KR970050120A - Speech Recognition Method in Word Spotting System - Google Patents

Speech Recognition Method in Word Spotting System Download PDF

Info

Publication number
KR970050120A
KR970050120A KR1019950067362A KR19950067362A KR970050120A KR 970050120 A KR970050120 A KR 970050120A KR 1019950067362 A KR1019950067362 A KR 1019950067362A KR 19950067362 A KR19950067362 A KR 19950067362A KR 970050120 A KR970050120 A KR 970050120A
Authority
KR
South Korea
Prior art keywords
similar
model
speech recognition
word
voice
Prior art date
Application number
KR1019950067362A
Other languages
Korean (ko)
Inventor
이승배
Original Assignee
구자홍
Lg 전자주식회사
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 구자홍, Lg 전자주식회사 filed Critical 구자홍
Priority to KR1019950067362A priority Critical patent/KR970050120A/en
Publication of KR970050120A publication Critical patent/KR970050120A/en

Links

Landscapes

  • Document Processing Apparatus (AREA)

Abstract

본 발명의 워드 스포팅 시스템에서의 음성인식방법은, 등록단어와 유사하게 발음되는 음성에 대한 별도의 유사음소모델을 만들어서 그 자체를 필라로 이용하거나 상기 등록단어와 같은 음소순서로 병합하여 유사단어모델로 이용함으로써 등록단어와 비슷한 음성이 발음되어도 등록단어와 구분하여 오인식으로 인한 검출결과를 줄임으로써 워드 스포팅이 응용되는 승차권발매기, 자동판매기, 항공기 예약 서비스, 은행 서비스 등과 같이 특정 등록단어만을 인식하여도 무방한 자동음성인식장치에 적용하는 경우 매우 유용하게 인식성능을 향상시킬 수 있다.In the word spotting system of the present invention, a similar phonetic model for a voice that is pronounced similarly to a registered word may be used as a pillar or merged into a phoneme sequence such as the registered word. Even if a voice similar to a registered word is pronounced, it can be distinguished from the registered word to reduce the detection result due to a misrecognition. When applied to any automatic speech recognition device can be very useful to improve the recognition performance.

Description

워스 스포팅 시스템에서의 음성인식방법Speech Recognition in Worth Spotting System

본 내용은 요부공개 건이므로 전문내용을 수록하지 않았음As this is a public information case, the full text was not included.

제4도는 본 발명에 의한 유사음소모델의 HMM 훈련과정을 도시한 것이다.Figure 4 shows the HMM training process of the pseudophoneme model according to the present invention.

Claims (4)

등록단어와 유사하게 발음되어 오인식이 발생할 수 있는 음성에대한 별도의 유사음소모델을 만들어서 상기 등록단어와 유사한 음성이 발음되어도 등록단어로 오인식하지 않도록 하는 것을 특징으로 하는 워드 스포팅 시스템에서의 음성인식방법.The speech recognition method of the word spotting system, characterized in that a similar phonetic model for a voice that is pronounced similarly to a registered word is generated so that a similar voice is not recognized as a registered word even when a voice similar to the registered word is pronounced. . 제1항에 있어서, 상기 유사음소모델은, 훈련용 음성 데이타를 선택하여 특징을 추출하고 상기 훈련용 음성 데이타와 유사하게 발음되어 오인식이 발생할 수 있는 유사음소에 대한 모델을 훈련시키기 위한 데이타를 생성하는 과정과, 상기 유사음소의 HMM 모델을 훈련시키는 과정으로 만들어지는 것을 특징으로 하는 워드 스포팅 시스템에서의 음성인식방법.The method of claim 1, wherein the similar phoneme model is configured to extract data by selecting training voice data and to generate data for training a model for similar phonemes that may be pronounced similarly to the training voice data so that misrecognition may occur. And a step of training the HMM model of the similar phoneme. 제1항 또는 제2항에 있어서, 상기 유사음소모델은 그 자체를 필라로 이용하는 것을 특징으로 하는 워드 스포팅 시스템에서의 음성인식방법.The speech recognition method of claim 1, wherein the pseudo phoneme model uses itself as a pillar. 제1항 또는 제2항에 있어서, 상기 유사음소모델은 상기 등록단어와 같은 음소순서로 병합하여 유사단어모델로 이용하는 것을 특징으로 하는 워드 스포팅 시스템에서의 음성인식방법.The speech recognition method of claim 1 or 2, wherein the similar phoneme model is used as a similar word model by merging in the same phonetic order as the registered word. ※ 참고사항 : 최초출원 내용에 의하여 공개하는 것임.※ Note: The disclosure is based on the initial application.
KR1019950067362A 1995-12-29 1995-12-29 Speech Recognition Method in Word Spotting System KR970050120A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1019950067362A KR970050120A (en) 1995-12-29 1995-12-29 Speech Recognition Method in Word Spotting System

