This paper introduces an algorithm for speaker identification based on multi-codebook vector quantization (MCVQ). MCVQ combines different size codebooks to achieve high recognition accuracy for text-independent speaker identification and reduce the number of distortion calculations during matching between test frame and speakers’ codebooks. Experimental work has shown that the proposed model sp...

Abstract In this paper we consider the motion segmentation problem on sparse and unstructured datasets involving rigid motions, motivated by multibody structure from motion. particular, assume only two-frame correspondences as input without prior knowledge about trajectories. Inspired success of synchronization methods, address introducing a two-stage approach: first, is addressed image pairs i...

In this paper, a within-class feature normalization (WCFN) framework operating in transformed segment-level (instead of frame-level) super-vector space is proposed for robust speech recognition. In this framework, each segment hypothesis in a lattice is represented by a high dimensional super-vector and projected to a class-dependent lower-dimensional eigensubspace to remove unwanted variabilit...

Most low bit rate speech coders employ linear predictive coding (LPC) which models the short-term spectral information within each speech frame as an all-pole lter. In this thesis, we examine various methods that can e ciently encode spectral parameters for every 20 ms frame interval. Line spectral frequencies (LSF) are found to be the most e ective parametric representation for spectral coding...

A 5.3-kbp fragment of the KHS gene was cloned from a genomic bank of Saccharomyces cerevisiae No. 115 constructed with an E. coli as the host and YEp13 as the vector. A non-killer yeast strain was transformed to a killer strain with the multi-copy vector containing the KHS gene, and the transformant could secrete 3-4 times more killer toxin into culture media than the donor, strain No. 115. The...