Lossless compression of binary audio signals

A binary data format, known as direct stream digital (DSD), has been proposed for the representation of audio signals. The typical sample rate of these binary audio signals is 64 x 44.1 kHz, so the bit rate equals 64 x 44100 = 2.8 Mbit/s (per audio channel). In order to reduce this rate, without affecting the audio quality in any way, lossless compression of these signals is necessary. Because the well-known general compression methods for binary data are too complex to be practically applied to these signals, we have developed a new lossless compression method that exploits only linear dependencies in the signal, which allows a real-time hardware implementation, even for contexts (memory) of 100 bits, or more. The encoder and decoder block diagrams for our method are shown