ESSV Konferenz Elektronische Sprachsignalverarbeitung

Title: Extending Monaural Speech and Audio Codecs by Inter-Channel Linear Prediction

Authors: Magnus Schäfer, Hauke Krüger, Peter Vary

Abstract:

In this contribution, we propose the application of a novel concept for a flexible hierarchical stereo extension of existing monaural speech and audio codecs. The concept is based on inter-channel linear prediction of the left and right channels from a sum signal and allows for a very flexible extension of existing speech and audio codecs. In contrast to theoretical examinations in earlier publications, a stereo codec is built in this contribution by combining the new stereo framing with the core transmission of the standardized Adaptive Multi Rate - WideBand codec (AMRWB) in a hierarchical manner. The proposed modification introduces just marginal additional system delay compared to mono AMR-WB. It will be shown by simulations that compared to an individual transmission of left and right channel, the application of the inter-channel linear prediction concept achieves an identical quality at a significantly lower data rate for an important class of stereo signals. This is due to a concentration of most of the signal energy in the sum signal and the filter coefficients while the prediction error is of lesser importance for the quality. This will be shown by gradually decreasing the transmission data rate for the prediction error to the point of a purely parametric solution where only the sum signal and the filter coefficients are transmitted.


Year: 2009
In session: Signalverarbeitung
Pages: 166 to 173