JPS6360484A - Enunciation training machine - Google Patents

Abstract

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

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a voice training device that listens to an audio output conversation, for example, in a foreign language, and repeats the conversation to practice pronunciation.

[Conventional technology]

2. Description of the Related Art Hitherto, various types of vocal practice devices, particularly devices for self-studying the pronunciation of foreign languages, have been put into practical use.

The pronunciation practice process for these models is to first play back model data recorded on magnetic tape, etc., and then LL learning, in which the pronunciation practitioner repeats the pronunciation in the same way while listening to the audio data, has become widespread. Many software are on sale.

[Problem that the invention seeks to solve]

However, this kind of LL learning is a simple learning process in which you listen to the teacher's pronunciation recorded on tape and simply repeat it, so you do not know how accurately the teacher's pronunciation and your own pronunciation are pronounced. Since it is not possible to determine whether or not the learning results have been achieved, it is difficult to see learning results, and there are problems in that learning effects tend to vary from person to person. In addition, although foreigners' desire to learn Japanese has been rapidly increasing recently, there are currently only a few instructors who can teach accurate Japanese pronunciation, which is causing dissatisfaction among foreigners. There is a strong demand for the development of such training equipment.

This invention was made to solve the above problems, and by analyzing voice data uttered from different voice sources and comparing the voice patterns, the degree of similarity between the two is determined, and accurate The purpose of the present invention is to obtain a vocal training device that can notify learners of learning effects.

[Means for solving problems]

The voice training device according to the present invention includes a voice input means for inputting first voice data serving as a model and second voice data uttered by a voice practitioner based on the first voice data; The first and second input from the input means
a voice data storage means for storing voice data; and a voice data storage means for reading out first and second voice data stored in the voice data storage means, analyzing the frequency spectrum of each voice data, and extracting voice characteristic patterns, respectively. a pattern extracting means for determining the similarity, a determining means for comparing the respective audio feature patterns extracted by the pattern extracting means to determine the degree of similarity, and a display means for displaying the degree of similarity determined by the determining means in stages. It was established.

[Effect]

In this invention, the pattern extracting means receives the first pattern input from the voice input means and stores the voice data in the voice data storage means.
When the second audio data is read out and the frequency spectrum of each audio data is analyzed to extract each audio feature pattern, the determining means compares each extracted audio feature pattern to determine the degree of similarity, and determines the similarity. Display the results on the display means.

〔Example〕

FIG. 1 is a block diagram illustrating an example of a vocal practice machine showing an embodiment of the present invention. Reference numeral 1 denotes a microphone serving as a voice input means, and a learning tape or foreign language output from a radio cassette tape recorder (to be described later) is used. Broadcasts (first audio data) and physical sounds uttered by a vocal practitioner (second audio data) are collected. Reference numeral 2 denotes an input switching unit, which selects the first and second audio data collected by the microphone 1 using a switch 2a.
Switch by. 3 is a voice memory, which is composed of a model voice memory 3a, a practice voice memory 3b, etc.
The model audio memory 3a stores first audio data collected from the microphone 1, and the practice audio memory 3b stores second audio data collected from the microphone 1. Reference numeral 4 denotes a feature extraction section, which is composed of a model speech feature extraction section 4a, a practice speech feature extraction section 4b, etc., and the model speech feature extraction section 4a extracts the first speech data stored in the model speech memory 3a. After reading out and performing fast Fourier transform processing and analyzing the frequency spectrum of the first voice data to obtain a frequency power spectrum, a cepstrum that is effective for vowel voice correspondence is extracted, and the practice voice feature extraction unit 4b The second voice data stored in the practice voice memory 3b is read out, fast Fourier transform processing is performed, and the frequency spectrum of the second voice data is analyzed to obtain a frequency power spectrum, which is effective for vowel voice correspondence. Extract the cepstrum.

Reference numeral 5 denotes a pattern matching processing unit that processes the patterns extracted from the model speech feature extraction unit 4a, , and the learned speech feature extraction unit 4b.
The distance between the cepstrum feature parameters of the second audio data and the second audio data is determined while aligning the time axis, and pattern similarity data between the two is output to the similarity determination section 6 based on this distance. The similarity determination unit 6 is a pattern matching processing unit 5.
The pattern similarity data outputted from is evaluated in three stages, such as "Good pronunciation,""A little more," and "Let's practice many times." indicate.

