JPS58113998A - Voice recognition system - Google Patents
Voice recognition systemInfo
- Publication number
- JPS58113998A JPS58113998A JP56214036A JP21403681A JPS58113998A JP S58113998 A JPS58113998 A JP S58113998A JP 56214036 A JP56214036 A JP 56214036A JP 21403681 A JP21403681 A JP 21403681A JP S58113998 A JPS58113998 A JP S58113998A
- Authority
- JP
- Japan
- Prior art keywords
- group
- word
- conjugation
- stem
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
(57)【要約】本公報は電子出願前の出願データであるた
め要約のデータは記録されません。(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.
Description
【発明の詳細な説明】 （１）＠明の稜術分野 本発明は音声認識が式に関する。[Detailed description of the invention] (1) @ Ming's Edge Technique Field The present invention relates to speech recognition equations.
（２）従来技術と問題点
従来の文節単位のｆ７’認織方式においては、各季語に
名詞・動詞・形容詞・助動詞等があり活用・語尾変化が
ある場合、各活用形を各々別の単鎖として扱っている。(2) Prior art and problems In the conventional clause-based f7' recognition method, when each seasonal word has nouns, verbs, adjectives, auxiliary verbs, etc., and there are conjugations and inflections, each conjugate is separated into a separate word. Treated as a chain.
そのため、語い数を増やすと辞書データの記憶域が増大
するだけでなく、認識率の低下及び認識処理時間の増大
などの欠点がある。Therefore, increasing the number of words not only increases the storage area of dictionary data, but also has drawbacks such as a decrease in recognition rate and an increase in recognition processing time.
（３）発明の目的
本発明の目的は、辞書データに文法情報を付加し、その
文法情報に則り辞書データを結合しながら入力データと
の比較を行なうことにより、暗い数が増大しても辞書デ
ータの記憶域をそれほど増やさず、能率的に文節単位の
音声認識を行なう音声ｎ方式を提供することにある。(3) Purpose of the Invention The purpose of the present invention is to add grammatical information to dictionary data, and to combine the dictionary data according to the grammatical information and compare it with the input data, so that even if the number of dark numbers increases, the dictionary The object of the present invention is to provide a speech n method that efficiently performs speech recognition in units of phrases without significantly increasing the data storage area.
（４）発明の構成
り記目的を達成するための本発明の特徴は、辞蒼データ
を語幹グループと活用語尾グループに分割し、各々の語
幹にどの活用語尾グループと納付できるかという品詞に
応じた文法情報を付加する。そして、辞書検索の際、ま
す語幹グループのうちどのグループに属するか前処理し
、そのグループ内で前述の文法情報からその語幹と活用
飴尾を結合し、入力データと比較することにある。(4) Structure of the invention A feature of the present invention to achieve the purpose is to divide dictionary data into word stem groups and conjugation ending groups, and to determine which conjugation ending group can be assigned to each stem, depending on the part of speech. Add grammatical information. Then, when searching the dictionary, preprocessing is performed to determine which group of word stem groups it belongs to, and within that group, the word stem and the conjugated Ameo are combined based on the above-mentioned grammatical information, and compared with the input data.
（５）発明の実施例
第１図は本発明が適用さｎる汁声ｇａシステムのブロッ
ク図である。(5) Embodiment of the Invention FIG. 1 is a block diagram of a GA system to which the present invention is applied.
音声入力は音声の特徴パラメータ抽出部１に入力されて
特徴パラメータが抽出され、このパラメータにより予め
登録しである辞４２内の辞書データが検索される。The voice input is input to the voice feature parameter extracting section 1 to extract feature parameters, and dictionary data in the dictionary 42, which has been registered in advance, is searched using these parameters.
そして検索の結果が出力部３から出力される。The search results are then output from the output unit 3.
第２図は本発明の一実施例を説明する図である。FIG. 2 is a diagram illustrating an embodiment of the present invention.
辞書２内の辞書データは語幹グループ２１と活用・四尾
グループ２２を有する。さらに語幹グループ２１内は例
えばその音節数により、第１グループ２１−１． ・
・・・、第１グループ２１−１・・・・・・の小グルー
プに分けられている。The dictionary data in the dictionary 2 has a word stem group 21 and a conjugation/four-tail group 22. Further, within the stem group 21, for example, depending on the number of syllables, the first group 21-1.・
..., the first group 21-1, and so on.
