• The Journal of the Acoustical Society of Korea
• /
• v.18 no.6
• /
• pp.57-64
• /
• 1999

In this paper, we investigated the robust endpoint detection algorithm in noisy environment. A new feature parameter based on a discrete wavelet transform is proposed for word boundary detection of isolated utterances. The sum of standard deviation of wavelet coefficients in the third coarse and weighted first detailed scale is defined as a new feature parameter for endpoint detection. We then developed a new and robust endpoint detection algorithm using the feature found in the wavelet domain. For the performance evaluation, we evaluated the detection accuracy and the average recognition error rate due to endpoint detection in an HMM-based recognition system across several signal-to-noise ratios and noise conditions.

• Journal of the Korean Institute of Telematics and Electronics T
• /
• v.36T no.4
• /
• pp.17-23
• /
• 1999

This paper described a security system using text-independent speaker recognition algorithm. Security system is based on PIC16F84 and sound card. Speaker recognition algorithm applied a k-means based model and weighted cepstrum for speech features. As the experimental results, recognition rate of the training data is 100%, non-training data is 99%. Also false rejection rate is 1%, false acceptance rate is 0% and verification mean error rate is 0.5% for registered 5 persons.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.1
• /
• pp.115-122
• /
• 1999

In this paper, we confirmed that a speech signal is a chaos signal, and in order to use it as a speaker recognition parameter, analyzed chaos dimension. In order to raise speaker identification and pattern recognition, by making up the strange attractor involving an individual's vocal tract characteristics very well and applying fuzzy membership function to correlation dimension, we proposed fuzzy correlation dimension. By estimating the correlation of the points making up an attractor are limited according space dimension value, fuzzy correlation dimension absorbed the variation of the reference pattern attractor and test pattern attractor. Concerning fuzzy correlation dimension, by estimating the distance according to the average value of discrimination error per each speaker and reference pattern, investigated the validity of speaker recognition parameter.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.2
• /
• pp.145-152
• /
• 2003

In this paper, assuming that the score of speech utterance is the product of HMM log likelihood and HMM weight, we propose a new method that HMM weights are adapted iteratively like the general MCE training. The proposed method adjusts HMM weights for better performance using delta coefficient defined in terms of misclassification measure. Therefore, the parameter estimation and the Viterbi algorithms of conventional 1:.um can be easily applied to the proposed model by constraining the sum of HMM weights to the number of HMMs in an HMM set. Comparing with the general segmental MCE training approach, computing time decreases by reducing the number of parameters to estimate and avoiding gradient calculation through the optimal state sequence. To evaluate the performance of HMM-based speech recognizer by weighting HMM likelihood, we perform Korean isolated digit recognition experiments. The experimental results show better performance than the MCE algorithm with state weighting.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.6 no.2
• /
• pp.186-192
• /
• 2011

In this paper, we propose a new procedure for improvement of the linear predictive coding. To reduce the error power incurred by the coding, we interchanged the order of the two procedures of windowing on the signal and linear prediction. This scheme corresponds to LPC extraction with windowed autocorrelation. The proposed method requires more calculational time because it necessitates matrix inversion on more parameters than the conventional technique where an efficient Levinson-Durbin recursive procedure is applicable with smaller parameters. Experimental test over various speech phonemes showed, however, that our procedure yields about 5 % less power distortion compared to the conventional technique. Consequently, the proposed method in this paper is thought to be preferable to the conventional technique as far as the fidelity is concerned. In a separate study of speaker-dependent speech recognition test for 50 isolated words pronounced by 40 people, our approach yielded better performance too.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1993.06a
• /
• pp.125-127
• /
• 1993

오디오 혹은 비디오화의, 방송 고품질전화 등의 음성신호의 전송을 위해 마련된 CCITT Recommendation G.722에 의거 Codec을 구성하고 이를 통과한 연속음성을 CMU의 불특정 화자 연속음성인식 시스템인 SPHINX에 입력하여 인식률을 조사 한 후 CODING전의 인식결과와 비교하였다. 이때 CODEC은 크게 네 부분(Trans Quarature Mirror Filter, Encoder, Decoder, Receive QMF)으로 구성하고 입력음성 데이터는 150화자에 의한 1018문장을 훈련용으로, 140문장을 테스트용으로 하였을 때의 단어 인식률을 인식률로 하였다. 또 이때 특징벡터로는 12차 Melcepstrum 계수를 사용하였다. 인식결과 코딩전(close talk Mic를 이용하여 직접입력)의 단어 인식률이 86.7%인데 비해 코딩후의 인식률은 85.6%로 나타나 약 1%의 인식률 저하를 가져와 코딩으로 인한 Error에 비해 비교적 양호한 결과를 얻을 수 있었다. 인식률 저하의 원인으로서는 코딩시의 BER(Bit Error Rate)에 의한 것으로 생각된다.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.6
• /
• pp.5-14
• /
• 1994

