JP2615649B2 - Word speech recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] The present invention generates a sound-repelling pattern from “na-line” and “ma-line” syllable patterns, creates and stores a voice pattern of a word having a repelling sound, The present invention relates to a word speech recognition device that performs word recognition. The object of the invention is to provide a word speech recognition device having a high word speech recognition rate for a word having a syllable syllable. And a consonant interval extracting unit for extracting a consonant section of a monosyllable pattern of "ma line" and "ma line" and "ma line" based on the pattern of the consonant period extracted by the consonant period extracting unit.
A means for generating each sound-repelling pattern and a sound-repelling pattern dictionary for storing the sound-repelling pattern are provided, and a sound-repelling pattern selection rule for storing a rule for selecting the sound-repelling pattern by syllable coupling is provided in the synthetic pattern creating section. Then, the words stored in the word dictionary unit are extracted by collating with the sound-repelling pattern selection rule unit, and the extracted words are stored in the single syllable pattern stored in the syllable template dictionary unit and the sound-repelling pattern dictionary unit. It is characterized in that a sound-repelling pattern is selected and a synthetic pattern is created in the synthetic pattern creating section.

[Industrial applications]

The present invention relates to a word-speech recognition device that generates a sound-repelling pattern from “na-line” and “ma-line” syllable patterns, creates and stores a voice pattern of a word having a repelling sound, and performs word recognition of an input voice. is there.

The word speech recognition device is a device that patterns a single syllable of an operator's voice, generates a synthesized pattern of words from the single syllable pattern, and recognizes a word of the voice input by the operator.

However, when a word is actually pronounced, it may not be pronounced according to the monosyllables that make up the word, but pronounced in different syllables.

The spelling pattern "n" that appears in a word is significantly different from the syllabic pattern of the "n" that was solely plucked.
, The word containing "" has a lower recognition rate.
There is a demand for a method of improving the recognition rate of words including "".

[Conventional technology]

FIG. 4 is a diagram for explaining the configuration of a conventional word speech recognition apparatus.

 First, a procedure for registering a syllable pattern will be described.

In FIG. 4, voice parameters of a voice input to the input unit 1 are extracted by a parameter extracting circuit 11.

That is, components such as a frequency component and a spectrum intensity of the input voice are extracted as a feature amount in accordance with a clock every fixed time.

Next, the section detection circuit 12 discriminates the length of a single syllable section due to the utterance of the input voice, detects each single syllable section, and separates it into single syllable sections.

The separated single syllable is registered as a single syllable pattern in the syllable template dictionary unit 2.

On the other hand, words to be recognized by the recognition device are represented by a symbol string of codes such as hiragana, katakana, alphabet, and phonetic symbols, and are stored in the word dictionary unit 3.

The synthesis pattern creation unit 4 includes the syllable template dictionary unit 2
By combining the monosyllabic patterns registered in the word dictionary unit 3
Create a composite pattern for each of the words stored in.

Then, the synthesized pattern of the words created for each word is stored in the synthesized pattern dictionary unit 5.

 Next, a procedure for recognizing an input voice will be described.

The voice input is formed into a pattern by the input unit 1, input to the pattern matching unit 6, and is checked one by one against a word pattern stored in the combined pattern dictionary unit 5, and a word of the matched pattern is output as a recognition result.

More specifically, for example, when “gaikoku” is registered as a word to be recognized in the word dictionary unit 3, “ga”, “i”, “K” and “K” are extracted, and a word pattern “Goku” is synthesized by the synthesizing unit 4 and registered in the synthesizing pattern dictionary unit 5.

When the voice input “gaikoku” is input, the pattern is formed into a pattern by the input unit 1, and the pattern is compared with the combined pattern of the words stored in the combined pattern dictionary unit 5 by the pattern matching unit 6. The pattern is compared with "Koku" and output as a recognition result.

[Problems to be solved by the invention]

There are syllables that need to be used depending on the syllables before and after the single syllable constituting the word when uttering.

For example, the utterance of the sound-repelling sound “n” is different from “n” of the word “Sendai” and “n” of “Gunma”.

