10th Speech in Noise Workshop, 11-12 January 2018, Glasgow

The effects of SNR driven amplitude compression in hearing aids on output SNR and signal envelope distortion

Christophe Lesimple(a), Miquel Sans
Bernafon AG, Bern, Switzerland

(a) Presenting

Compressing the amplitude of the incoming signal is a standard application of hearing aid amplification. The amount of compression is initially defined by the selected fitting rationale that indicates how much gain should be applied to a speech signal at different input levels. The results from speech intelligibility tests, in controlled environments, indicate that compression can compensate for the loss of audibility for soft speech (Marriage & Moore, 2003; Davies-Venn et al., 2009).
Difficulties appear when noise degrades the speech signal. In this case, the compressor still applies gain based on the input level without distinguishing the signal type. Over-amplification of noise by compressive amplification was found to result in degradation of the output SNR (Naylor & Johannesson, 2009) and speech envelope flattening (Jenstad & Souza, 2007).
A decision block applying corrections to the amplification, based on the SNR estimation, at a phonemic resolution, was designed to address these two negative effects. The principle of SNR driven compression is to estimate if the signal is useful or not, i.e. for noise the effective compression is released to avoid over-amplification. This qualification is not restricted by pre-defined rules for listening environment detection so that it can measure small and fast changes in daily situations.
The effect of different hearing aid signal processing algorithms was measured individually and in different combinations. The output SNR and envelope distortion index were calculated from different spatial distributions of speech and noise sources. When noise is spatially separated from the signal, traditional algorithms, such as directionality, improve the output SNR and the envelope distortion index. However, directional microphone algorithms cannot avoid an output SNR degradation caused by compression at positive input SNRs. SNR driven compression partially compensates for this loss and improves the contrast of the envelope at the output of the hearing aid. The resulting changes in output SNR and the envelope distortion index with different outcome measures will be presented and discussed.

Last modified 2017-11-17 15:56:08