Efficient coding of high‐quality sound by an adaptive prediction scheme

A new digital coding system at 64 kb/s, which can transmit wide‐band speech and music signals in digital telephone network, is presented. Adaptive prediction and adaptive bit allocation constitute this system. the input signal is divided into several frequency bands and predictive coding is applied to each band. the dynamic bit allocation, which uses redundancy based on the frequency spectrum concentration, waveform timewise concentration, and periodicity in the case of coding prediction residual signals along with the interchannel power deviation in the case of the stereo signals, is applied and the information compression is achieved. the system parameters of this coding scheme are chosen to maximize the signal to quantization noise ratio. the transmission band was chosen as 8 kHz wide based on subjective tests. the performance of coding schemes is evaluated by the psychological equivalence relation to the log‐companding PCM and by the psychological scale using the category judgment method. As a result, it is clarified that information compression more than twice that of PCM is achieved for the speech and music sound, and the quality close to the original sound can be achieved at 64 kb/s for the monophonic signal and at 128 kb/s for the stereo.