第１グループ２１−１は「関係」、「美し」の妬＜３ｆ
節、第、グーープ；、−４は「私」、「あなた」の如く
３音節である。The first group 21-1 is “relationship” and “beauty” jealousy <3f
Clause, first, goop;, -4 is three syllables, like ``I'' and ``you.''
一ガ、活用・語尾グループ２２は助詞、助動詞等の種類
毎に第１グループ２２−１．・・・・・・第１グループ
２２１．・・・・と分けである。The conjugation/suffix group 22 includes a first group 22-1 for each type of particles, auxiliary verbs, etc. ...First group 221. It is divided into...
まず、入力データ２３が入ると前選択の処理が行なわれ
、語幹グループ２１内のある特定グループが選択される
。ここでは、人力データ「音声は」であり、音節数は５
である。従ってこの前選択により、音節数５になり得る
音節数４のグループ２１−１が選択される。次にその小
グループ内で。First, when the input data 23 is input, a pre-selection process is performed, and a certain specific group within the word stem group 21 is selected. Here, the human data is ``sound is'', and the number of syllables is 5.
It is. Therefore, this pre-selection selects the group 21-1, which has a number of syllables of four, which can have a number of five syllables. Then within that small group.
どの語幹はどの活用語尾グループが適切であるかという
文法情報２１ａ＝Ｌ ・・・・・・、２１ａ−１゜・・
・・・を活用し、その小グループ内の語幹についてのみ
、活用・語尾と次々に結合し、その結合されてでき九文
節あるいは一語と、入力データとの比較を行ない、最も
似ているものを認識結東とするものである。ここで、語
幹と＠鳩を結合する方法としては例えば、■語尾の先頭
音節が母ｉの場合は、音節故に応じた比率で結合する。Grammatical information 21a=L, 21a-1゜, 21a-1゜, which word stem is appropriate for which conjugation ending group.
..., only the stems in that small group are combined with conjugations and endings one after another, and the nine clauses or single words formed by these combinations are compared with the input data to find the most similar one. This is the recognition of the following. Here, as a method of combining the word stem and @ pigeon, for example, ■ If the first syllable at the end of the word is the mother i, it is combined at a ratio according to the syllable.
■語尾の先頭音節が摩擦音・ハレツ音である場合は、間
にノイズを混ぜて音節数に応じた比率で、結合する方法
がある。このような結合情報２２ａ−１，１１０，・を
活用語尾グループに付加しておく。■If the first syllable at the end of a word is a fricative or a haretsu sound, there is a way to mix noise in between and combine them at a ratio depending on the number of syllables. Such combination information 22a-1, 110, . is added to the inflection ending group.
（６） 発明の効果
以上説明した本発明によれば、胎い数が増大しても辞書
検索で吸うデータ数は少なく、辞書記憶領域が莫大にな
らず、認識処理時間が短縮される。(6) Effects of the Invention According to the present invention described above, even if the number of babies increases, the amount of data required for dictionary search is small, the dictionary storage area does not become enormous, and the recognition processing time is shortened.
８１図は本発明が適用される音声認識システムを示すブ
ロック図、第２図は本発明の一実施例を説明する図であ
る。
２：辞 書、２１：語幹グループ、２２：活用・語尾
グループ、２１−１． ・・・・、２１−１゜・・・
および２２−１．・・・・・、２２−１．・・・・・：
小グループ、 ２１ ａ−Ｉ！ ・・・・・、２１
ａ−１・・・弓第 １ 図FIG. 81 is a block diagram showing a speech recognition system to which the present invention is applied, and FIG. 2 is a diagram illustrating an embodiment of the present invention. 2: Dictionary, 21: Stem group, 22: Conjugation/Ending group, 21-1. ..., 21-1°...
