• The Journal of the Acoustical Society of Korea
• /
• v.19 no.5
• /
• pp.13-19
• /
• 2000

The blind segmentation method, which segments input speech data into recognition unit without any prior knowledge, plays an important role in continuous speech recognition system and corpus generation. As no prior knowledge is required, this method is rather simple to implement, but in general, it suffers from bad performance when compared to the knowledge-based segmentation method. In this paper, we introduce a method to improve the performance of a blind segmentation of Korean continuous speech by postprocessing the segment boundaries obtained from the blind segmentation. In the preprocessing stage, the candidate boundaries are extracted by a clustering technique based on the GLR(generalized likelihood ratio) distance measure. In the postprocessing stage, the final phoneme boundaries are selected from the candidates by utilizing a simple a priori knowledge on the syllabic structure of Korean, i.e., the maximum number of phonemes between any consecutive nuclei is limited. The experimental result was rather promising : the proposed method yields 25% reduction of insertion error rate compared that of the blind segmentation alone.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.1
• /
• pp.23-27
• /
• 2003

Recently attempts to we the syllable as the recognition unit to enhance performance in continuous speech recognition hate been reported. However, syllables are worse in their trainability than phones and the former have a disadvantage in that contort-dependent modeling is difficult across the syllable boundary since the number of models is much larger for syllables than for phones. In this paper, we propose a method to enhance the trainability for the syllables in Korean and phoneme-context dependent syllable modeling across the syllable boundary. An experiment in which the proposed method is applied to word recognition shows average 46.23% error reduction in comparison with the common syllable modeling. The right phone dependent syllable model showed 16.7% error reduction compared with a triphone model.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.1
• /
• pp.11-15
• /
• 2000

Speaker adaptive continuous speech recognition for the RM speech corpus is described in this paper. Learning of hidden markov models for the reference speaker is performed for the training data of RM corpus. For the evaluation, evaluation data of RM corpus are used. Parts of another training data of RM corpus are used for the speaker adaptation. After dynamic time warping of another speaker's data for the reference data is accomplished, error back propagation neural network is used to transform the spectrum between speakers to be recognized and reference speaker. Experimental results to get the best adaptation by tuning the neural network are described. The recognition ratio after adaptation is substantially increased 2.1 times for the word recognition and 4.7 times for the word accuracy for the best.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.8
• /
• pp.105-113
• /
• 1996

In this paper, a new korean digita speech recognition technique was proposed using muktolayer perceptron (MLP). In spite of its weakness in dynamic signal recognition, MLP was adapted for this model, cecause korean syllable could give static features. It is so simle in its structure and fast in its computing that MLP was used to the suggested system. MLP's input vectors was transformed using karhunen-loeve transformation (KLT), which compress signal successfully without losin gits separateness, but its physical properties is changed. Because the suggested technique could extract static features while it is not affected from the changes of syllable lengths, it is effectively useful for korean numeric recognition system. Without decreasing classification rates, we can save the time and memory size for computation using KLT. The proposed feature extraction technique extracts same size of features form the tow same parts, front and end of a syllable. This technique makes frames, where features are extracted, using unique size of windows. It could be applied for continuous speech recognition that was not easy for the normal neural network recognition system.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.2
• /
• pp.49-57
• /
• 1997

In this paper, we propose the stochastic pronunciation lexicon model for large vocabulary continuous speech recognition system. We can regard stochastic lexicon as HMM. This HMM is a stochastic finite state automata consisting of a Markov chain of subword states and each subword state in the baseform has a probability distribution of subword units. In this method, an acoustic representation of a word can be derived automatically from sample sentence utterances and subword unit models. Additionally, the stochastic lexicon is further optimized to the subword model and recognizer. From the experimental result on 3000 word continuous speech recognition, the proposed method reduces word error rate by 23.6% and sentence error rate by 10% compare to methods based on standard phonetic representations of words.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.26 no.2
• /
• pp.117-121
• /
• 2015

