JP2000250571A - Method and device for editing speech data, and medium recorded with program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make a functional word, etc., included in speach easy to listen. SOLUTION: An analog speech signal is converted into a series of speech data by an A/D converter 21, and a series of speech data are divided into the word data by a word blocking means 23, and the word data whose signal level is lower than a prescribed threshold value R among the divided word data are extracted by a word data extraction means 24. The signal level of the extracted word data is amplified by an amplifier means 25, and the amplified word data and the other data are synthesized in order of input speech by a synthetic means 26, converted into the analog speech signal, and outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technology for editing a voice so that it is easy to hear.

[0002]

2. Description of the Related Art For example, as a method of learning a foreign language,
The LL learning method of actually learning by listening to the foreign language by ear is effective.

In such LL learning, a method is used in which a foreign language voice is output from a master unit used by a teacher or the like to a plurality of slave units for students, and each student listens through a headphone. I have.

[0004] However, among a plurality of students, not only those who are accustomed to the actual speed of a foreign language but also those who are not accustomed to it at all, such beginners are sent from the parent machine. In many cases, the incoming voice cannot be understood.

[0005] In order to solve this, it is conceivable to convert the input voice into a slow voice using a speech speed conversion technique.

[0006]

However, simply converting the voice to a slow speed, for example, in the case of English, such as prepositions and articles such as for, to, the, etc. (these are called functional words) There is a problem that words that are short and weakly pronounced cannot be clearly recognized.

An object of the present invention is to solve this problem and to provide a voice data editing method, a voice data editing device, and a medium on which a program is recorded, which makes it easy to hear function words contained in voice.

[0008]

According to a first aspect of the present invention, there is provided a voice data editing method comprising: converting a series of voice data obtained by converting an analog voice signal into digital data into word data; Segmenting, amplifying word data having a signal level lower than a predetermined threshold value among the segmented word data, and synthesizing the amplified word data and other word data in the original voice order. And

According to a second aspect of the present invention, there is provided an audio data editing apparatus for converting an input analog audio signal into digital audio data, and an output from the A / D converter. Word blocking means for dividing a series of voice data into word data; word data extracting means for extracting word data whose signal level is lower than a predetermined threshold value among word data divided by the word blocking means; Amplifying means for amplifying the signal level of the word data extracted by the word data extracting means; and word data not extracted by the word data extracting means among the word data segmented by the word blocking means. Synthesizing means for synthesizing the word data whose signal level has been amplified by the amplifying means in the order of the input voice; and And a D / A converter audio data and outputs the converted analog audio signal synthesized me.

[0010] Further, a medium on which the program according to claim 3 of the present invention is recorded is a computer, comprising: a computer configured to divide a series of voice data into word data; Word data extracting means for extracting word data having a signal level lower than a predetermined threshold value; amplifying means for amplifying a signal level of the word data extracted by the word data extracting means; and the word blocking means Therefore, a program for functioning as synthesizing means for synthesizing word data not extracted by the word data extracting means among the word data divided and word data whose signal level has been amplified by the amplifying means in the order of input voices. Is recorded.

[0011]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an audio editing device 2 according to the embodiment.
FIG. 2 shows an example of the internal configuration of the main part.

In FIG. 1, an A / D converter 21 samples an audio signal input to an input terminal 20a at a predetermined cycle, converts the audio signal into digital audio data, and sequentially stores the digital audio data in an input buffer 22.

The word blocker 23 divides a series of voice data stored in the input buffer 22 into word data and silence data.

For example, an audio pitch is extracted from a series of audio data stored in the input buffer 22 using an autocorrelation method, a zero-crossing counting method, or the like, and a series of audio data is unvoiced based on the extracted pitch. It is divided into a section, a voiced section, and a silent section, and a data string of a continuous voiced part composed of a combination of the unvoiced section and the voiced section is defined as word data W without sandwiching the silent section, and a data string of a silent section between the word data is defined. Let it be silence data S.

Here, the unvoiced section refers to, for example, a high-frequency part (consonant part) Va at the beginning of the sound in the signal of one sound shown in FIG. 3, and the voiced section refers to a low-frequency part Vb.
(A vowel portion), and a silent section indicates a portion N where there is no sound (noise level). Note that FIG. 3 shows one sound, but in the actual voice, a signal as shown in FIG. Are interposed, and a data string of a continuous voiced section composed of a combination of a voiceless section Va and a voiced section Vb is defined as word data W, and a data string of a silent section N is defined as silent data S.

