting the response, and the user can produce correct vocalizations.

By the way, when performing the above operation, the recognition unit must always be in a state of waiting for utterance. However, if noise is picked up from the microphone, the response output may be stopped by mistake, which is inconvenient. Therefore, as shown in Figure 2 (a), after a voice section is detected (time point t in the figure), the voice response output is stopped only if it continues for a certain period of time (T in the figure). do. Appropriately, this value T is approximately 150 ms. This is because there are generally no words shorter than that, and noise is often shorter than that.

FIG. 3 is a block diagram for explaining one embodiment of the speech recognition device according to the present invention that performs the above-mentioned operation.
Reference numeral 1 designates a timer, and other parts having the same functions as those in the embodiment shown in FIG. 1 are given the same reference numerals as in FIG. Similarly to the operation of the embodiment shown in FIG. 1, when a voice section is detected while outputting a voice response, the response control section detects the voice section continuously for a certain period of time based on a signal from the timer. detect whether or not the If the voice section signal is less than a certain time T as shown in FIG. 2(b), the output of the voice response section is maintained. On the other hand, if the length becomes longer than T as shown in FIG. 2 (,), the output of the response section is stopped (at the point L2 in the figure). By doing this, the response output will not stop due to noise.

As is clear from the above explanation, in the voice recognition device of claim 1, if the user starts speaking during a voice response, the response output is stopped, so the voice becomes stable and recognition is improved. rate increases. Furthermore, in the speech recognition device according to the second aspect of the present invention, there is no erroneous stoppage of the speech response due to surrounding noise, and the user can hear stable speech response output.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining one embodiment of a speech recognition device according to the present invention, FIG. 2 is a time chart for explaining the operation of the present invention, and FIG. FIG. 2 is a configuration diagram for explaining an example. 1...Microphone, 2...Feature extraction unit, 3...
- Voice section detection unit, 4... Recognition unit, 5... Voice response unit, 6... Response control unit, 7... Voice response data unit, 8
...Voice response section, 9..Speech output section, 10..Speaker, 11..Timer section. Figure Figure Figure Figure

Claims (1)

[Scope of Claims] 1. A microphone that picks up voices, a feature amount extraction unit that extracts voice features, a voice section detection means that detects voice sections, a voice recognition dictionary, and a voice recognition system for input voice. a recognition unit that compares with the dictionary and outputs the most similar dictionary as the correct answer; a voice response unit; a voice response data unit; a voice output unit; a response control unit that controls the operation of the voice response unit;
What is claimed is: 1. A voice recognition device having a timer unit, characterized in that when a voice to be recognized during a voice response is detected by a voice section detection unit, the output of the voice response unit is stopped. 2. A microphone that picks up the voice, a feature extractor that extracts the features of the voice, a voice section detector that detects the voice section, a voice recognition dictionary, and compares the input voice with the voice recognition dictionary to find the most similar one. A voice recognition device comprising: a recognition unit that outputs a dictionary that is correct as a correct answer, a voice response unit, a voice response data unit, a voice output unit, a response control unit that controls the operation of the voice response unit, and a timer unit, A voice recognition device characterized in that when a voice to be recognized during a voice response is continuously detected by a voice section detection unit for a certain period of time, the output of the voice response unit is stopped.