Even if such a word is synthesized by applying a syllable pattern to the notation of the word, unlike a synthesized pattern of a word obtained from an actual utterance, word recognition cannot be performed by the above-described conventional speech recognition apparatus. Or incorrect recognition.

The present invention has been made in view of such a point, and creates a repellent sound "n" from a consonant section of a syllable pattern of "na line" and "ma line" by creating a repellent sound "n". It is an object of the present invention to provide a word-speech recognition device with high word-speech recognition by selecting a "" pattern and synthesizing a word pattern.

[Means for solving the problem]

In order to achieve the above object, a speech recognition device
A consonant interval extraction unit for extracting a consonant interval of a single syllable pattern of “na line” and “ma line”, a unit for creating a sound repelling pattern based on the extracted consonant period, and a sound repelling pattern dictionary unit for storing the sound repelling pattern And a sound pattern selecting rule storing a rule for selecting a sound pattern of the word pattern generated by the syllable combination in the synthetic pattern creating section.

(Operation)

When a word matching the rule stored in the sound-repelling-pattern selection rule unit is extracted from the words stored in the word dictionary unit and a word pattern is generated by the synthetic pattern generating unit, the sound-repelling selection rule includes the sound-repelling sound. It is created by selecting and using a sound-repelling syllable pattern of a word pattern.

For this purpose, the consonant sections of the syllable patterns “na-row” and “ma-row” registered in the syllable template dictionary section are extracted by the consonant section extraction section, and the characteristics of the predetermined section of the consonants of the syllable pattern are extracted by the average value calculation section. An average value of the amount is calculated, and a pattern in which the average value of the characteristic amount continues for a predetermined time is defined as a sound-repelling syllable pattern.

Then, the created sound-repellent syllable pattern is stored in the sound-repelling pattern dictionary unit as a single syllable pattern when a word conforming to the sound-repelling selection rule is synthesized.

The word pattern in which the two sound-repelling patterns are selectively applied in accordance with the sound-repelling selection rule is stored in the combined pattern storage unit together with the other word patterns.

Therefore, the synthesized pattern storage unit stores the synthesized pattern of words close to the utterance word, and the pattern matching unit improves the recognition rate of the input word to be recognized.

〔Example〕

Hereinafter, embodiments of the word voice input device of the present invention will be described with reference to the drawings.

 FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a diagram for explaining an extraction section, and FIG. 3 is a diagram showing an example of a syllable syllable.

 The same reference numerals indicate the same objects throughout the drawings.

First, when registering the voice of the operator, the voice of the operator is patterned by the input unit 1, separated into single syllables, and stored as a syllable pattern in the syllable template dictionary unit 2.

In the stored syllable pattern, as shown in FIG. 2, the consonant interval extraction unit 7 sets the point where the variation of the parameter between adjacent frames in the single syllable pattern is the maximum as the center of the transition part of the consonant and vowel. The section from the beginning of the pattern to the transition point is extracted.

That is, the peak of the variation of the adjacent frame destination parameter is generated at the beginning of the sound, the transition point of the consonant and the vowel, and the end of the sound in the normal single syllable pattern (excluding the small peak of silence at the time of the sound).

Therefore, a portion between the first peak and the second peak from the utterance start point is defined as a consonant extraction section.

Then, the average value calculating unit 8 calculates the “na” of the “na row” single syllable.
In the consonant extraction section of “ni”, “nu”, “ne”, and “no”, the average value of the entire feature amount in a predetermined section (usually the first half section) where consonant features are remarkable is calculated.

Similarly, the single syllables “ma”, “mi”, “mu” and “me”
The average value of the entire feature amount in a predetermined section of the “mo” consonant extraction section is calculated.

A pattern having the average value of the feature values of the consonants “na-line” and “ma-line” and having a predetermined duration is defined as a syllable pattern of the repellent sound as “n *” and “n #”, respectively. It is memorized.

On the other hand, there are two cases of the sound utterance of the word to be uttered. (1) When uttered with "n *" like "n" of "Sendai", (2) When uttered with "n #" like "n" of "Gunma", There is.

