Guangyu Zhou

University of Central Florida, Orlando, Florida, United States

Are you Guangyu Zhou?

Claim your profile

Publications (4)0.33 Total impact

  • Guangyu Zhou, W.B. Mikhael
    [Show abstract] [Hide abstract]
    ABSTRACT: In this work, speaker identification (SI) approach which is based on vector quantization (VQ) is presented. The method employs adaptive techniques to select the optimal parameters of the discriminative function. The proposed adaptive discriminative VQ based SI (ADVQSI) technique considers the interspeaker variation between each speaker and all speakers in the SI group in order to enlarge the speakers' template differences. For each speaker, the speech feature vector space is divided into subspaces. Different discriminative weights are given to different subspaces. Subspaces with larger discriminative weights play a more important role in the SI decision. The performance of ADVQSI is analyzed and tested experimentally. The experimental results confirm the performance improvement employing the proposed technique in comparison with the existing VQ technique for SI (VQSI) and recently reported discriminative VQ techniques for SI (DVQSI).
    Circuits and Systems, 2005. 48th Midwest Symposium on; 09/2005
  • Guangyu Zhou, Wasfy B. Mikhael, Brent Myers
    [Show abstract] [Hide abstract]
    ABSTRACT: A novel Discriminative Vector Quantization method for Speaker Identification (DVQSI) is proposed, and its parameters selection is discussed. In the training mode of this approach, the vector space of speech features is divided into a number of regions. Then, a Vector Quantization (VQ) codebook for each speaker in each region is constructed. For every possible combination of speaker pairs, a discriminative weight is assigned for each region, based on the region's ability to discriminate between the speaker pair. Consequently, the region, which contains a larger distribution difference between the speech feature vector sets of the two speakers in the speaker pair, plays a more important role by assigning it a larger discriminative weight, in identifying the better speaker match from the two speakers. In the testing mode, to identify an unknown speaker, discriminative weighted average VQ distortion pairs are computed for the unknown speaker input waveform. Then, a technique is described that figures out the best match between the unknown waveform and speakers' templates. The proposed DVQSI approach can be considered a generalization of the existing VQ technique for Speaker Identification (VQSI). The method presented here yields better Speaker Identification (SI) accuracy by employing the discriminative weights and space segmentation as design parameters. This is confirmed experimentally. In addition, a computationally efficient implementation of the DVQSI technique is given which uses a tree-structured-like approach to obtain the codebooks.
    Journal of Circuits System and Computers 06/2005; 14(3):581-596. DOI:10.1142/S0218126605002404 · 0.33 Impact Factor
  • Guangyu Zhou, W.B. Mikhael
    [Show abstract] [Hide abstract]
    ABSTRACT: Different approaches have been proposed for speaker identification (SI). Distortion outputs of template-based SI are generally in compatible with probability measures. Frequently, data fusion is used for SI that uses the two kinds of distortion measures, which give rise to incompatibility problems. A technique, which converts the distortion outputs of template-based SI classifiers into compatible probability measures at the same scale for the SI data fusion problem at the measurement level, is presented. In the proposed approach, for each template-based classifier, the stochastic model for each distortion output of the classifier and each speaker, given that the unknown utterance comes from this speaker, is estimated. Then, a posteriori probability of the unknown utterance belonging to each speaker is calculated for each given distortion output. Compatible probability measures of the distortion outputs are obtained based on the posteriori probabilities. Experimental results confirm the effectiveness of the proposed approach for SI data fusion at the measurement level.
    Circuits and Systems, 2004. ISCAS '04. Proceedings of the 2004 International Symposium on; 06/2004
  • Source
    Guangyu Zhou, W.B. Mikhael
    [Show abstract] [Hide abstract]
    ABSTRACT: A novel discriminative vector quantization method for speaker identification (DVQSI) is proposed, and its parameters selection is discussed. The vector space of speech features is divided into a number of subspaces and the distribution of the inter speaker variation inside the speech feature vector space is considered. Discriminative weighted average distortion instead of equally weighted average distortion is used in speaker identification (SI). The proposed approach can be considered a generalization of the existing vector quantization (VQ) technique and the experimental results confirm the improved SI accuracy
    Circuits and Systems, 2003 IEEE 46th Midwest Symposium on; 01/2004

Publication Stats

5 Citations
0.33 Total Impact Points

Institutions

  • 2004–2005
    • University of Central Florida
      • Department of Electrical Engineering & Computer Science
      Orlando, Florida, United States