This paper is about an experiment of speaker-independent automatic Korean spoken words recognition using Multi-Layered Perceptron and Error Back-propagation algorithm. The object words are 50 citynames of D.D.D local numbers. 43 of those are 2 syllables and the rest 7 are 3 syllables. The words were not segmented into syllables or phonemes, and some feature components extracted from the words in equal gap were applied to the neural network. That led independent result on the speech duration, and the PARCOR coefficients calculated from the frames using linear predictive analysis were employed as feature components. This paper tried to find out the optimum conditions through 4 differerent experiments which are comparison between total and pre-classified training, dependency of recognition rate on the number of frames and PAROCR order, recognition change due to the number of neurons in the hidden layer, and the comparison of the output pattern composition method of output neurons. As a result, the recognition rate of $89.6\%$ is obtaimed through the research.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.447-453
• /
• 2020

In this paper, we propose a system to extract effective speaker representations from a speech signal using a deep learning method. Based on the fact that speech signal contains identity unrelated information such as text content, emotion, background noise, and so on, we perform a training such that the extracted features only represent speaker-related information but do not represent speaker-unrelated information. Specifically, we propose an auto-encoder based disentanglement method that outputs both speaker-related and speaker-unrelated embeddings using effective loss functions. To further improve the reconstruction performance in the decoding process, we also introduce a discriminator popularly used in Generative Adversarial Network (GAN) structure. Since improving the decoding capability is helpful for preserving speaker information and disentanglement, it results in the improvement of speaker verification performance. Experimental results demonstrate the effectiveness of our proposed method by improving Equal Error Rate (EER) on benchmark dataset, Voxceleb1.

• Phonetics and Speech Sciences
• /
• v.7 no.3
• /
• pp.139-149
• /
• 2015

This paper aims to analyze pronunciation variations of loanwords produced by Korean and improve the performance of pronunciation modeling of loanwords in Korean by using syllable-based segmentation and phonological knowledge. The loanword text corpus used for our experiment consists of 14.5k words extracted from the frequently used words in set-top box, music, and point-of-interest (POI) domains. At first, pronunciations of loanwords in Korean are obtained by manual transcriptions, which are used as target pronunciations. The target pronunciations are compared with the standard pronunciation using confusion matrices for analysis of pronunciation variation patterns of loanwords. Based on the confusion matrices, three salient pronunciation variations of loanwords are identified such as tensification of fricative [s] and derounding of rounded vowel [ɥi] and [$w{\varepsilon}$]. In addition, a syllable-based segmentation method considering phonological knowledge is proposed for loanword pronunciation modeling. Performance of the baseline and the proposed method is measured using phone error rate (PER)/word error rate (WER) and F-score at various context spans. Experimental results show that the proposed method outperforms the baseline. We also observe that performance degrades when training and test sets come from different domains, which implies that loanword pronunciations are influenced by data domains. It is noteworthy that pronunciation modeling for loanwords is enhanced by reflecting phonological knowledge. The loanword pronunciation modeling in Korean proposed in this paper can be used for automatic speech recognition of application interface such as navigation systems and set-top boxes and for computer-assisted pronunciation training for Korean learners of English.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.4
• /
• pp.407-414
• /
• 2002

In this paper, we present a method that automatically generates concatenate morpheme based language models to improve the performance of Korean large vocabulary continuous speech recognition. The focus was brought into improvement against recognition errors of monosyllable morphemes that occupy 54% of the training text corpus and more frequently mis-recognized. Knowledge-based method using POS patterns has disadvantages such as the difficulty in making rules and producing many low frequency concatenate morphemes. Proposed method automatically selects morpheme-pairs from training text data based on measures such as frequency, mutual information, and unigram log likelihood. Experiment was performed using 7M-morpheme text corpus and 20K-morpheme lexicon. The frequency measure with constraint on the number of morphemes used for concatenation produces the best result of reducing monosyllables from 54% to 30%, bigram perplexity from 117.9 to 97.3. and MER from 21.3% to 17.6%.