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1019950067362A KR970050120A (en) 1995-12-29 1995-12-29 Speech Recognition Method in Word Spotting System

Publications (1)

Publication Number Publication Date
KR970050120A true KR970050120A (en) 1997-07-29

Family

ID=66637409

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1019950067362A KR970050120A (en) 1995-12-29 1995-12-29 Speech Recognition Method in Word Spotting System

Country Status (1)

Country Link
KR (1) KR970050120A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100484493B1 (en) * 2002-12-12 2005-04-20 한국전자통신연구원 Spontaneous continuous speech recognition system and method using mutiple pronunication dictionary

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100484493B1 (en) * 2002-12-12 2005-04-20 한국전자통신연구원 Spontaneous continuous speech recognition system and method using mutiple pronunication dictionary

Similar Documents

Publication Publication Date Title
KR970707529A (en) SPEECH RECOGNITION FOR SPEECH RECOGNITION DEVICE AND SPEECH RECOGNITION DEVICE
KR870009322A (en) Speaker array language recognition system
US5425129A (en) Method for word spotting in continuous speech
EP0867858A3 (en) Pronunciation generation in speech recognition
EP0874353A3 (en) Pronunciation generation in speech recognition
DE69822179D1 (en) METHOD FOR LEARNING PATTERNS FOR VOICE OR SPEAKER RECOGNITION
EP0735736A3 (en) Method for automatic speech recognition of arbitrary spoken words
ES2086345T3 (en) METHOD FOR THE RECOGNITION OF THE SPEECH ADAPTABLE TO THE USER.
ATE203119T1 (en) LANGUAGE RECOGNITION SYSTEM FOR COMPOUND WORD LANGUAGES
Byrne et al. Pronunciation modelling using a hand-labelled corpus for conversational speech recognition
Kurzekar et al. Continuous speech recognition system: A review
Hon et al. Towards large vocabulary Mandarin Chinese speech recognition
Lamel et al. Continuous speech recognition at LIMSI
Price et al. Combining linguistic with statistical methods in modeling prosody
Cutler The recognition of spoken words with variable representations
US7430503B1 (en) Method of combining corpora to achieve consistency in phonetic labeling
KR970050120A (en) Speech Recognition Method in Word Spotting System
Gauvain et al. Identification of non-linguistic speech features
Kepuska et al. Wake-up-word speech recognition application for first responder communication enhancement
Elhadj et al. An accurate recognizer for basic arabic sounds
Deekshitha et al. Speech Signal Based Broad Phoneme Classification and Search Space Reduction for Spoken Term Detection
Takahashi et al. Interactive voice technology development for telecommunications applications
KR970050115A (en) Speech Recognition Method of Variation Unit using Korean Variation Grouping Tree
JP3110025B2 (en) Utterance deformation detection device
KR930014267A (en) Speaker dependent speech recognition system and voice matching method

Legal Events

Date Code Title Description
A201 Request for examination
E902 Notification of reason for refusal
E601 Decision to refuse application