An improved block sparse method for feedback suppression in hearing aids

The conventional algorithms such as least mean square (LMS) and normalized LMS (NLMS) exhibit degraded performance for colored signal as input in the presence of impulsive noise. To circumvent this, the family of affine projection sign algorithm (APSA) has been introduced for improved performance under impulsive noise interference. Further, to obtain significant reduction in the computations while maintaining the accuracy, considering the memory of previous gain factors, block sparse memory improved proportionate APSA (BS-MIP-APSA) algorithm has been suggested. This paper introduces an improved BS-MIP-APSA (IB-MIP-APSA) algorithm based feedback canceler for hearing aids, incorporating the benefits of variable step-size and BS-MIP-APSA for faster convergence rate, enhanced steady state performance and less computations in the situation of impulsive noise intervention. Simulation studies confirm the performance of the proposed feedback canceler in terms of misalignment, ASG for input signals such as white noise and speech segment.