• The Journal of the Acoustical Society of Korea
• /
• v.22 no.5
• /
• pp.403-407
• /
• 2003

Recently, eigenvoice approach has been widely used for rapid speaker adaptation. However, even in the eigenvoice approach, Performance improvement using very small amount of adaptation data is relatively small in comparison with that using somewhat large adaptation data because the reliable estimation of weights of eigenvoice is difficult. In this paper, we propose a rapid speaker adaptation method based on eigenvoice using the weight distribution characteristics to improve the performance on a small adaptation data. In the Experimental results on vocabulary-independent word recognition task (using PBW 452 database), the weight threshold method alleviates the problem of relatively low performance for a tiny small adaptation data. When single adaptation word is used, word error rate is reduced about 9-18% by the weight threshold method.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.2
• /
• pp.84-89
• /
• 1998

본 논문에서는 임의의 단어를 인식하기 위하여 음성학적으로 최적화된 (phonetically-optimized word) 음성 데이터베이스를 사용하여 훈련된 가변어휘 고립단위 음 성인식기의 실제 인식기 사용 환경에서의 성능을 평가하였다. 이를 위하여, 훈련 데이터베이 스에서와 상이한 환경에서 수집된 음성학적으로 균형 잡힌(phonetically-balanced word) 고 립 단어 음성을 테스트 데이터로 사용하였다. 테스트 데이터는 일반적인 사무실에서 작동하 는 노트북 PC에서 내장 마이크를 사용하여 녹음되었다. 이렇게 녹음된 음성을 사용하여 고 립단어 인식기의 인식률을 측정하였다. 이 인식기는 최대 사후(maximum a posteriori) 추정 알고리듬을 사용하여 화자의 변화에 적응하였다. 컴퓨터 모의실험 결과에 의하면 화자 적응 을 하지 않은 기본 시스템은 깨끗한 음성에 대하여 81.3%에서 사무실 환경 음성에 대하여 69.8%로 인식률이 저하되었다. 사무실 환경 음성에 대하여, 비교사 점진(unsupervised incremental) 모드에서 최대 사후 추정 화자 적응 알고리듬을 적용하였을 경우에는 화자적 응을 하지 않은 경우에 비하여 9%의 에러를 감소시키며, 50단어의 적응 단어를 사용하여 교사 묶음(supervised batch) 모드에서 최대 사후 추정 화자 적응 알고리듬을 적용하였을 경우에는 16%의 에러를 감소시켰다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1998.06e
• /
• pp.309.2-314
• /
• 1998

본 연구에서는 음성인식을 위한 화자적응화 기법에 대해 연구하였다. 첫째로 적응화에 포함되지 않은 카테고리 음절에 대해 적응화 효과를 줄 수 있는 보간적응화 방법에 대해 연구하였다. 표준모델과 소량의 음성 데이터만으로 적응화가 가능한 MAPE(최대사후확률추정)으로 적응화한 모델의 평균벡터 변화정도를 적응화 발화에 포함되지 않은 모델에 보간적응하는 방법이다. 둘째로 음절단위 모델을 구축한 후 적응화 하고자 하는 화자의 데이터를 연결학습법과 Viterbi 알고리즘으로 음절단위의 추출을 자동화 한 후 MAPE으로 적응화하는 방법에 대해 각각 실험을 하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.5 no.2
• /
• pp.90-97
• /
• 2000

This paper describes an speaker-adaptive speech recognition system which make a reliable recognition of speech signal for new speakers. In the Proposed method, an speech spectrum of new speaker is adapted to the reference speech spectrum by using Parameters of a 1st linear transformation network at the front of phoneme classification neural network. And the recognition system is based on semicontinuous HMM(hidden markov model) which use the multilayer perceptron as a fuzzy vector quantizer. The experiments on the isolated word recognition are performed to show the recognition rate of the recognition system. In the case of speaker adaptation recognition, the recognition rate show significant improvement for the unadapted recognition system.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.3
• /
• pp.40-52
• /
• 1992

In this paper, we propose the speaker adaptive word recognition method using a mapped membership function, in order to absorb a fluctuation owing to personal difference which is a problem of speaker independent speech recognition. In the training procedure of this method, the mapped membership function is made with the fuzzy theory introducded into a mapped codebook, between an unknown speaker's spectrum pattern and a standard speaker's one. In the recognition procedure, an input pattern of an unknown speaker is reconstructed to the pattern which is adapted to that of a standard speaker by the mapped membership function. To show the validity of this method, word recognition experiments are carried out using 28 DDD area names. The recognition rate of the conventional speaker-adaptive method using a mapped codebook by VQ is 64.9[%], and that made by a fuzzy VQ is 76.2[%]. Throughout the experiment using a mapped membership function, we can achieve 95.4[%] recognition rate. This shows that our proposed method is more excellent in recognition performance. Moreover, this method doesn't need an iterative training procedure to make the mapped membership function, and memory capacity and computation requirements for this method are reduced to 1/30 and 1/500 time of those for the conventional method using a mapped codebook, respectively.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2003.06a
• /
• pp.6-9
• /
• 2003