FIG. 2 is a schematic diagram showing an example of a pronunciation training form using the vocal training machine shown in FIG. 1, and the same parts as in FIG. 1 are given the same reference numerals.

In this figure, reference numeral 11 denotes a radio cassette tape recorder (RCTR) that serves as a first audio data source, and outputs audio from a speaker of a commercially available learning conversation tape or foreign language learning program that serves as the first audio data. 12 is the second
A practitioner who serves as a source of voice data listens to the first voice data outputted from the RCTRII and utters a real voice that becomes the second voice data.

Next, the pronunciation practice evaluation control operation according to the present invention will be explained based on the flowchart shown in FIG.

FIG. 3 is a flowchart illustrating the pronunciation practice evaluation control operation according to the present invention. In addition, (1) to (9)
indicates each step.

By operating the changeover switch 2a, the first and second audio data collected by the microphone 1 are stored in the model audio memory 3a, and the second audio data is stored in the practice audio memory. Store (1). Next, the model speech feature extraction unit 4a reads out the first speech data stored in the model speech memory 3a2), performs fast Fourier transform processing to obtain the frequency power spectrum, and then calculates the vowel speech correspondence. Perform feature extraction to extract the effective cepstrum and output parameters to the pattern matching processing unit 5 (3) 0 Next, read out the second audio data stored in the training audio memory 3b 4) Fast Fourier After performing the conversion process and obtaining the frequency power spectrum,
Feature extraction is performed by extracting a cepstrum that is effective for vowel speech correspondence and outputting parameters to the pattern matching processing section 5 (5).Subsequently, the pattern matching processing section 5 extracts the model speech feature extraction section 4a, ! ! Pattern matching is performed to find the distance #D of the cepstrum feature parameters of the first and second audio data extracted from the learned speech feature extraction unit 4b while aligning the time axis (B), and by this distance #D, both output the pattern similarity data (distance ftD) to the similarity determination unit 6 (7) 0 Then, the similarity determination unit 6
．． The parameter distance @D obtained by the pattern matching processing section 5 is converted into predetermined distance data Nl, N2.
8), and the indicators 7a to 7 are displayed based on the difference.
Light up one of C (9).

In addition, in the above embodiment, the case where the pronunciation similarity is evaluated in three stages using the 1 display units 7a to 7C was explained.
By increasing the number of evaluations, learners can accurately grasp their learning progress.

〔Effect of the invention〕

As explained above, the present invention includes a voice input means for inputting first voice data serving as a model and voice data of $2 uttered by a vocal practitioner based on the first voice data, and A voice data storage means for storing first and second voice data inputted from the input means; and a frequency of each voice data by reading out the first and second voice data stored in the voice data storage means. A pattern extracting means for analyzing the spectrum and extracting each audio feature pattern, a determining means for comparing the audio feature patterns respectively extracted by the pattern extracting means and determining the degree of similarity, and the similarity determined by the determining means. Since we have provided a display means to display the degree in stages,
In addition to being able to practice pronunciation using commercially available learning tapes or broadcast foreign language courses as models, it also accurately displays the pronunciation status of the model and the practitioner, thereby increasing the practitioner's motivation to learn. It has the advantage of greatly promoting learning and exhibiting extremely excellent learning effects.

[Brief explanation of the drawing]

FIG. 1 is a block diagram illustrating an example of a vocal training machine according to an embodiment of the present invention, FIG. 2 is a schematic diagram showing an example of a pronunciation training form using the vocal training machine shown in FIG. The figure is a flowchart illustrating the pronunciation practice evaluation control operation according to the present invention. In the figure, 1 is a microphone, 2 is an input switching unit, 3 is an audio memory,
4 is a feature extraction section, 5 is a pattern matching section, 6 is a similarity determination section, and 7a to 7C are indicators.

Claims (1)

[Claims] a voice input means for inputting first voice data serving as a model and second voice data uttered by a voice practitioner based on the first voice data; A voice data storage means for storing second voice data, and a voice characteristic pattern is determined by reading the first and second voice data stored in the voice data storage means and analyzing the frequency spectrum of each voice data. A pattern extracting means for extracting the respective patterns, a determining means for comparing the audio feature patterns respectively extracted by the pattern extracting means to determine the degree of similarity, and a display means for displaying the degree of similarity determined by the determining means in stages. A vocal training machine characterized by comprising the following.