ABSTRACT

Audio compression algorithms are used to obtain compact digital representations of high-fidelity audio signals for the purpose of efficient transmission over larger distances and storage. This paper presents an audio coder for real-time mutimedia applications. This coder applies a discrete wavelet transform to decompose audio test files into subbands to eliminate redundant data using spectral and temporal masking properties. This architecture is combined with a psychoacoustic model characterised by an external and middle ear model as a first part and a dynamic Gammachirp filter as a second part whose connections are selected in order to come close to the critical bands of the ear. Experimental results show the best performance of this architecture.