Reduced-bit-rate source coding method for digital audio signal transmission - applying neural network or fuzzy logic in all or parts of encoding and decoding procedures

摘要

The audio signal (10) to be compressed is source-coded by means of a simulated neural network or fuzzy logic (4), with optional use of a separate signal analysis (3). Additional preprocessing and postprocessing steps (6,7) may be applied, the former involving e.g. signal transformation in the frequency domain. The postprocessing may involve standardisation of the signal or matching of the source-coded transmission (20) to the requirements of the channel. Psycho-acoustic frequency and level parameters of tonal and noisy maskings may be detected in layers of the neural network. ADVANTAGE - Efficiency of coding is increased with redn. of irrelevancy by exploitation of properties of human hearing.