In this paper, we studied on the difference of speaker adaptation according to the phoneme classification for Korean Speech recognition. In order to study of speech adaptation according to the weight of difference of phoneme as recognition unit, we used SCHMM as recognition system. And Speaker adaptation method used in this paper was MAPE(Maximum A Posteriori Probability Estimation), Linear Spectral Estimation. In order to evaluate the performance of these methods, we used 10 Korean isolated numbers as the experimental data. It is possible for the first and the second methods to be carried out unsupervised learning and used in on-line system. And the first method was shown performance improvement over the second method, and hybrid adaptation showed the better recognition results than those which performed each method. And the result of Speaker adaptation using the variable weight according to the phoneme had better than the result using fixed weight.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.3
• /
• pp.268-272
• /
• 2009

In this paper, we propose a Gaussian Mixture Model (GMM) based incremental robust adaptation with a forgetting factor for the speaker verification. Speaker recognition system uses a speaker model adaptation method with small amounts of data in order to obtain a good performance. However, a conventional adaptation method has vulnerable to the outlier from the irregular utterance variations and the presence noise, which results in inaccurate speaker model. As time goes by, a rate in which new data are adapted to a model is reduced. The proposed algorithm uses an incremental robust adaptation in order to reduce effect of outlier and use forgetting factor in order to maintain adaptive rate of new data on GMM based speaker model. The incremental robust adaptation uses a method which registers small amount of data in a speaker recognition model and adapts a model to new data to be tested. Experimental results from the data set gathered over seven months show that the proposed algorithm is robust against outliers and maintains adaptive rate of new data.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.3 no.4
• /
• pp.3-20
• /
• 1993

본 연구에서는 퍼지사상화(fuzzy mapping)에 의한 사상된(mapped) 코드북을 사용하는 화자적은 음성합성 알고리즘을 제안한다. 입력화자와 기준화자의 코드북은 신경망 클러스터링 알고리즘인 자율경쟁 학습을 사용하여 작성된다. 사상된 코드북은 입력 음성벡터에 대한 두 화자의 대응 코드벡터의 소속갑(membership value)으로 퍼지 히스토그랩을 작성하여 이들을 1차 결합함으로써 얻어지는 퍼지사상화에 의하여 작성된다. 음성합성시에는 사상된 코드북을 사용하여 입력화자의 음것을 퍼지 벡터양자화한 다음, CFM 연산으로 합성함으로써 입력화자에 적응된 합성음을 얻는다. 실험에서 여러 입력화자로 30대의 남성, 20대의 여성음을 사용하였고 기준음석으로 입력음성과는 다른 20대의 여성음성을 사용하였다.실험에 사용된 음성데이타는 문장/안녕하십니까/와/굿모닝/이다. 실험결과는 각각의 입력화자에 기준화자 음성이 적응된 합성음을 얻었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.268-270
• /
• 2002

최근 음성 인식 시스템의 성능 향상을 위해 화자 적응 (speaker adaptation)에 대한 연구가 활발히 진행되고 있다. HMM 기반 인식 시스템의 모델 파라미터를 수정하는 화자 적응의 경우, MAP방법과 MLLR 방법에 대한 연구가 주류를 이루고 있다. 두 방법은 adaptation data의 양에 따라서 서로 다른 성능을 보인다. 본 논문에서는 기존 두 방법을 Maximum-likelihood Estimation(MLE)를 이용하여 화자 적응을 수행하는 방법을 제안한다. 제안한 방법을 KAIST 통신연구실에서 구축한 한국어 도시이름 500단어 인식 시스템에 적용하여 adaptation data의 양에 상관없이 항상 높은 성능을 나타냈으며, 기존의 방법에 대해서 최고 4.37%의 인식률 향상을 보였다.

• The Journal of the Acoustical Society of Korea
• /
• v.15 no.2
• /
• pp.5-11
• /
• 1996

The purpose of this paper is to propose a method improving the performance of a semi-continuous hidden Markov model(SCHMM) speaker adaptation system which uses Bayesian Parameter reestimation approach. The performance of Bayesian speaker adaptation could be degraded in case that the features of a new speaker are severely different from those of a reference codebook. The excessive codewords of the reference codebook still remain after adaptation proess. which cause confusion in recognition process. To solve such problems, the proposed method uses formant information which is extracted from the cepstral coefficients of the reference codebook and adaptation data. The reference codebook is adapted to represent the formant distribution of a new speaker and it is used for Bayesian speaker adaptation as an initial codebook. The proposed method provides accurate correspondence between reference codebook and adaptation data. It was observed that the excessive codewords were not selected during recognition process. The experimental results showed that the proposed method improved the recognition performance.