ecvqencode encodes trainingfile (DAT format) using entropy-constrained vector quantization (ECVQ). ecvqencode uses input codebookfile (CBK format) as the ECVQ codebook and outputs a channel of indices to channelfile (CHN format). ecvqencode implements full-search VQ with a rate-distortion-based cost measure in the form J = D + lambda * R where D is the squared-error distortion, R is a rate estimate (based on codeword probabilities), and lambda is a "rate-distortion" parameter that controls the trade-off between rate and distortion in the cost function. For more details on ECVQ, refer to the paper by Chou, Lookabaugh, and Gray cited below.
The ECVQ codebook, codebookfile, can be trained using ecvqtrain(1) . The channel symbols range from 0 to N-1, where N is the number of codewords in codebookfile. The output channel, channelfile, can be decoded (to produce a quantized datafile) using vqdecode(1) .
P. A. Chou, T. Lookabaugh, and R. M. Gray, "Entropy-constrained Vector Quantization," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 37, pp. 31-42, January 1989.