and 22-1. ..., 22-1.・・・・・・：
Small group, 21 a-I! ...,21
a-1...Bow Figure 1
Claims (1)
ータとを比較しｇ鷹する手段を備え１文節率位の認識を
行なう音声認識方式において、辞書データは接続すべき
活用語吊金指示する文法情報を待つ語幹グループと該文
法情報により指示される活用・語尾グループを有し、か
つ該語幹グループと該活用・語尾グループはそれぞれ複
数の小グループに分かれ、前選択処理で検索すべき該小
グループの１つを選択し、その中で語幹と活用・語尾を
結合しつつ、入力音声データと一致検索をすることを特
徴とする音声認識方式。In a voice recognition method that is equipped with a means to compare and compare pre-registered dictionary data with newly input voice data and perform recognition at a rate of one phrase, the dictionary data instructs the conjugated word to be connected. It has a word stem group that waits for grammatical information and a conjugation/inflection group that is instructed by the grammatical information, and each of the stem group and the conjugation/inflection group is divided into a plurality of small groups. A speech recognition method characterized by selecting one of the groups and searching for a match with input speech data while combining the stem, conjugation, and ending of the word.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP56214036A JPS58113998A (en) | 1981-12-26 | 1981-12-26 | Voice recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP56214036A JPS58113998A (en) | 1981-12-26 | 1981-12-26 | Voice recognition system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPS58113998A true JPS58113998A (en) | 1983-07-07 |
| JPH0139599B2 JPH0139599B2 (en) | 1989-08-22 |
Family
ID=16649208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP56214036A Granted JPS58113998A (en) | 1981-12-26 | 1981-12-26 | Voice recognition system |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JPS58113998A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01217399A (en) * | 1988-02-25 | 1989-08-30 | Fujitsu Ltd | System for recognizing voice inputted at every clause |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS5592923A (en) * | 1978-12-29 | 1980-07-14 | Fuji Xerox Co Ltd | Typewriter for japanese letter |
-
1981
- 1981-12-26 JP JP56214036A patent/JPS58113998A/en active Granted
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS5592923A (en) * | 1978-12-29 | 1980-07-14 | Fuji Xerox Co Ltd | Typewriter for japanese letter |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01217399A (en) * | 1988-02-25 | 1989-08-30 | Fujitsu Ltd | System for recognizing voice inputted at every clause |
Also Published As
| Publication number | Publication date |
|---|---|
| JPH0139599B2 (en) | 1989-08-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bhattacharya et al. | Deep Speaker Embeddings for Short-Duration Speaker Verification. | |
| Mei et al. | An encoder-decoder based audio captioning system with transfer and reinforcement learning for DCASE challenge 2021 task 6 | |
| Abidin et al. | Deep neural network for automatic speech recognition from Indonesian audio using several lexicon types | |
| Wingfield et al. | Word recognition from acoustic onsets and acoustic offsets: Effects of cohort size and syllabic stress | |
| JPS58113998A (en) | Voice recognition system | |
| Seeber | Ideal languages in the french and english imaginary voyage | |
| Enoch | Jewish Georgian | |
| Brown | Automatic recognition of geographically-proximate accents using content-controlled and content-mismatched speech data. | |
| JPS616693A (en) | Accent decision method | |
| Dioubina et al. | An IPA vowel diagram approach to analysing L1 effects on vowel production and perception. | |
| JPS63221475A (en) | Analyzing method for syntax | |
| Gable et al. | Speech errors and task demand. | |
| Boumehdi et al. | Comparison of New Approaches of Semi-syllable Units for Speech Recognition of Any Arabic Word | |
| Sekine et al. | NYU/BBN 1994 CSR evaluation | |
| Laitinen | Would you like to have some beta with those slopers?: anglicisms in Finnish sport climbing jargon | |
| Huang | Effects of context, stress, and speech style on american vowels. | |
| Albrektson | Singing or Pruning? | |
| Kashino et al. | The influence of knowledge and experience during the processing of spoken words: non-native listeners. | |
| Yomie et al. | Application of the multilingual acoustic representation model XLSR-53 for the transcription of Ewondo | |
| JPS59185400A (en) | Monosyllable sound recognition system | |
| Miller | Morphologically determined allomorphs in spoken Tibetan | |
| Normanskaya et al. | Accuracy Analysis of Mansi Transcription in Various Sources of the Beginning of the 20th Century. | |
| Scotton et al. | Loan words and the borrowing process in two Ateso dialects | |
| JPS5968798A (en) | Inputting of voice | |
| JPS5958490A (en) | Voice recognition system |