Dynamic range compression of sound mixtures

Abstract

Nearly all advanced hearing aids use dynamic range compression (DRC) to make quiet sounds louder and loud sounds quieter. A form of nonlinear gain control, DRC can help to improve listening comfort, but it can also introduce distortion in the presence of multiple competing sounds. The gain applied to all sounds in a mixture is determined by the level of the loudest sound; the signals therefore modulate each other. It has been widely observed that DRC performs poorly in noisy environments, but there has been little mathematical analysis of the interaction between compression and noise. In this work, we use an idealized model, equating the envelope of a signal with its statistical variance, to analyze the effects of DRC on mixtures of uncorrelated signals. We show that when DRC is applied to a mixture, the effective compression applied to each sound is weaker than it would have been in isolation. Similarly, we analyze how DRC algorithms alter the long-term signal-to-noise ratio of a sound mixture. This analysis can help us to develop DRC algorithms that are more robust to noise and improve the performance of hearing aids in the complex environments where they are needed most.