Lossless compression for μ-law (A-law) and IMA ADPCM on the basis of a fast RLS algorithm

Lossless compression methods are introduced for μ-law (A-law) and IMA ADPCM standards. Lossless compression means that for a given original input we generate the exactly same output as that of these standards while reducing the bit rates of the compressed files. To keep the same quality, we use the same quantization methods as that in the standards. To reduce the bit rates, we use prediction and entropy coding techniques. The prediction is based on a fast recursive least square (RLS) algorithm which requires less computation than existing RLS algorithms. The entropy coding is based on Huffman coding. The prediction, quantization and coding steps are integrated into an adaptive scheme. Then we can keep the same quality while reducing the bit rates per sample. Comparing to 8 bit per sample μ-law or A-law, only 3.24 bits per sample are needed for coding of an audio signal with 44100 Hz sampling frequency and 4.72 bits for speech/audio signals at 11025 Hz. Some improvements for IMA ADPCM standard are also obtained