ABSTRACT:
This paper describes the suboptimal two-stage, 11 bit vector quantizer (VQ) for coding Line Spectral Frequency (LSF) parameters for speech communication. The proposed VQ possesses low search complexity as compared with single stage vector quantization, using exhaustive search technique while maintaining nearly seme quelity. This proposed two-stage VQ techniqua quantizes LSF parameters at 11 bits/frame with a frama size of 10ms. Intra frame ordaring propertias and features that LSF are bounded within a given range are exploited in codebook design. A pre-selection codebook designed from the second stage codebooks is used as first stage codebook in order to reduce the search complexity. Moreover, this reduction in search complexity reducas the search deley as well as the computational complexity of the search.. Codebook search is based on the maximum matching of input LSF vector and the codebook vector using minimum mean square error (MMSE) as the distanca criterion. This paper describas the two staga codebook dasign, and codebook search techniques. The performance of the proposed VQ has been tested and results are reported in this paper.
Cite this article:
Khan, Shukla and Khetan (2007). A Sub-Optimal, Low Search Complexity, Vector Quantization (11-Bit) of LSF Parameters of Speech. Journal of Ravishankar University (Part-B: Science), 20(1), pp.13-18.