• The Journal of the Acoustical Society of Korea
• /
• v.19 no.3
• /
• pp.45-50
• /
• 2000

This paper proposes a delta energy method based on Parameter Filtering(PF), which is a speech segmentation algorithm for text dependent speaker verification system over telephone line. Our parametric filter bank adopts a variable bandwidth along with a fixed center frequency. Comparing with other methods, the proposed method turns out very robust to channel noise and background noise. Using this method, we segment an utterance into consecutive subword units, and make models using each subword nit. In terms of EER, the speaker verification system based on whole word model represents 6.1%, whereas the speaker verification system based on subword model represents 4.0%, improving about 2% in EER.

• Journal of the Korean Society for information Management
• /
• v.14 no.2
• /
• pp.213-234
• /
• 1997

A stemming algorithm for the Korean language free-text retrieval system has been designed and implemented. The algorithm contains three major parts and it operates iteratively ; firstly, stop-words are removed with a use of a stop-word list ; secondly, a basic removing procedure proceeds with a rule table 1, which contains the suffixes, the postpositional particles, and the optionally adopted symbols specifying an each stemming action ; thirdly, an extended stemming and rewriting procedures continue with a rule table 2, which are composed of th suffixes and the optionally combined symbols representing various actions depending upon the context-sensitive rules. A test was carried out to obtain an indication of how successful the algorithm was and to identify any minor changes in the algorithm for an enhanced one. As a result of it, 21.4 % compression is achieved and an error rate is 15.9%.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.20 no.5
• /
• pp.683-687
• /
• 2010

This paper studied some methods which use frequency warping method that is the one of the speech transformation method to develope the robust speech recognition system for the emotional variation. For this purpose, the effect of emotional variations on the speech signal were studied using speech database containing various emotions and it is observed that speech spectrum is affected by the emotional variation and this effect is one of the reasons that makes the performance of the speech recognition system worse. In this paper, new training method that uses frequency warping in training process is presented to reduce the effect of emotional variation and the speech recognition system based on vocal tract length normalization method is developed to be compared with proposed system. Experimental results from the isolated word recognition using HMM showed that new training method reduced the error rate of the conventional recognition system using speech signal containing various emotions.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.1
• /
• pp.47-50
• /
• 2019

This paper proposes an acoustic model adaptation method for effective speech recognition in noisy environments. In the proposed algorithm, the noise corruption function is estimated employing DNN (Deep Neural Network), and the function is applied to the model parameter estimation. The experimental results using the Aurora 2.0 framework and database demonstrate that the proposed model adaptation method shows more effective in known and unknown noisy environments compared to the conventional methods. In particular, the experiments of the unknown environments show 15.87 % of relative improvement in the average of WER (Word Error Rate).

• Phonetics and Speech Sciences
• /
• v.10 no.4
• /
• pp.111-119
• /
• 2018

This paper proposes a method to improve speech recognition performance by extracting various pronunciations of the pseudo-morpheme unit from an eojeol unit corpus and generating a new recognition unit considering pronunciation variations. In the proposed method, we first align the pronunciation of the eojeol units and the pseudo-morpheme units, and then expand the pronunciation dictionary by extracting the new pronunciations of the pseudo-morpheme units at the pronunciation of the eojeol units. Then, we propose a new recognition unit that relies on pronunciation by tagging the obtained phoneme symbols according to the pseudo-morpheme units. The proposed units and their extended pronunciations are incorporated into the lexicon and language model of the speech recognizer. Experiments for performance evaluation are performed using the Korean speech recognizer with a trigram language model obtained by a 100 million pseudo-morpheme corpus and an acoustic model trained by a multi-genre broadcast speech data of 445 hours. The proposed method is shown to reduce the word error rate relatively by 13.8% in the news-genre evaluation data and by 4.5% in the total evaluation data.

