This paper presents a transient analysis of a hearing aids adaptive feedback canceller. The system employs an LMS adaptive estimator and an LMS adaptive predictor operating simultaneously. The nature of the practical problem makes the input to the adaptive estimator and the interference to its output statistically correlated. First, a modification is proposed to the original structure that accelerates convergence without compromising the cancellation level. The modified structure is then analyzed for slow adaptation and for an autoregressive input signal. Analytical models are derived for the mean behavior of the adaptive weights. Monte Carlo simulations verify the accuracy of the derived model.
展开▼
