• MALSORI
• /
• no.54
• /
• pp.73-83
• /
• 2005

The goal of our research is to build a text-independent speaker identification system that can be used in any condition without any additional adaptation process. The performance of speaker recognition systems can be severely degraded in some unknown mismatched microphone and noise conditions. In this paper, we show that PCA(principal component analysis) can improve the performance in the situation. We also propose an augmented PCA process, which augments class discriminative information to the original feature vectors before PCA transformation and selects the best direction for each pair of highly confusable speakers. The proposed method reduced the relative recognition error by 21%.

• Speech Sciences
• /
• v.9 no.3
• /
• pp.3-16
• /
• 2002

We examined some problems in speaker identification procedures: transformation of acoustic parameters into auditory scales, invalid measurement values, and comparability of spectral energy values across the frequency range. To resolve those problems, we analyzed the acoustic spectral energy of three Korean numbers produced by ten female students from narrow-band spectrograms at 19 proportional time points of each voiced segment. Then, cells of the first five spectral matrices were averaged to form a matrix model for each speaker. The correlation coefficients and sum of the absolute amplitude difference in each pair of the spectral models of the ten subjects were obtained. Also, some individual matrix models were compared to those of the same subject or the other subject with a similar spectral model. Results showed that in numbers '2' and '9' subjects could not be clearly distinguished from the others but in number '4' it shed some possibility of setting threshold values for speaker identification if we employed the coefficients and the sum of absolute difference. Further studies would be desirable on various combinations of the range of long-term average spectra and the degree of signal pre-emphasis.

• Speech Sciences
• /
• v.10 no.2
• /
• pp.249-258
• /
• 2003

This study attempted to examine a speaker identification method using difference sum and correlation coefficient determined from a pair of intensity level matrices of band-pass-filtered numeric sounds produced by ten female speakers of similar age and height. Subjects recorded three digit numbers at a quiet room at a sampling rate of 22 kHz on a personal computer. Collected data were band-pass-filtered at five different band ranges. Then, matrices of five intensity levels at 100 proportional time points were obtained. Pearson correlation coefficients and the sum of absolute intensity differences between a pair of given matrices were determined within and across the speakers. Results showed that very high correlation coefficient and small difference sum generally occurred within each speaker but some individual variation was also observed. Thus, the matrix pair with a higher coefficient and a smaller difference sum was averaged to form each individual's model. Comparison among the speakers yielded generally low coefficients and large differences, which suggests successful speaker identification, but among them there were a few cases with very high coefficients and small differences. Future studies will focus on finer band ranges and additional spectral parameters at some peak points of the intensity contour at a low frequency band.

• Journal of Internet Computing and Services
• /
• v.5 no.3
• /
• pp.57-62
• /
• 2004

In this paper, an efficient speaker identification based on robust vector quantizationprincipal component analysis (VQ-PCA) is proposed to solve the problems from outliers and high dimensionality of training feature vectors in speaker identification, Firstly, the proposed method partitions the data space into several disjoint regions by roust VQ based on M-estimation. Secondly, the robust PCA is obtained from the covariance matrix in each region. Finally, our method obtains the Gaussian Mixture model (GMM) for speaker from the transformed feature vectors with reduced dimension by the robust PCA in each region, Compared to the conventional GMM with diagonal covariance matrix, under the same performance, the proposed method gives faster results with less storage and, moreover, shows robust performance to outliers.

• Journal of KIISE:Software and Applications
• /
• v.27 no.5
• /
• pp.583-592
• /
• 2000

This paper proposes feature parameter transformation method using independent component analysis (ICA) for speaker identification. The proposed method assumes that the cepstral vectors from various channel-conditioned speech are constructed by a linear combination of some characteristic functions with random channel noise added, and transforms them into new vectors using ICA. The resultant vector space can give emphasis to the repetitive speaker information and suppress the random channel distortions. Experimental results show that the transformation method is effective for the improvement of speaker identification system.

• Speech Sciences
• /
• v.14 no.1
• /
• pp.163-174
• /
• 2007

All observation has uncertainty due to noise or channel characteristics. This uncertainty should be counted in the modeling of observation. In this paper we propose a modified optimization object function of a GMM training considering inexact observation. The object function is modified by introducing the concept of observation confidence as a weighting factor of probabilities. The optimization of the proposed criterion is solved using a common EM algorithm. To verify the proposed method we apply it to the speaker recognition domain. The experimental results of text-independent speaker identification with VidTimit DB show that the error rate is reduced from 14.8% to 11.7% by the modified GMM training.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.1
• /
• pp.35-41
• /
• 1992

A speaker identification scheme using the speaker-based VQ codecook of a sustained vowel is proposed and tested. With the pitch synchronous LPC vector of the sustained vowel /i/ as a feature vector, a VQ codebook size of 4 was found to be suitable to characterize each speaker's feature space. For 40 normal speakers (20 males, 20 females), we achieved the correct identification rate of 99.4% with a training data set, and 89.4% with a test data set with speech samples of only 50 pitch periods.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.45-48
• /
• 2007

This paper proposes a flexible selection method of feature vectors for speaker identification. In speaker identification, overlapped region between speaker models lowers the accuracy. Recently, a method was proposed which discards overlapped feature vectors without regard to the source causing the overlap. We suggest a new method using both overlapped features among speakers and non-overlapped features to mitigate the overlap effects.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.3
• /
• pp.89-97
• /
• 2021

Voice interface can not only make daily life more convenient through artificial intelligence speakers but also improve the working environment of the factory. This paper presents a voice-assisted work management system that supports both speech and speaker recognition. This system is able to provide machine control and authorized worker authentication by voice at the same time. We applied two speech recognition methods, Google's Speech application programming interface (API) service, and DeepSpeech speech-to-text engine. For worker identification, the SincNet architecture for speaker recognition was adopted. We implemented a prototype of the work management system that provides voice control with 26 commands and identifies 100 workers by voice. Worker identification using our model was almost perfect, and the command recognition accuracy was 97.0% in Google API after post- processing and 92.0% in our DeepSpeech model.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.3
• /
• pp.348-354
• /
• 2003

This proposes a new algorithm for advanced personal identification system using iris pattern and speech signal. Since the proposed algorithm adopts a fusion scheme to take advantage of iris recognition and speaker identification, it shows robustness for noisy environments. For evaluating the performance of the proposed scheme, we compare it with the iris pattern recognition and speaker identification respectively. In the experiments, the proposed method showed more 56.7% improvements than the iris recognition method and more 10% improvements than the speaker identification method for high quality security level. Also, in noisy environments, the proposed method showed more 30% improvements than the iris recognition method and more 60% improvements than the speaker identification method for high quality security level.