Transient noise reduction for hearing aid

This paper proposes speech enhancement to reduce transient noise. The transient noise is composed of a wideband component which is attack of the noise, and a narrowband component due to a natural oscillation of an object. The duration of a narrowband component is long. Unfortunately, it affects a conversation more than the wideband component. Therefore, we proposed the transient noise reduction for speech. It uses a backward linear predictor using the adaptive algorithm with 4th order cumulant. However, there is a problem that the convergence rate is slow. We propose a RLFC (Recursive least 4th order cumulant) algorithm that convergence rate is faster than the previous one.