JPS59124395A - 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] [Technical Field of the Invention] The present invention examines whether the level of input speech is appropriate based on the parameters of the input speech, and if the level is outside the appropriate level range, the speaker The present invention relates to a speech recognition device that is designed to call attention to the input level.

[Prior art and problems]

FIG. 1 is a diagram showing an example of a conventional configuration of a speech recognition device.

In FIG. 1, 1 is an audio input section, 2 is an input parameter extraction section, 3 is a level detector, 4 is a matching section, 5 is a dictionary, and 6 is a host processor.

In FIG. 1, the input parameter extraction section 2 is the voice input section 1.
The dictionary 5 is used to extract feature amounts, so-called parameters, of the input speech from the input voice, and the dictionary 5 is used to store parameters.

The matching unit 4 compares the parameters of the input voice extracted by the input parameter extraction unit 2 with the parameters stored in the dictionary 5, and notifies the host processor 6 of the most matching string as a recognized character string. It is. The level detector 3 detects and displays the peak value of the input audio from the audio input section 1. In such conventional speech recognition devices, it is necessary to convert analog amounts inputted into voice into digital amounts for internal processing. Since the digital quantity is also represented by a finite value, there is always a risk of clipping or overflow.

For this purpose, a level detector 3 is provided, and normally notifies the speaker of the level so that the volume of the input voice is appropriate.

However, in cases where voice input is performed from a location far from the voice recognition device, for example, it is possible to pay attention to the level when registering, such as when creating a dictionary, but during actual work, the level of the input voice is It was not possible to know whether the input voice was appropriate or not using the level detector 3, and there was no way for the speaker to obtain information about the level of the input voice.

[Purpose of the invention]

The present invention is based on the above considerations, and provides a speech recognition device that can notify the speaker of the level even when inputting speech at a location away from the speech recognition device. The purpose is to

[Structure of the invention]

To this end, the speech recognition device of the present invention includes a speech input section, an input parameter extraction section that extracts the parameters of the input speech from the speech input section, a dictionary that stores the parameters, and an input parameter extraction section that stores the parameters. A speech recognition device includes a matching section that compares speech parameters with parameters stored in the dictionary and sends the closest recognized character string to a host processor. The time when the power exceeds a certain value is stored as part of the parameter [7. The matching section calculates the power of the input audio from the parameters extracted by the input parameter extraction section, and determines whether the power is within the above-mentioned constant value. The time exceeding the value is counted, and the time is compared with the time stored in the dictionary to check whether the input audio power level is appropriate. If it is determined that it is not appropriate, the notification method is used to notify the user. This feature is characterized in that it is configured to notify the user.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a diagram showing the configuration of one embodiment of the present invention, FIG. 3 is a diagram showing an example of the relationship between time and audio power, and FIG. 4 is a diagram showing an example of the configuration of a parameter section of a dictionary.

In FIG. 2, numerals 6 to 6 indicate those corresponding to those in FIG. 1, 7 indicates an audio output device, and 8 indicates a speaker.

In FIG. 2, the dictionary 5 contains a certain value PFH for the voice power parameter as shown in FIG.
The time Tott exceeding Tott is stored as part of the parameters. Assuming that the configuration of the variation part in the conventional dictionary 5 is as shown in FIG. 4(A), the dictionary 5 shown in FIG. The input time Tov exceeding a certain constant value PWH is stored in the parameter section as part of the parameters. Then, the matching section 4 starts counting the time when the power of the input voice exceeds a certain value Pry, and continues counting the time until the point when the power of the input voice exceeds Prrr. Here, the parameter F of the input audio extracted by the input parameter extraction unit 2 is, for example, an output value obtained by frequency analysis of the input audio sampled at a certain fixed time interval, and therefore a value represented by the sum of the values, for example, 16 channels. If the parameter consists of, ΣFi becomes the power I++Q−PF. Therefore, in the matching section 4, the sum of the parameter values of the audio input extracted by the input parameter extraction section 2 is taken as the power, and the time when the power exceeds a certain value PWH is counted, and the count value T6F is calculated. In the recognition operation, the ToV of the recognized word in the dictionary is compared. As a result, if TaV is within a certain range, the audio input level is determined to be normal, and if it is outside a certain range, a comment is sent from the speaker 8 to the speaker through the audio output device 7. . Audio output device 7
The device stores messages in itself, and outputs audio by specifying a value called a voice number associated with the message.

Usually a speech synthesizer is used.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, it is possible to determine whether or not the level of input audio is appropriate based on the output of the input parameter extracting section, rather than relying on a level detector that normally displays based on the peak value of input audio. and send appropriate comments to the speaker through the audio output device.
Even when voice input is performed at a location away from the voice recognition device, recognition processing can always be performed with an appropriate level of input, and the recognition rate can be improved.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an example of the conventional configuration of a speech recognition device, FIG. 2 is a diagram showing the configuration of one embodiment of the present invention, FIG. 3 is a diagram showing an example of the relationship between time and voice power, and FIG. The figure shows an example of the configuration of a parameter section of a dictionary. 1... Audio input section, 2... Input parameter extraction section, 3
...Level detector, 4...Matching section, 5...
Dictionary, 6... host processor, 7... audio output device! :, 8...Speaker. Patent applicant: Soshitsu Co., Ltd. Representative patent attorney: Yotsube Kyotani

Claims (1)

[Claims] a voice input section, an input frameta extraction section for extracting the frameta of the input voice from the voice input section, a dictionary for accumulating the frameta, and the above (of the input voice extracted by the frameta extraction section). In the speech recognition device, the speech recognition device includes a matching unit that compares the character string with the parameters stored in the dictionary and sends the closest recognized character string to the host processor. The matching section calculates the power of the input voice from the parameter extracted by the above-described parameter extraction section, and calculates the power of the input audio from the parameter extracted by the parameter extraction section. Count the time when the noise level exceeds the certain value, compare the time with the time stored in the dictionary, check whether the noise level of the input voice is appropriate, and determine that it is not appropriate. A voice recognition device characterized in that the voice recognition device is configured to notify by the notification means when the notification is made.