Accordingly, a synthetic pattern of a word including a sound repellent is created by the synthetic pattern creating section 4 and stored in the synthetic pattern dictionary section 5.

The use example of the sound-repelling pattern has been specifically described above. As a general rule for selecting “n *” and “n #”, the consonant of the syllable connected immediately after the rule
Or when the syllable is "ぱ row" ([m] [p]
)), “N #” is used for sound repelling, and “n *” is used otherwise.

Such a rule is stored in the sound-repelling-pattern selection rule unit 10, and the words registered in the word dictionary unit 3 are applied when the combined pattern creating unit 4 creates a word pattern.

As a result, a synthesized pattern of a word having a high similarity to the uttered sound is prepared in the synthesized pattern dictionary unit 6, and therefore, the probability of matching with the input voice in the pattern matching unit 6 at the time of recognition is increased.

Note that the combined pattern dictionary unit 5 may have only one pattern area without constantly holding all the combined patterns, and may replace the patterns sequentially created by the combined pattern creation unit at the time of comparison.

By doing so, the memory capacity can be significantly reduced.

For the sound-repelling pattern, the average value extracted from each of the consonant sections of “na line” and “ma line” was used.
It can be created using a specific value obtained from the consonant section of “ma row”.

〔The invention's effect〕

As described above, according to the present invention, the sound-repellent “n”
Syllable pattern is created from the consonant sections of the “na-line” and “ma-line” syllable patterns, selected and used according to the connected syllable pattern, and the synthesized pattern of words that approximates the word utterance of the actual voice is stored. Therefore, the recognition rate of word speech recognition is improved, which is extremely useful in practice.

[Brief description of the drawings]

 FIG. 1 is a block diagram showing the configuration of a word speech recognition apparatus according to the present invention, FIG. 2 is a diagram for explaining an extraction section, FIG. 3 is a diagram showing an example of a lyric syllable, and FIG. FIG. In the figure, 1 is an input section, 2 is a syllable template dictionary section, 3 is a word dictionary section, 4 is a synthetic pattern creating section, 5 is a synthetic pattern dictionary section, 6 is a pattern matching section, 7 is a consonant section extracting section, and 8 is An average value calculation unit, 9 is a sound-repelling pattern dictionary unit, and 10 is a sound-repelling pattern selection rule unit.

Claims (1)

(57) [Claims] 1. An input unit (1) for extracting and patternizing a feature amount from an input speech, a syllable template dictionary unit (2) for storing a single syllable pattern patterned for each syllable of the speech, and a symbol string. A word dictionary unit (3) that stores the words described in (1), and a word in the word dictionary unit (3) is used to create a composite pattern of the words using the single syllable patterns stored in the syllable template dictionary unit (2). And a composite pattern dictionary unit (5) for storing the composite pattern of the words created by the composite pattern creation unit (4).
And a pattern matching unit (6) for matching the combined pattern stored in the combined pattern dictionary unit (5) with the voice pattern input to the input unit (1). ) Is stored in the syllable template dictionary unit (2), and the syllable template dictionary unit (2) of the voice recognition device that recognizes the unknown word voice input to the input unit (1) is stored. ), A consonant interval extracting unit (7) for extracting a consonant interval of a monosyllable pattern of “na line” and “ma line”, and the consonant interval extracting unit (7)
A means (8) for generating a sound-repelling pattern for each of "na-line" and "ma-line" based on the pattern of the consonant section extracted in step (8), and a sound-repelling pattern dictionary unit (9) for storing the sound-repelling pattern, and The synthetic pattern creation section (4) is provided with a sound-repelling pattern selection rule section (10) for storing a rule for selecting the sound-repelling pattern by syllable combination, and the words stored in the word dictionary section (3) are converted to the sound-repelling patterns. The syllable template dictionary unit (2) selects a single syllable pattern stored in the syllable template dictionary unit (2) and a sound-repelling pattern stored in the sound-repelling pattern dictionary unit (9). A word-speech recognition apparatus, wherein a synthesized pattern is created by the synthesized-pattern creating unit (4).