JP2737122B2 - Voice dictionary creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] The present invention relates to a speech dictionary creating apparatus for creating and storing speech patterns of words having unvoiced syllables and performing word recognition of input speech. An input unit for outputting a word pattern or a monosyllabic pattern of an input voice, comprising: a parameter extracting unit to be extracted; a syllable interval detecting circuit for detecting a syllable interval of the extracted feature quantity; and storing the monosyllabic pattern. A syllable template dictionary unit, a consonant interval extraction unit that extracts a consonant interval from the single syllable patterns stored in the syllable template dictionary unit, and an average value calculation unit that calculates an average value of the feature amount in the consonant interval An unvoiced syllable pattern dictionary unit that stores the output from the average calculation unit as a unvoiced syllable pattern; a word dictionary unit that stores word pronunciation in advance; A voiceless syllable from the output of the unvoiced rule unit that stores the rules of the syllables, the syllable plate dictionary, the voiceless syllable pattern dictionary, the word dictionary, and the voiceless rule And a synthesis pattern creation unit that creates a synthesis pattern of a word including. BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech dictionary creation device that creates and stores speech patterns of words having unvoiced syllables and performs word recognition of input speech. The word speech recognition device is a device that patterns a single syllable of an operator's utterance, generates a synthesized pattern of words from the single syllable pattern, and recognizes a word of a voice input by the operator. However, some words contain syllables without vocal fold vibration when the word is actually pronounced. Since these words have a low recognition rate, there is a need for a method of improving the recognition rate. [Prior Art] FIG. 4 is a diagram for explaining a configuration of a conventional word speech recognition apparatus. First, a procedure for registering a syllable pattern will be described. In FIG. 4, voice parameters (features) are extracted from a voice input to the input unit 1 by a parameter extraction circuit 11. That is, a frequency component, a spectrum intensity (power), and the like of the input voice in accordance with a clock at regular time intervals. Is extracted as a feature quantity. Next, the section detection circuit 12 discriminates the length of a single syllable section by utterance of the input speech, detects each single syllable section, and separates the single syllable into single syllables. The syllables are registered as a patterned single syllable pattern in the syllable template dictionary unit 2. On the other hand, words to be recognized by this recognition device are represented by hiragana, katakana, alphabet, or a symbol string of codes such as phonetic symbols. Is stored in the word dictionary unit 3. The synthetic pattern creating unit 4 is configured to store 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 input of the voice to be recognized is patterned by the input unit 1, input to the pattern matching unit 6, collated with the word pattern stored in the combined pattern dictionary unit 5 one by one, and the words of the matched pattern are output as the recognition result. More specifically, for example, as shown in Example 1 of FIG. 3, when “Nichiden” is registered as a word to be recognized in the word dictionary unit 3, a single word registered in advance by the voice of the operator is used. From the syllable patterns, "ni", "chi", "de",
“N” is extracted, the word pattern “Nichiden” is synthesized by the synthesizing unit 4 and registered in the synthesizing pattern dictionary unit 5. When a recognized voice input “Nichiden” is input, the input unit 1 forms a pattern, and the pattern matching unit 6 compares the pattern with the combined pattern of the words stored in the combined pattern dictionary unit 5. Then, the pattern is compared with "Nichiden" and output as a recognition result. [Problems to be Solved by the Invention] Among words, there are syllables uttered without vocal cord vibration when uttering, that is, words having unvoiced syllables. For example, a normal utterance is uttered without vocal cord vibration, such as "chi" of the word "Hitachi". Even if such a word is synthesized by applying a voice pattern to a kana notation of a word, unlike a synthesized pattern of a word obtained from an actual utterance, word recognition cannot be performed by the above-described conventional voice recognition device. Or incorrect recognition. The present invention has been made in view of such a point, and an object of the present invention is to provide a word-speech recognition device having high word-speech recognition by synthesizing a word pattern having unvoiced syllables. [Means for Solving the Problems] The object described above comprises a parameter extraction unit for extracting a feature amount of the input speech, and a syllable interval detection circuit for detecting a syllable interval of the extracted feature amount, An input unit for outputting a word pattern or a monosyllabic pattern, a syllable template dictionary for storing the monosyllabic pattern, and a consonant interval for extracting a consonant interval among the monosyllabic patterns stored in the syllable template dictionary An extraction unit; an average value calculation unit that calculates an average value of the feature amount in the consonant section; an unvoiced syllable pattern dictionary unit that stores an output from the average calculation unit as a voiceless syllable pattern; A word dictionary section stored in advance, a voiceless rule section storing rules of voiceless syllables, the syllable plate dictionary section, and the voiceless syllable pattern A speech dictionary creating apparatus configured to have a combined pattern creating unit that creates a combined pattern of words including unvoiced syllables from the output of the writing unit, the word dictionary unit, and the unvoiced rule unit. Is done. [Operation] When words that match the rules stored in the unvoiced rule section are extracted from the words stored in the word dictionary section and the combined pattern creating section creates a word pattern, A syllable pattern that is usually voiced and uttered is created using a special syllable pattern that is devoiced. For that purpose, the consonant section of the required syllable pattern registered in the syllable template dictionary section is extracted by the consonant section extracting section, and the average value calculating section calculates the average value of the characteristic values of the consonant section of the syllable pattern, and the consonant section Are extracted, the average value thereof is obtained, and the pattern having the average value as the amplitude is used as the syllable pattern of the unvoiced syllable. Then, the created syllable pattern indicating the unvoiced portion is stored in the unvoiced syllable pattern dictionary as a single syllable pattern when a word conforming to the unvoiced rule is synthesized. The word pattern created in accordance with the devoicing rule in this manner is stored in the combined pattern storage unit together with 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. [Embodiment] Hereinafter, an embodiment of a word speech 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 the configuration of a syllable, and FIG. 3 is a diagram showing an example of a voiceless 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, for example, for "chi", as shown in FIG. 2, the consonant interval extraction unit 7 determines the consonant interval [t 域] and the vowel interval [i ] Is determined. The average value calculating section 8 calculates the average value of the characteristic values in the consonant section [t∫], and has the average value of the characteristic values of the spectrum [t∫] in the syllable section “chi” and has a predetermined duration. “Chi *” is stored in the unvoiced syllable pattern dictionary unit 9 as a single syllable pattern. On the other hand, the word "chi" to be uttered may have three cases depending on the syllables before and after as shown in FIG. That is, (1) When voice is pronounced like "Chi" of "Nichiden", (2) When voice is voiced with a high probability like "Chi" of "Hitachi", ( 3) Like “chi” of “ichi”, in any of (1) and (2), therefore, in the case of (1), the syllable pattern “chi” is used, and in the case of (2), “ "*" In the case of (3)
And a composite pattern of words using both “*” and “chi *” is created by the composite pattern creation unit 4 and stored in the composite pattern dictionary unit 5. Although the case of the syllable pattern "chi" was specifically described above,
As a general rule for unvoicing vowels, Rule 1 "K""K""S""S""J""T""H"
The beats of "F", "B", "B", "S" are "K", "S"
When it comes just before the beat of each line such as "ta", "c", "ba". Rule 2 "ki""ku""shi""su""chi""tsu""hi"
When a beat such as “F”, “B”, “B”, or “S” comes right before the break, and the accent of that beat is low. [Refer to NHK Japanese Pronunciation Accent Dictionary] These rules are stored in the devoicing rule section 10 and the words registered in the word dictionary section 3 are applied when the composite pattern creating section 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 section 5, and therefore, the probability of matching with the input voice in the pattern matching section 6 at the time of recognition is increased. Note that the composite pattern dictionary unit 5 does not need to hold all the composite patterns at all times, but may have only one pattern area, and replace the patterns sequentially at the time of comparison. By doing so, the memory capacity can be significantly reduced. Further, the unvoiced syllable pattern was created using the average value of the characteristic values of the consonant sections. However, as another method, for example, a method of synthesizing a pattern of a predetermined duration by complementing or thinning out the extracted consonant pattern is also used. good. [Effects of the Invention] As described above, according to the present invention, an unvoiced syllable pattern is created from a consonant section of an original monosyllable pattern, and a synthesized pattern of words that approximates the word pronunciation of an actual voice is stored. Therefore, the recognition rate of word speech recognition is improved, and it is extremely useful in practice.