Background : Cepstral analysis which is obtained from Fourier transformation of spectrum has been known to be effective indicator to analyze the voice disorder. To evaluate the voice disorder, phonation of sustained vowel /a/ sound or continuous speech have been used but the former was limited to capture hoarseness properly. This study is aimed to compare the effectiveness in analysis of cepstrum between the sustained vowel /a/ sound and continuous speech. Methods : From March 2012 to December 2014, total 72 patients was enrolled in this study, including 24 unilateral vocal cord palsy, vocal nodule and vocal polyp patients, respectively. The entire patient evaluated their voice quality by VHI (Voice Handicap Index) before and after treatment. Phonation of sustained vowel /a/ sample and continuous speech using the first sentence of autumn paragraph was subjected by cepstral analysis and compare the pre-treatment group and post-treatment group. Results : The measured values of pre and post treatment in CPP-a (cepstral peak prominence in /a/ vowel sound) was 13.80, 13.91 in vocal cord palsy, 16.62, 17.99 in vocal cord nodule, 14.19, 18.50 in vocal cord polyp respectively. Values of CPP-s (cepstral peak prominence in text-based speech) in pre and post treatment was 11.11, 12.09 in vocal cord palsy, 12.11, 14.09 in vocal cord nodule, 12.63, 14.17 in vocal cord polyp. All 72 patients showed subjective improvement in VHI after treatment. CPP-a showed statistical improvement only in vocal polyp group, but CPP-s showed statistical improvement in all three groups (p<0.05). Conclusion : In analysis of cepstrum, text-based analysis is more representative in voice disorder than vowel sound speech. So when the acoustic analysis of voice by cepstrum, both phonation of sustained vowel /a/ sound and text based speech should be performed to obtain more accurate result.

• Proceedings of the IEEK Conference
• /
• 1999.06a
• /
• pp.1107-1110
• /
• 1999

In speech signal processing, the accurate decision of the voiced/unvoiced sound is important for robust word recognition and analysis and a high coding efficiency. In this paper, we propose the mehod of the voiced/unvoiced decision using the LSP parameter which represents the spectrum characteristics of the speech signal. The voiced sound has many more LSP parameters in low frequency region. To the contrary, the unvoiced sound has many more LSP parameters in high frequency region. That is, the LSP parameter distribution of the voiced sound is different to that of the unvoiced sound. Also, the voiced sound has the minimun value of sequantial intervals of the LSP parameters in low frequency region. The unvoiced sound has it in high frequency region. we decide the voiced/unvoiced sound by using this charateristics. We used the proposed method to some continuous speech and then achieved good performance.

• Phonetics and Speech Sciences
• /
• v.6 no.3
• /
• pp.155-164
• /
• 2014

This paper proposes a new method to determine the recognition units for large vocabulary continuous speech recognition (LVCSR) in Korean by applying unsupervised segmentation and merging. In the proposed method, a text sentence is segmented into morphemes and position information is added to morphemes. Then submorpheme units are obtained by splitting the morpheme units through the maximization of posterior probability terms. The posterior probability terms are computed from the morpheme frequency distribution, the morpheme length distribution, and the morpheme frequency-of-frequency distribution. Finally, the recognition units are obtained by sequentially merging the submorpheme pair with the highest frequency. Computer experiments are conducted using a Korean LVCSR with a 100k word vocabulary and a trigram language model obtained by a 300 million eojeol (word phrase) corpus. The proposed method is shown to reduce the out-of-vocabulary rate to 1.8% and reduce the syllable error rate relatively by 14.0%.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.7
• /
• pp.59-68
• /
• 1998

This paper is to develop an algorithm for the unlimited korean speech synthesis. We present the results controlled of prosodic factors with isolated words as aynthesis basis unit int he time domain. With a new pitch-synchronous and parametric speech synthesis mehtod in the time domain here we mainly present the results of controlled prosody factors such a spitch periods, energy envelops and durations and the evaluaton of synthetic speech qualities. In the case of synthesis, it is possible ot synthesize connected words by controlling of a continuous unified prosody that makes to improve the naturalities. In the results of experiment, it also has been to be improved uncontinuities of pitch and zeroing of energy in the junction parts of speech waveforms. Specially it has been to be possible to synthesize speeches with unlimitted durations and tones. So on it makes the noisiness and the clearness better by improving the degradation effects from the phase distortion due to the discontinuities in the waveform connection parts.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.4E
• /
• pp.11-16
• /
• 1998

For large vocabulary speech recognition using HMMs, context-dependent subword units have been often employed. However, when context-dependent phone models are used, they result in a system which has too may parameters to train. The problem of too many parameters and too little training data is absolutely crucial in the design of a statistical speech recognizer. Furthermore, when building large vocabulary speech recognition systems, unseen triphone problem is unavoidable. In this paper, we propose the modified phonetic decision tree algorithm for the automatic prediction of unseen triphones which has advantages solving these problems through following two experiments in Japanese contexts. The baseline experimental results show that the modified tree based clustering algorithm is effective for clustering and reducing the number of states without any degradation in performance. The task experimental results show that our proposed algorithm also has the advantage of providing a automatic prediction of unseen triphones.