The word data extracting means 24 compares the signal level of each word data segmented by the word blocking means 23 with a predetermined threshold value R, and outputs word data Wa whose signal level is lower than the threshold value R. Extraction and amplification means 25
And the word data W whose signal level is equal to or higher than the threshold value R
b and the silence data S are output to the synthesizing means 26. The amplification unit 25 amplifies the signal level of the word data Wa lower than the threshold value R by one or more amplification degrees A.

The synthesizing means 26 determines that the signal level is equal to the threshold R
The word data Wb and the silence data S described above and the word data Wa ′ amplified by the amplifying unit 25 are synthesized in the order of the input voice, and stored in the intermediate buffer 27.

The speech speed conversion means 28 expands the unvoiced section, voiced section and each silent data of each word data stored in the intermediate buffer 27 at a predetermined expansion rate B, and inputs each expanded word block. The data is stored in the output buffer 29 in the order of voice.

The expansion rate B may be uniform for each of the unvoiced section, voiced section and silent section, or may be different for each section.

The D / A converter 30 converts the word blocks stored in the output buffer 29 into an analog audio signal in the order of the input audio and outputs the analog audio signal from the output terminal 20b.

Speakers and headphones are connected to the output terminal 20b via an amplifier or the like.
You can hear the result of editing the audio that was input to.

On the other hand, the threshold value R of the word data extraction means 24, the amplification degree A of the amplification means 25, and the expansion rate B of the speech speed conversion means 28 are set by the editing parameter setting means 31.

Here, the editing parameter setting means 31
As the threshold R and the amplification A, values calculated based on the level of a series of audio data stored in the input buffer 22 are set.

That is, as shown in FIG. 2, the level detection means 32 detects the maximum value H and the minimum value L of the absolute value of the signal level of a series of audio data stored in the input buffer 22.

The threshold setting means 33 includes a coefficient setting means 3
4, a coefficient α equal to or less than 1 and a maximum value H,
From the minimum value L, the following calculation, R = α · (HL), is performed to calculate the threshold value R, and the word data extraction means 2
Set to 4. The coefficient α set in the count setting means 34 is set to, for example, about 0.2 in order to discriminate a weakly pronounced word from another word by using the threshold value R.

The amplification degree setting means 35 sets the amplification degree A according to the coefficient α set in the coefficient setting means 34 in the amplification means 25.

Here, the amplification degree A is not less than 1/1/1 so that the signal level of the amplified word data does not exceed the maximum value H.
If the coefficient α is variable within a range equal to or less than α, for example, 1 / (2 ·
α) and so on.

The expansion rate setting means 36 can arbitrarily set the expansion rate B of the speech speed conversion means 28 within a predetermined range. For example, the expansion ratio B is set to 1 for each section.
, The voice becomes the voice of the same speed as the original voice, and if the expansion ratio B is made larger than 1 for each section, the voice is converted into a voice slower than the original voice. If it is smaller than 1, the voice is converted to a voice faster than the original voice.

The A / D converter 21 and the D / A converter 3 among the constituent elements of the audio data editing device 20 described above.
Each buffer 22, 27, 29 other than 0 and each of the means 23 to 26, 28, 31 can be operated by a computer.

In this case, the A / D converter 21 and the D / A
The buffers 22, 27, 29, the means 23 to 26, 2
The program is loaded from a medium (a floppy disk, a CD-ROM, an MO, etc.) on which a program for causing the programs 8 and 31 to function is recorded, and the computer executes the program.

For the audio data editing apparatus 20 configured as described above, for example, the English sentence “It's difficult for met” composed of eight words is as follows.
o finish thehomework. Is input, an analog audio signal (shown by a dashed line) shown in FIG. 4A is input to the input terminal 20a, and is converted into a series of audio data by the A / D converter 21. Is done.

Then, the maximum value H and the minimum value L are detected from the series of audio data, and the threshold value R and the amplification degree A are determined based on these values and the coefficient α as described above.

On the other hand, a series of voice data converted by the A / D converter 21 is converted into eight word data W1 to W8 and nine silence data S1 by the word blocker 23.
~ S9.

With respect to these word data W1 to W8,
The word data extraction means 24 extracts word data Wa whose signal level is lower than the threshold value R.

Here, as shown in FIG. 4A, word data Wa having a signal level lower than the threshold value R is used.
Means "for", "to", "the" as function words
, And these word data W3, W5, W7 are output to the amplifying means 25 and amplified at the amplification degree A.

