Research On Feedback Cancellation Algorithm Of Digital Hearing Aids

Hearing is one of the most important part when human communicate with surroundings.In Recent years, due to deterioration of environment, aging of population and some other reasons,the number of people with hearing loss is increasing sharply and it has seriously hampered the development of our society. At this stage, Hearing aid is the most effective way to improve patients’aural comprehension, however feedback problems seriously affect its practical effects. Based on research of existing algorithms, this paper propose a Variable Step Size Improved Proportionate Normalized Least Mean Square algorithm and an improved algorithm for howling detection and suppression.The main work of this paper can be shown as follows:1. This paper make deeply research on the principle of adaptive filtering and two widely used adaptive filtering algorithms-LMS and RLS, then campare the advantages and disadvantages of LMS and RLS algorithm through simulations and complexity anylsis.2. A new feedback cancellation algorithm-Variable Step Size Improved Proportionate Normalized Least Mean Square algorithm(VSS-IPNLMS) based on state classification is proposed in this paper.This paper applies IPNLMS algorithm to feedback cancellation algorithm in digital hearing aids and use coefficients’energy of the adaptive filter to divide state of adaptive filter into three, namely convergence state, trans it ion state and steady state. When adaptive filter is in convergence state or steady state, IPNLMS algorithm is used to update its coefficients and while it’s in transition state,the proposed VSS-IPNLMS algorithm is used to update adaptive filter’s coefficients.The proposed algorithm use the normalized distance between coeffocients’long term average and short term average for global step control.Simulations and experiments demonstrated that proposed VSS-IPNLMS algorithm can achieve faster and more accurate convergence than IPNLMS and NLMS algorithm.3. An improved frequency domain howling detection and suppression algorithm based on research of howling generation mechanism and existing frequency domian and time domain howling detection algorithm is proposed in this paper.The proposed algorithm combined improved peak to neighboring power ratio(IPNPR) and interframe peak magnitude persistence(IPMP) for howling detection and then use frequency calculation algorithm for accurate frequency calculation.Once howling is detected,notch filter is used for howling suppression.Simulations and experiments demonstrated that the proposed algorithm has better accuracy and robustness.4. Applied feedback cancellation algorithms and other digital hearing aids algorithms to embedded system and use a software running on PC to control the whole system.Experiments based on the embedded system verifies the validity of algorithms proposed in this paper.