• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.2
• /
• pp.28-34
• /
• 1994

In this paper, a new time-domain parameter extraction method for speech recognition is proposed. The suggested emthod is based on the fact that the local peak-to-peak interval, i.e., the interval between maxima and minima of speech waveform is closely related to the frequency component of the speech signal. The parameterization is achieved by a sort of filter bank technique in the time domain. To test the proposed parameter extraction emthod, an isolated word recognizer based on Vector Quantization and Hidden Markov Model was constructed. As a test material, 22 words spoken by ten males were used and the recognition rate of 92.9% was obtained. This result leads to the conclusion that the new parameter extraction method can be used for speech recognition system. Since the proposed method is processed in the time domain, the real-time parameter extraction can be implemented in the class of personal computer equipped onlu with an A/D converter without any DSP board.

• MALSORI
• /
• no.60
• /
• pp.145-164
• /
• 2006

This paper proposes class histogram equalization (CHEQ) to compensate noisy acoustic features for robust speech recognition. CHEQ aims to compensate for the acoustic mismatch between training and test speech recognition environments as well as to reduce the limitations of the conventional histogram equalization (HEQ). In contrast to HEQ, CHEQ adopts multiple class-specific distribution functions for training and test environments and equalizes the features by using their class-specific training and test distributions. According to the class-information extraction methods, CHEQ is further classified into two forms such as hard-CHEQ based on vector quantization and soft-CHEQ using the Gaussian mixture model. Experiments on the Aurora 2 database confirmed the effectiveness of CHEQ by producing a relative word error reduction of 61.17% over the baseline met-cepstral features and that of 19.62% over the conventional HEQ.

• MALSORI
• /
• no.61
• /
• pp.113-123
• /
• 2007

The environmental differences between training and testing mode are generally considered to be the critical factor for the performance degradation in speaker recognition systems. Especially, general speaker recognition systems try to get as clean speech as possible to train the speaker model, but it's not true in real testing phase due to environmental and channel noise. So in this paper, the new method of weighting the frame-based likelihood according to frame SNR is proposed in order to cope with that problem. That is to make use of the deep correlation between speech SNR and speaker discrimination rate. To verify the usefulness of this proposed method, it is applied to the context dependent speaker identification system. And the experimental results with the cellular phone speech DB which is designed by ETRI for Koran speaker recognition show that the proposed method is effective and increase the identification accuracy by 11% at maximum.

• MALSORI
• /
• no.46
• /
• pp.117-126
• /
• 2003

This paper presents our language model adaptation for Korean spontaneous speech recognition. Korean spontaneous speech is observed various characteristics of content and style such as filled pauses, word omission, and contraction as compared with the written text corpus. Our approaches focus on improving the estimation of domain-dependent n-gram models by relevance weighting out-of-domain text data, where style is represented by n-gram based tf/sup */idf similarity. In addition to relevance weighting, we use disfluencies as Predictor to the neighboring words. The best result reduces 9.7% word error rate relatively and shows that n-gram based relevance weighting reflects style difference greatly and disfluencies are good predictor also.

• Speech Sciences
• /
• v.11 no.4
• /
• pp.43-52
• /
• 2004

Model parameters in HMM based speech recognition systems are normally estimated using Maximum Likelihood Estimation(MLE). The MLE method is based mainly on the principle of statistical data fitting in terms of increasing the HMM likelihood. The optimality of this training criterion is conditioned on the availability of infinite amount of training data and the correct choice of model. However, in practice, neither of these conditions is satisfied. In this paper, we propose a training algorithm, MCE(Minimum Classification Error), to improve the performance of a speech recognizer detecting mispronunciation of a foreign language. During the conventional MLE(Maximum Likelihood Estimation) training, the model parameters are adjusted to increase the likelihood of the word strings corresponding to the training utterances without taking account of the probability of other possible word strings. In contrast to MLE, the MCE training scheme takes account of possible competing word hypotheses and tries to reduce the probability of incorrect hypotheses. The discriminant training method using MCE shows better recognition results than the MLE method does.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.1
• /
• pp.24-31
• /
• 1994