• Phonetics and Speech Sciences
• /
• v.13 no.1
• /
• pp.45-51
• /
• 2021

End-to-end (E2E) automatic speech recognition (ASR) has achieved promising performance gains with the introduced self-attention network, Transformer. However, due to training time and the number of hyperparameters, finding the optimal hyperparameter set is computationally expensive. This paper investigates the impact of hyperparameters in the Transformer network to answer two questions: which hyperparameter plays a critical role in the task performance and training speed. The Transformer network for training has two encoder and decoder networks combined with Connectionist Temporal Classification (CTC). We have trained the model with Wall Street Journal (WSJ) SI-284 and tested on devl93 and eval92. Seventeen hyperparameters were selected from the ESPnet training configuration, and varying ranges of values were used for experiments. The result shows that "num blocks" and "linear units" hyperparameters in the encoder and decoder networks reduce Word Error Rate (WER) significantly. However, performance gain is more prominent when they are altered in the encoder network. Training duration also linearly increased as "num blocks" and "linear units" hyperparameters' values grow. Based on the experimental results, we collected the optimal values from each hyperparameter and reduced the WER up to 2.9/1.9 from dev93 and eval93 respectively.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.483-489
• /
• 2020

This paper proposes acoustic model training using self-attention for low-resource speech recognition. In low-resource speech recognition, it is difficult for acoustic model to distinguish certain phones. For example, plosive /d/ and /t/, plosive /g/ and /k/ and affricate /z/ and /ch/. In acoustic model training, the self-attention generates attention weights from the deep neural network model. In this study, these weights handle the similar pronunciation error for low-resource speech recognition. When the proposed method was applied to Time Delay Neural Network-Output gate Projected Gated Recurrent Unit (TNDD-OPGRU)-based acoustic model, the proposed model showed a 5.98 % word error rate. It shows absolute improvement of 0.74 % compared with TDNN-OPGRU model.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.1
• /
• pp.97-102
• /
• 2000

Subband coding is to divide the signal frequency band into a set of uncorrelated frequency bands by filtering and then to encode each of these subbands using a bit allocation rationale matched to the signal energy in that subband. The actual coding of the subband signal can be done using waveform encoding techniques such as PCM, DPCM and vector quantizer(VQ) in order to obtain higher data compression. Most researchers have focused on the error in the quantizer, but not on the overall reconstruction error and its dependence on the filter bank. This paper provides a thorough analysis of subband codecs and further development of optimum filter bank design using vector quantizer. We compute the mean squared reconstruction error(MSE) which depends on N the number of entries in each code book, k the length of each code word, and on the filter bank coefficients. We form this MSE measure in terms of the equivalent quantization model and find the optimum FIR filter coefficients for each channel in the M-band structure for a given bit rate, given filter length, and given input signal correlation model. Specific design examples are worked out for 4-tap filter in 2-band paraunitary filter bank structure. These optimum paraunitary filter coefficients are obtained by using Monte Carlo simulation. We expect that the results of this work could be contributed to study on the optimum design of subband codecs using vector quantizer.

• The Journal of the Acoustical Society of Korea
• /
• v.14 no.5
• /
• pp.74-82
• /
• 1995

This paper is study on the voice dialing using HMM and post processing of the connected digits. HMM algorithm is widely used in the speech recognition with a good result. But, the maximum likelihood estimation of HMM(Hidden Markov Model) training in the speech recognition does not lead to values which maximize recognition rate. To solve the problem, we applied the post processing to segmental K-means procedure are in the recognition experiment. Korea connected digits are influenced by the prolongation more than English connected digits. To decrease the segmentation error in the level building algorithm some word models which can be produced by the prolongation are added. Some rules for the added models are applied to the recognition result and it is updated. The recognition system was implemented with DSP board having a TMS320C30 processor and IBM PC. The reference patterns were made by 3 male speakers in the noisy laboratory. The recognition experiment was performed for 21 sort of telephone number, 252 data. The recognition rate was $6\%$ in the speaker dependent, and $80.5\%$ in the speaker independent recognition test.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.1
• /
• pp.189-195
• /
• 2008

The OFDMA Physical layer in the WiMAX standard of IEEE 802.16e adopts 114 LDPC codes with various code rates and block sizes as a channel coding scheme to meet varying channel environments and different requirements for transmission performance. In this paper, the performances of the LDPC codes are evaluated according to various code rates and block-lengths throueh simulation studies using min-sum decoding algorithm in AWGN chamois. As the block-length increases and the code rate decreases, the BER performance improves. In the cases with code rates of 2/3 and 3/4, where two different codes ate specified for each code rate, the codes with code rates of 2/3A and 3/4B outperform those of 2/3B and 3/4A, respectively. Through the statistical analyses of the number of decoding iterations the decoding complexity and the word error rates of LDPC codes are estimated. The results can be used to trade-off between the performance and the complexity in designs of LDPC decoders.