Intelligibility Enhancement at the Receiving End of the Speech Transmission System - Effects of Far-End Noise Reduction

Post-processing methods can be used in mobile communications to improve the intelligibility of speech in adverse near-end background noise conditions. Generally, it is assumed that the input of the post-processing contains quantization noise only, that is to say, no far-end noise is present. However, this assumption is not entirely realistic. Therefore, the effect of farend noise with and without noise reduction on the performance of three post-processing methods is studied in this investigation. The performance evaluation is done using subjective intelligibility and quality tests in several far-end and near-end noise conditions. The results suggest that although the noise reduction generally improves performance in stationary far-end noise, the noise reduction does not improve intelligibility in unstationary far-end noise conditions but has a positive impact on perceptual quality for some of the post-processing methods.