Improving joint stereo audio coding by adaptive inter-channel prediction

Abstract

A method for exploiting inter-channel redundancies of stereophonic or multichannel audio signals is presented. In contrast to known stereo redundancy reduction techniques used in joint stereo audio coding. Where only the statistical dependencies between two concurrent samples of the left and right channel signals are considered, the adaptive inter-channel prediction also takes into account possible phase or time delay between the channels and exploits more than only one value of the cross-correlation function. The analysis of subjective listening test results has shown that this technique is especially effective for a class of test sequences which has proven to be most critical for the ISO MPEG Layer II and Layer III codecs at bit rates of 2/spl times/64 kbit/s. For these signals the gain due to the stereo redundancy reduction technique used in Layer III joint stereo coding is less than 5-10 dB, while in Layer II joint stereo coding no specific stereo redundancy reduction technique is used. In a first step, the adaptive inter-channel prediction has been applied to an ISO MPEG Layer II codec. The simulation results show that a prediction gain up to 30-40 dB can be achieved for large parts of the above mentioned signals.<>