Vector Quantization of Principal Spectral Components for Speech Coding at 4800 BPS
P.C. Loizou, Andreas Spanias · 2005
In this paper, we present some results from a study on the vector quantization of spectral Components. Three types of transform-based speech representation are examined, namely a least-squares representation, a subband representation, and a pitch-harmonic representation. The spectral components associated with these representations are encoded using vector quantization techniques. The results of all the coding schemes are evaluated using signal-to-noise ratios and informal listening tests.