In this paper, we report on the development of a speaker independent continuous speech recognition system using continuous hidden Markov models. The continuous hidden Markov model consists of mean and covariance matrices and directly models speech signal parameters, therefore does not have quantization error. Filter bank coefficients with their 1st and 2nd-order derivatives are used as feature vectors to represent the dynamic features of speech signal. We use the segmental K-means algorithm as a training algorithm and triphone as a recognition unit to alleviate performance degradation due to coarticulation problems critical in continuous speech recognition. Also, we use the one-pass search algorithm that Is advantageous in speeding-up the recognition time. Experimental results show that the system attains the recognition accuracy of $83\%$ without grammar and $94\%$ with finite state networks in speaker-indepdent speech recognition.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.4
• /
• pp.31-36
• /
• 2000

This paper describes the implementation and the evaluation of the speech recognition automatic exchange system. The system provides government or public offices, companies, educational institutions that are composed of large number of members and parts with exchange service using speech recognition technology. The recognizer of the system is a Speaker-Independent, Isolated-word, Flexible-Vocabulary recognizer based on SCHMM(Semi-Continuous Hidden Markov Model). For real-time implementation, DSP TMS320C32 made in Texas Instrument Inc. is used. The system operating terminal including the diagnosis of speech recognition DSP and the alternation of speech recognition candidates makes operation easy. In this experiment, 8 speakers pronounced words of 1,300 vocabulary related to automatic exchange system over wire telephone network and the recognition system achieved 91.5% of word accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.3
• /
• pp.51-56
• /
• 2000

Is this paper, we studied on the automatic speech control system in real-time windows environment using voice recognition. The applied reference pattern is the variable DMS model which is proposed to fasten execution speed and the one-stage DP algorithm using this model is used for recognition algorithm. The recognition vocabulary set is composed of control command words which are frequently used in windows environment. In this paper, an automatic speech period detection algorithm which is for on-line voice processing in windows environment is implemented. The variable DMS model which applies variable number of section in consideration of duration of the input signal is proposed. Sometimes, unnecessary recognition target word are generated. therefore model is reconstructed in on-line to handle this efficiently. The Perceptual Linear Predictive analysis method which generate feature vector from extracted feature of voice is applied. According to the experiment result, but recognition speech is fastened in the proposed model because of small loud of calculation. The multi-speaker-independent recognition rate and the multi-speaker-dependent recognition rate is 99.08% and 99.39% respectively. In the noisy environment the recognition rate is 96.25%.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.8
• /
• pp.498-504
• /
• 2005

We introduce a Korean speech recognition platform (ECHOS) developed for education and research Purposes. ECHOS lowers the entry barrier to speech recognition research and can be used as a reference engine by providing elementary speech recognition modules. It has an easy simple object-oriented architecture, implemented in the C++ language with the standard template library. The input of the ECHOS is digital speech data sampled at 8 or 16 kHz. Its output is the 1-best recognition result. N-best recognition results, and a word graph. The recognition engine is composed of MFCC/PLP feature extraction, HMM-based acoustic modeling, n-gram language modeling, finite state network (FSN)- and lexical tree-based search algorithms. It can handle various tasks from isolated word recognition to large vocabulary continuous speech recognition. We compare the performance of ECHOS and hidden Markov model toolkit (HTK) for validation. In an FSN-based task. ECHOS shows similar word accuracy while the recognition time is doubled because of object-oriented implementation. For a 8000-word continuous speech recognition task, using the lexical tree search algorithm different from the algorithm used in HTK, it increases the word error rate by $40\%$ relatively but reduces the recognition time to half.

• Speech Sciences
• /
• v.10 no.1
• /
• pp.71-84
• /
• 2003

In the acoustic modeling for large vocabulary speech recognition, a sparse data problem caused by a huge number of context-dependent (CD) models usually leads the estimated models to being unreliable. In this paper, we develop a new clustering method based on the C45 decision-tree learning algorithm that effectively encapsulates the CD modeling. The proposed scheme essentially constructs a supervised decision rule and applies over the pre-clustered triphones using the C45 algorithm, which is known to effectively search through the attributes of the training instances and extract the attribute that best separates the given examples. In particular, the data driven method is used as a clustering algorithm while its result is used as the learning target of the C45 algorithm. This scheme has been shown to be effective particularly over the database of low unknown-context ratio in terms of recognition performance. For speaker-independent, task-independent continuous speech recognition task, the proposed method reduced the percent accuracy WER by 3.93% compared to the existing rule-based methods.