[Brief description of the drawings] FIG. 1 is a configuration block diagram of an embodiment of the present invention, FIG. 2 is a diagram illustrating the configuration of syllables, FIG. 3 is a diagram illustrating an example of unvoiced syllables; FIG. 4 is a diagram illustrating the configuration of a conventional example. In the figure, 1 is an input unit, 2 is a syllable template dictionary, 3 is a word dictionary section, 4 is a composite pattern creation unit, 5 is a composite pattern dictionary unit, 6 is a pattern matching unit, 7 is a consonant section extraction unit, 8 is an average value calculation unit, 9 is a voiceless syllable pattern dictionary section, 10 is the voiceless rule dictionary, 11 is a feature amount extraction unit, Reference numeral 12 denotes a section detection circuit.

Claims (1)

(57) [Claims] A parameter extraction unit that extracts a feature amount of the input voice;
An input unit configured to output a word pattern or a single syllable pattern of the input speech, and a syllable template dictionary unit that stores the single syllable pattern; A consonant interval extraction unit that extracts a consonant interval from the single syllable patterns stored in the syllable template dictionary unit; an average value calculation unit that calculates an average value of a feature amount in the consonant interval; Unvoiced syllable pattern dictionary unit that stores the output of vocalized syllables as syllable patterns, word dictionary unit that stores the pronunciation of words in advance, and unvoiced syllable pattern selection that stores the rules of unvoiced syllables A rule unit, the syllable plate dictionary unit, the unvoiced syllable pattern dictionary unit, the word dictionary unit, and the output of the unvoiced syllable pattern selection rule unit. A synthesizing pattern creating unit for creating a synthesizing pattern of words including unvoiced syllables.