• MALSORI
• /
• no.62
• /
• pp.149-161
• /
• 2007

In this paper we proposed a modified reliability function for improving bimodal speaker identification(BSI) performance. The convectional reliability function, used by N. Fox[1], is extended by introducing an optimization factor. We evaluated the proposed method in BSI domain. A BSI system was implemented based on GMM and it was tested using VidTIMIT database. Through speaker identification experiments we verified the usefulness of our proposed method. The experiments showed the improved performance, i.e., the reduction of error rate by 39%.

• The Journal of the Acoustical Society of Korea
• /
• v.8 no.1
• /
• pp.81-87
• /
• 1989

This paper discribes the method associated with modeling of Korean phonemes. Hidden Markov models(HMM's) may be viewed as an effective technique for modeling the inherent nonstationarity of speech signal. We propose a 3-state phoneme model to represent the sequentially changing characteristics of phonemes, i.e., transition-to-stationary-to-transition. Also we clarify that the duration of a phoneme is an important factor to have an effect in recognition accuracy and show that improvement in recognition rate can be obtained by using duration-dependent 3-state hidden Markov models.

• Journal of the Institute of Convergence Signal Processing
• /
• v.9 no.2
• /
• pp.121-128
• /
• 2008

Accurate voice activity detection have a great impact on performance of speech applications including speech recognition, speech coding, and speech communication. In this paper, we propose methods for voice activity detection that can adapt to various car noise situations during driving. Existing voice activity detection used various method such as time energy, frequency energy, zero crossing rate, and spectral entropy that have a weak point of rapid. decline performance in noisy environments. In this paper, the approach is based on existing spectral entropy for VAD that we propose voice activity detection method using MFB(Met-frequency filter banks) spectral entropy, gradient FFT(Fast Fourier Transform) spectral entropy. and gradient MFB spectral entropy. FFT multiplied by Mel-scale is MFB and Mel-scale is non linear scale when human sound perception reflects characteristic of speech. Proposed MFB spectral entropy method clearly improve the ability to discriminate between speech and non-speech for various in noisy car environments that achieves 93.21% accuracy as a result of experiments. Compared to the spectral entropy method, the proposed voice activity detection gives an average improvement in the correct detection rate of more than 3.2%.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.6
• /
• pp.506-511
• /
• 2018

This paper proposes an GMM(Gaussian Mixture Model)-DNN(Deep Neural Network) hybrid-based feature compensation method for effective speech recognition in noisy environments. In the proposed algorithm, the posterior probability for the conventional GMM-based feature compensation method is calculated using DNN. The experimental results using the Aurora 2.0 framework and database demonstrate that the proposed GMM-DNN hybrid-based feature compensation method shows more effective in Known and Unknown noisy environments compared to the GMM-based method. In particular, the experiments of the Unknown environments show 9.13 % of relative improvement in the average of WER (Word Error Rate) and considerable improvements in lower SNR (Signal to Noise Ratio) conditions such as 0 and 5 dB SNR.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.5
• /
• pp.12-20
• /
• 1999

In this paper, we compared speaker verification performance of the speech data collected in clean environment and in channel environment. For the improvement of the performance of speaker verification gathered in channel, we have studied on the efficient feature parameters in channel environment and on the preprocessing. Speech DB for experiment is consisted of Korean doublet of numbers, considering the text-prompted system. Speech features including LPCC(Linear Predictive Cepstral Coefficient), MFCC(Mel Frequency Cepstral Coefficient), PLP(Perceptually Linear Prediction), LSP(Line Spectrum Pair) are analyzed. Also, the preprocessing of filtering to remove channel noise is studied. To remove or compensate for the channel effect from the extracted features, cepstral weighting, CMS(Cepstral Mean Subtraction), RASTA(RelAtive SpecTrAl) are applied. Also by presenting the speech recognition performance on each features and the processing, we compared speech recognition performance and speaker verification performance. For the evaluation of the applied speech features and processing methods, HTK(HMM Tool Kit) 2.0 is used. Giving different threshold according to male or female speaker, we compare EER(Equal Error Rate) on the clean speech data and channel data. Our simulation results show that, removing low band and high band channel noise by applying band pass filter(150～3800Hz) in preprocessing procedure, and extracting MFCC from the filtered speech, the best speaker verification performance was achieved from the view point of EER measurement.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.5
• /
• pp.383-389
• /
• 2004