Then, the amplified word data Wa '(W
3 ', W5', W7 ') are other word data Wb (W1, W1
W2, W4, W6, W8, W9) and silence data S1
Along with S9, the images are synthesized in the order of input by the synthesis means 26,
It is stored in the intermediate buffer 27.

Therefore, in the intermediate buffer 27, FIG.
(B), “for”, “to”, “th”
A series of audio data in which only the portion corresponding to "e" is amplified is stored.

If the expansion rate B of the speech speed conversion means 28 is 1 for each section, the D / A converter 30 outputs
The audio signal shown in FIG. 4B is output. Since this speech signal is obtained by selectively amplifying a word having a particularly weak pronunciation in the input speech shown in FIG. Even beginners can easily hear these words.

When the expansion rate B of the speech speed conversion means 28 is set to be larger than 1 for each section, the time width of each word data and silence data increases as shown in FIG. The user can hear the input voice at a slow speed, and the input voice can be heard more easily in addition to the word data amplification process described above.

It should be noted that the threshold value R, the amplification factor A and the expansion factor B set by the editing parameter setting means 31 may be fixed, or may be variable by an operation unit (not shown). As described above, by making the operation unit variable, it is possible to set the sound of the LL learning system, the television, the video, the radio, and the like so that the user can hear it most easily.

Further, in the above description, the data sequence of continuous voiced parts composed of a combination of unvoiced sections and voiced sections is used as word data to be distinguished from unvoiced data. A set of word data may be obtained by combining the data string of the part and the silence data preceding or following it. In this case, when the word data whose signal level is determined to be lower than the threshold value R is amplified by the amplifying means 25, only the sound part of the word data is amplified, and the silent data part is replaced with the original data. By outputting the signal as it is, a decrease in S / N can be prevented.

[0042]

As described above, according to the present invention, a series of voice data obtained by digitizing an analog voice signal is divided into word data, and word data having a signal level lower than a predetermined value are converted from the word data. The signal level is extracted and amplified, and the amplified word data and other data are combined in the order of the input voice and converted into an analog voice signal.

For this reason, it is possible to enhance the recognizability of a weakly pronounced word in the voice, and it becomes easy to hear a voice in a foreign language or the like.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration example of a main part of the embodiment.

FIG. 3 is a diagram showing a correspondence between an audio signal and each section.

FIG. 4 is a signal diagram for explaining the operation of the embodiment;

[Explanation of symbols]

 Reference Signs List 20 audio data editing device 21 A / D converter 23 word blocking means 24 word data extracting means 25 amplifying means 26 synthesizing means 28 speech speed converting means 30 D / A converter 31 editing parameter setting means 32 level detecting means 33 threshold Value setting means 34 coefficient setting means 35 amplification degree setting means 36 elongation magnification setting means

[Procedure amendment]

[Submission date] April 8, 1999 (1999.4.8)

[Procedure amendment 1]

[Document name to be amended] Drawing

[Correction target item name] All figures

[Correction method] Change

[Correction contents]

FIG.

FIG. 2

FIG. 3

FIG. 4

Claims (3)

[Claims] A step of classifying a series of voice data obtained by digitally converting an analog voice signal into word data; and word data having a signal level lower than a predetermined threshold value among the divided word data. And a step of synthesizing the amplified word data and other word data in the original voice order. 2. An A / D converter for converting an input analog voice signal into digital voice data, and a word block for dividing a series of voice data output from the A / D converter into word data Means, word data extraction means for extracting word data having a signal level lower than a predetermined threshold value among word data divided by the word blocking means, and word data extracted by the word data extraction means. Amplifying means for amplifying the signal level; word data not extracted by the word data extracting means among the word data divided by the word blocking means; and word data whose signal level has been amplified by the amplifying means. Means for synthesizing the voice data in the order of the input voice, and converting the voice data synthesized by the synthesis means into an analog voice. Audio data editing apparatus and a D / A converter for converting the items. 3. A computer, comprising: a word blocking unit for dividing a series of voice data into word data; and word data whose signal level is lower than a predetermined threshold value among the word data divided by the word blocking unit. Word data extracting means for extracting the word data; amplifying means for amplifying the signal level of the word data extracted by the word data extracting means; and the word data extracting means among the word data divided by the word blocking means. A medium on which is recorded a program for functioning as a synthesizing means for synthesizing word data not extracted in step (a) and word data whose signal level has been amplified by the amplifying means in the order of input speech.