JPS6370298A - Double consonant recognition equipment - 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 [Object of the Invention] (Field of Industrial Application) The present invention relates to a consonant letter recognition device for voice recognition.

(Prior art and its problems) In the conventional speech recognition, recognition of consonants is indispensable. Conventional consonant processing involves, for example, pronouncing ``tsu'' and converting it into ``tsu'' by operating a key or the like.

However, this method has the problem that key operations are troublesome. Also, the "tsu" part is silent for a long time, in other words, in "kakutoku katsuta" and "kata (kata)", the silent section between "ka" and "ta" is better in "katsuta". Automatic recognition was carried out by taking advantage of the fact that it was longer than the ``type''.

However, with this method, the length of the silent interval varies from utterance to utterance. This variation becomes even larger when the speakers are different, making it extremely difficult to recognize consonants.

On the other hand, if the utterer is the same, the sound is relatively small, but even so, it may become louder depending on the phoneme before and after the consonant, and it is not possible to clearly distinguish when a consonant is present and when it is not, leading to misrecognition. The problem was that it was often put away.

The present invention has been made to address these problems.
The purpose is to provide a consonant recognition device that can reliably recognize consonant consonants and has good operability.

[Configuration of the Invention (Means for Solving Problems) In order to achieve the above object, the present invention provides an audio input unit that converts audio into an electrical signal, and an analysis method that determines the acoustic parameters of the audio that has been converted into an electrical signal. a silence/sound detecting unit that determines silence or utterance based on the acoustic parameters; a key input unit that inputs a key operation indicating a unit of audio input; and a silent section length determined from the output of the silence/sound detecting unit. is detected, and when it is determined that the length of the silent section has reached a certain threshold, it is recognized as a consonant and control is performed to display the consonant, and when the length of the silent section is greater than the threshold and reaches another threshold, the consonant is displayed. a display control unit that performs control to cancel the display of a consonant at the end of the voice input unit when a key indicating a voice input unit is pressed from the key input unit; The present invention is characterized by comprising a display section that displays phonemes.

(Function) When the silent interval length is smaller than a threshold, a consonant is not displayed, when the silent interval length is smaller than another threshold, a consonant is displayed, and when the silent interval length is greater than another threshold, a consonant is displayed. The display is erased, and the consonants are also erased by key operation.

(Example) Hereinafter, the industrial water invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a consonant recognition device according to the present embodiment. As shown in the figure, this consonant recognition device includes a voice input section 1, an analysis section 2, a voiceless utterance detection section 3, and a display control section. Part 4,
It consists of a display section 5 and a key input section 6.

An audio signal input via a microphone or the like is converted into an electrical signal via an audio input section 1 and guided to an analysis section 2. The analysis section 2 is composed of, for example, a group of 16 channel band-pass filters, and is used to find acoustic parameters that effectively express the characteristics of the voice. Note that this acoustic parameter is not limited to the output of the band-pass filter group described above, but may be given as a part of various parameters obtained by cepstral coefficients or correlation analysis, or as a combination of multiple parameters. good.

However, in this case, the analysis unit 2 calculates the power in the entire band together with the data of the acoustic parameters, and also calculates the power in the entire band.
Let it be the lV parameter.

The silence/speech detection unit 3 determines whether the audio frame is silent or speech based on the acoustic parameters. The determination of silence or sound is made by a threshold value. For example, the soundless/sound detecting unit 3 determines that there is a sound when the power is equal to or higher than the interrogation TP, and that there is no sound when the power is smaller than the threshold Tp. Then, a signal indicating whether there is no sound or sound is sent to the display control section 4.

The display control unit 4 is equipped with a silence counter that counts silent frames, and uses the counted number of silent frames as the silent interval length, and when it determines that the silent interval length has reached a certain threshold value Ta, it recognizes it as a consonant and displays the consonant. When the silent interval length reaches another threshold value Tb larger than the threshold value, the display of the consonant is canceled, and when the key indicating the voice input unit is pressed from the key input unit 6, the end of the voice input unit is pressed. Performs 1til control to cancel the display of the consonant. The display unit 5 displays the recognized voice. The key input unit 6 has operation keys arranged, and the operator can perform key operations indicating voice input units.