In this paper. we propose the bias compensation methods and the eigenvoice method using the mean of dimensional eigenvoice to improve the performance of rapid speaker adaptation based on eigenvoice under mismatch between training and test environment. Experimental results for vocabulary-independent word recognition task (using PBW 452 DB) show that the proposed methods yield improvements for small adaptation data. We obtained about 22∼30% relative improvement by the bias compensation methods as amount of adaptation data varied from 1 to 50, and obtained 41% relative improvement in error rate by the eigenvoice method using the mean of dimensional eigenvoice with only single adaptation word.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.115-118
• /
• 2002

Parameters used in a speaker recognition system are desirable expressing speaker's characteristics filly and have in a speech. That is to say, if inter-speaker than intra-speaker variance a big characteristic, it is useful to distinguish between speakers. Also, to make minimum error between speakers, it is required the improved recognition technology as well as the distinguishing characteristics. When we see the result of recent simulation performance, we obtain more exact performance by using dynamic characteristics and constant characteristics by a speaking habit. Therefore we suggest it to solve this problem as followings. The prosodic information is used by a characteristic vector of speech. Characteristics vector generally using in speaker recognition system is a modeling spectrum information and is working for a high performance in non-noise circumstance. However, it is found a problem that characteristic vector is distorted in noise circumstance and it makes a reduction of recognition rate. In this paper, we change pitch line divided by segment which can estimate a dynamic characteristic and it is used as a recognition characteristic. we confirmed that the dynamic characteristic is very robust in noise circumstance with a simulation. We make a decision of acceptance or rejection by comparing test pattern and recognition rate using the proposed algorithm has more improvement than using spectrum and prosodic information. Especially stational recognition rate can be obtained in noise circumstance through the simulation.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.9
• /
• pp.103-112
• /
• 1999

In this paper, a speaker-independent isolated world recognition system for a car navigation system is implemented using a general digital signal processor. This paper presents a method combining SNR normalization with RAS as a noise processing method. The semi-continuous hidden markov model is adopted and TMS320C31 is used in implementing the real-time system. Recognition word set is composed of 69 command words for a car navigation system. Experimental results showed that the recognition performance has a maximum of 93.62% in case of a combination of SNR normalization and spectral subtraction, and the performance improvement rate of the system is 3.69%, Presented noise processing method showed good speech recognition performance in 5dB SNR in car environment.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.4
• /
• pp.103-111
• /
• 2004

The method of speaker normalization has been known as the successful method for improving the accuracy of speech recognition at speaker independent speech recognition system. A frequency warping approach is widely used method based on maximum likelihood for speaker normalization. This paper propose a new power spectrum warping approach to making improvement of speaker normalization better than a frequency warping. Th power spectrum warping uses Mel-frequency cepstrum analysis(MFCC) and is a simple mechanism to performing speaker normalization by modifying the power spectrum of Mel filter bank in MFCC. Also, this paper propose the hybrid VTN combined the Power spectrum warping and a frequency warping. Experiment of this paper did a comparative analysis about the recognition performance of the SKKU PBW DB applied each speaker normalization approach on baseline system. The experiment results have shown that a frequency warping is 2.06%, the power spectrum is 3.06%, and hybrid VTN is 4.07% word error rate reduction as of word recognition performance of baseline system.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.1
• /
• pp.82-89
• /
• 2004

This paper describes how to reduce the effect of an occupation threshold by that the transform of mixture components of HMM parameters is controlled in hierarchical tree structure to prevent from over-adaptation. To reduce correlations between data elements and to remove elements with less variance, we employ PCA (principal component analysis) and ICA (independent component analysis) that would give as good a representation as possible, and decline the effect of over-adaptation. When we set lower occupation threshold and increase the number of transformation function, ordinary WLLR adaptation algorithm represents lower recognition rate than SI models, whereas the proposed MLLR adaptation algorithm represents the improvement of over 2% for the word recognition rate as compared to performance of SI models.