• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 10월 학술대회지
• /
• pp.145-148
• /
• 2003

To improve the recognition performance of Korean connected digit telephone speech, in this paper, both Aurora feature extraction method that employs noise reduction 2-state Wiener filter and DWFBA method are investigated and used. CMN and MRTCN are applied to static features for channel compensation. Telephone digit speech database released by SITEC is used for recognition experiments with HTK system. Experimental results has shown that Aurora feature is slightly better than MFCC and DWFBA without channel compensation. And when channel compensation is included, Aurora feature is slightly better than DWFBA with MRTCN.

• 한국음향학회지
• /
• 제21권4호
• /
• pp.401-406
• /
• 2002

한국어 숫자는 모두 단음절로 이루어져 있으며, 연속적으로 발음될 때 인접 숫자들의 상호조음현상에 의해 각 숫자의 고유 발음이 변화하고, 또한 그 숫자들의 경계도 모호해지는 문제점이 있다. 이러한 문제점들과 더불어 배경잡음이나 채널에 의한 왜곡에 따른 문제점들로 인해 한국어 연결숫자의 인식 성능은 만족스럽지 못한 것이 현실이다. 본 논문에서는 연결숫자의 인식성능 향상을 위해서 한국어 숫자들의 음운변화를 고려하여 유사음소 (phonelike units: PLUs)군을 정의하고, 사용자의 여러 가지 발성형태에 따른 다양한 음운 현상의 변화를 흡수할 수 있도록 인식 시스템을 구성하는 방식을 검토하였다. 전화망 4연숫자를 이용한 화자독립 인식 실험을 수행한 결과 제안된 방법의 숫자열 인식률은 상태당 믹스쳐 (mixture) 개수가 1인 경우 83.2%로, 기준 시스템 (baseline)에 대한 오류감소률이 7.2%였고 가장 높은 성능을 나타낸 믹스쳐 개수가 11인 경우 숫자열 인식률은 91.8% 오류감소율은 4.7%였다.

• 한국음향학회지
• /
• 제23권2호
• /
• pp.187-191
• /
• 2004

연결 단어 인식기, 특히 연결 숫자음 인식기를 제작하기 위한 음성 데이터베이스를 구축하는데 있어서 완전하고 효율적인 발성목록을 작성하기 위한 알고리즘을 제안한다. 기존의 음성 DB에서 사용되는 목록은 주로 난수 발생기에 의하여 만들어지거나 사용자의 전화번호, 우편번호 등을 이용하여 만들어져 왔으므로 다양한 환경의 음소 또는 단어를 균일하게 포함하고 있지 못하다. 따라서 본 논문에서는 하나의 단어에 대하여 전후에 모든 단어가 연결되는 조합을 모두 한번씩 포함하는 목록을 만드는 효율적인 알고리즘을 제안한다. 본 알고리즘으로 7연 숫자 목록을 만들면 200개의 문장으로 모든 조합을 포함할 수 있게 된다. 본 논문에서는 알고리즘 예제와 본 알고리즘의 완전성과 효율성에 대하여 기술하였다.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2002년도 11월 학술대회지
• /
• pp.201-204
• /
• 2002

As a preliminary study for improving recognition performance of the connected digit telephone speech, we investigate feature parameters as well as channel compensation methods of telephone speech. The CMN and RTCN are examined for telephone channel compensation, and the MFCC, DWFBA, SSC and their delta-features are examined as feature parameters. Recognition experiments with database we collected show that in feature level DWFBA is better than MFCC and for channel compensation RTCN is better than CMN. The DWFBA+Delta_ Mel-SSC feature shows the highest recognition rate.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 5월 학술대회지
• /
• pp.119-122
• /
• 2003

In general, triangular shape filters are used in the filter bank when we get the MFCCs from the spectrum of speech signal. In [1], a new feature extraction approach is proposed, which uses specific filter shapes in the filter bank that are obtained from the spectrum of training speech data. In this approach, principal component analysis technique is applied to the spectrum of the training data to get the filter coefficients. In this paper, we carry out speech recognition experiments, using the new approach given in [1], for a large amount of telephone speech data, that is, the telephone speech database of Korean connected digit released by SITEC. Experimental results are discussed with our findings.

• 대한음성학회지:말소리
• /
• 제44호
• /
• pp.105-113
• /
• 2002

In Korean, each digit is monosyllable and some pairs are known to have high confusability, causing performance degradation of connected digit recognition systems. To improve the performance, in this paper, we employ various discriminant analyses (DA) including Linear DA (LDA), Weighted Pairwise Scatter LDA WPS-LDA), Heteroscedastic Discriminant Analysis (HDA), and Maximum Likelihood Linear Transformation (MLLT). We also examine several combinations of various DA for additional performance improvement. Experimental results show that applying any DA mentioned above improves the string accuracy, but the amount of improvement of each DA method varies according to the model complexity or number of mixtures per state. Especially, more than 20% of string error reduction is achieved by applying MLLT after WPS-LDA, compared with the baseline system, when class level of DA is defined as a tied state and 1 mixture per state is used.

• 대한음성학회지:말소리
• /
• 제45호
• /
• pp.79-91
• /
• 2003

Utterance verification aims at rejecting both out-of-vocabulary (OOV) utterances and low-confidence-scored in-vocabulary (IV) utterances. For utterance verification on Korean connected digit recognition task, we investigate several methods to construct filler and anti-digit models. In particular, we propose a substitution error correction method based on 2-best decoding results. In this method, when 1st candidate is rejected, 2nd candidate is selected if it is accepted by a specific hypothesis test, instead of simply rejecting the 1st one. Experimental results show that the proposed method outperforms the conventional log likelihood ratio (LLR) test method.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 10월 학술대회지
• /
• pp.149-152
• /
• 2003

In this paper, data-driven temporal filter methods[1] are investigated for robust feature extraction. A principal component analysis technique is applied to the time trajectories of feature sequences of training speech data to get appropriate temporal filters. We did recognition experiments on the Korean connected digit telephone speech database released by SITEC, with data-driven temporal filters. Experimental results are discussed with our findings.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 5월 학술대회지
• /
• pp.145-148
• /
• 2003

In this paper, we carried out comparative study about various feature parameters for the effective speaker recognition such as LPC, LPCC, MFCC, Log Area Ratio, Reflection Coefficients, Inverse Sine, and Delta Parameter. We also adopted cepstral liftering and cepstral mean subtraction methods to check their usefulness. Our recognition system is HMM based one with 4 connected-Korean-digit speech database. Various experimental results will help to select the most effective parameter for speaker recognition.

• 한국음향학회:학술대회논문집
• /
• 한국음향학회 1998년도 제15회 음성통신 및 신호처리 워크샵(KSCSP 98 15권1호)
• /
• pp.250-253
• /
• 1998

In digital communicatioin networks, speech recognition systems conventionally reconstruct speech followed by extracting feature [parameters. In this paper, we consider a useful approach by incorporating speech coding parameters into the speech recognizer. Most speech coders employed in the networks represent line spectral pairs as spectral parameters. In order to improve the recognition performance of the LSP-based speech recognizer, we introduce two different ways: one is to devise weighed distance measures of LSPs and the other is to transform LSPs into a new feature set, named a pseudo-cepstrum. Experiments on speaker-independent connected-digit recognition showed that the weighted distance measures significantly improved the recognition accuracy than the unweighted one of LSPs. Especially we could obtain more improved performance by using PCEP. Compared to the conventional methods employing mel-frequency cepstral coefficients, the proposed methods achieved higher performance in recognition accuracies.