Current digital methods for economic coding and transmission of auditory signals rely heavily on the inability of the auditory system to resolve certain signals. This inability, typically called masking, is exploited to achieve bit rate reductions that are truly remarkable. It is now well-established that auditory information perceptually indistinguishable from the 16 bit/sample compact disc quality can be obtained with an encoding of an average of about 3 bits/sample (e.g. Brandenburg, 1988). Such success has been achieved by allocating bits to filtered segments of the signal's spectrum in such a manner that the signal itself masks or "hides" any quantization noise that inevitably occurs. This idea has been applied to speech coding by Schroeder, Atal and Hall (1979).
展开▼
