Sub-band feedback cancellation with variable step sizes for music signals in hearing aids

Standard adaptive feedback cancellation algorithms in hearing aids suffer from a biased adaptation if the input signal is spectrally colored, as it is for tonal signals, like music. Due to that, distortion artifacts (entrainment) are generated. In this paper, a sub-band feedback cancellation system is presented combined with an adaptation control to deal with those signals. Two control concepts for determining the variable step sizes [1, 2], known from general adaptive filter algorithms, are theoretically and practically analyzed and evaluated for an application to feedback cancellation. For feedback cancellation the control is combined with known methods to reduce the bias, such as prediction error filtering or frequency shifting. Based on this combination, a completely new setup for feedback cancellation is proposed. It relies entirely on signals accessible in real systems, shows a low computational complexity, and therefore has a strong practical relevance.