A computer-implemented method for performing digital signal processing, comprising:
- obtaining a first set of digitized coefficients from source data;
- determining a best-fit distribution of a generalized Gaussian distribution for the set of digitized coefficients;
- applying a quantization algorithm to the first set of plurality of digitized coefficients to obtain a second set of quantizers, wherein the quantization algorithm is based at least in part on the determined best-fit distribution;
- and
- providing second set of quantizers as a compressed representation of the source data.