Next, the operation of this embodiment will be explained.

When the operator pronounces "ka", this audio data is converted into an electrical signal by the audio input section 1, acoustic parameters are determined by the analysis section 2, and the sound is determined to be uttered by the utterance detection section 3. "Ka" is displayed on the display section 5 via the rough display 1llllJ te1 section 4.

Next, the operator pronounces "ta", and the cases will be explained based on the time from pronouncing "ka" to pronouncing "ta".

Figure 2(a) shows the case where ``ka'' and ``ta'' are pronounced consecutively. When the pronunciation of "ka" is finished, the silence detection section 3 determines that there is no sound, and the silence counter of the display control section 4 is reset, the number of silent frames is counted, and the length of the silent period until "ta" is pronounced is detected. Ru. In this case, since the silent interval length is smaller than the preset threshold value Ta, the display control section 4 assumes that "ka" and "da" are pronounced consecutively, and the display section 5 displays "kata". Ru.

FIG. 2(b) shows a case where the operator pronounces "It was". Similarly to FIG. 2(a), when the pronunciation of "h" ends, the silence counter in the display control section 4 is reset, and the length of the silent section is detected.

Since the silent interval length at this time is greater than the threshold value Ta, the display control section 4 assumes that a consonant is necessary between "ka" and "ta", and displays "ita" on the display section 5.

FIG. 2(C) shows a case where the operator pronounces "ka", pauses the pronunciation, and then pronounces "ta" after a while.

In this case, after the pronunciation of "ka" ends, the silence counter operates to detect the length of the silent period.

Since this silent interval length is greater than the threshold value Ta, the display control section 4 assumes that there is a consonant between "ka" and "ta", and the display section 5 temporarily displays "Kat".

However, since this silent interval length is larger than the second threshold Tb, the display control section 4 erases the consonant and the display section 5 displays "kata".

Figure 2 (d> is to prevent a consonant from entering at the end of an input unit such as a segment. When the operator finishes pronouncing "ka", the silence counter operates and the length of the silent interval in this case is calculated. is larger than the lower threshold a, so "and" is displayed on the display section 5. Since a consonant is displayed on the display section 5 at the end of the phrase, the operator operates a predetermined key on the key input section 6. The display control section 4 erases the consonant in response to a command from the key input section 6, and the display section 5 displays "ka".

In this way, in this embodiment, "
There is no need to convert tsu to tsu. Furthermore, it is possible to clearly distinguish between the presence and absence of a consonant, and to perform reliable speech recognition.

[Effects of the Invention] As described above in detail, according to the present invention, it is possible to provide a consonant recognition device that reliably recognizes consonants and has good operability.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a consonant recognition device according to an embodiment of the present invention, and FIG. 2 is an explanatory diagram showing the operation of the embodiment. 1...Voice input 2...Analysis section 3...Speechless sound detection section 4...Display control section 5...Display section 6...Key input section

Claims (1)

[Scope of Claims] An audio input unit that converts audio into an electrical signal, an analysis unit that determines acoustic parameters of the audio converted into an electrical signal, and a silence/sound detection unit that determines whether there is silence or utterance based on the acoustic parameters. a key input unit for inputting a key operation indicating a voice input unit; detecting a silent interval length from the output of the voiceless utterance detection unit, and recognizing it as a consonant when determining that the silent interval length has reached a certain threshold; control is performed to display a consonant, and when the silent interval length reaches another threshold value greater than the threshold, control is performed to cancel the display of the consonant, and a key input indicating a voice input unit is input from the key input section. A display control unit that performs control to cancel display of a consonant at the end of a voice input unit; and a display unit that displays a phoneme in accordance with a command from the display unit.