Lossy Audio Compression Using a Statistical Sub-Band Model of Quantization Noise

The algorithm, which is based on MPEG AAC, minimizes the perceived distortion due to compression. This distortion is related to the quantization noise over frequency sub-bands. A statistical model of the quantization noise is used for computation of the optimal scalefactors for sub-bands, which is a key step in the algorithm.

(The project thumbnail is from Wikimedia Commons.)