Gaussian Source Coding using a Simple Switched Quantization Algorithm and Variable Length Codewords

Abstract

This paper introduces an algorithm based on switched scalar quantization utilizing a novel Îź-law quantization model (optimized in terms of minimal distortion) and variable length codewords, for high-quality encoding of the signals modeled by Gaussian distribution. The implemented Îź-law quantizer represents an improvement of the standard Îź-law quantizer in terms of bit rate, at the same time providing the equal signal quality. The main concept of the algorithm is to divide the range of the input signal variances into a certain number of sub-ranges, and to design the optimal quantizer for each sub-range. The signal is processed frame-by-frame, and for each frame the best performing quantizer is chosen, where the estimated frame variance is used as the switching criterion. Theoretical results indicate that the proposed algorithm achieves performance comparable to the standard Îź-law quantizer, enabling the compression of about 0.5 bit/sample. The simulation results are provided to confirm the correctness of the proposed model.

Keywords

• gaussian distribution
• quantization
• source coding
• signal processing algorithms
• signal to noise ratio