| Adaptive filtering algorithm is more and more common in practical application.The classical adaptive filtering algorithms,such as the Least Mean Square algorithm and the Affine Projection Algorithm,have excellent performance in a variety of adaptive processes.However,when the channel is sparse impulse response,the convergence rate of these two adaptive algorithms will be seriously degraded.The proportional adaptive algorithm uses the echo path impulse response sparse structural features widely used in the field of echo cancellation.In this paper,we improve the performance of the Affine Projection Like algorithm in sparse system(acoustic echo channel system and network echo channel system)by adjusting the APL algorithm by proportional thinking,M-estimation and convex combination.First of all,in order to improve the APL algorithm for sparse system,the slow convergence rate of the problem,this paper will be proportional into the APL algorithm.Three standard APL algorithms(standard PAPL algorithm,IPAPL algorithm,MPAPL algorithm)are proposed according to different proportional control factor calculation criteria.The experimental results show that the three algorithms can be used to deal with the sparse system with respect to the convergence speed of APL algorithm.Secondly,in order to make the algorithm in the impulse noise environment performance is not affected,this paper introduces the idea of M-estimation into proportional APL algorithm,and proposes a proportional affine projection algorithm based on M-estimation(standard PAPLM algorithm,IPAPLM algorithm,MPAPLM algorithm).The experimental results show that the proposed algorithm can maintain good performance in the impulse noise environment.Finally,in order to solve the problem that the fixed-step algorithm can not balance the convergence rate and steady-state error,this paper introduces the idea of convex combination into M-proportional proportional affine projection algorithm.Through the weight transfer strategy and IPAPLM algorithm.A convex combination IPAPLM algorithm for weight vector transfer strategy is proposed.The experimental results show that the proposed algorithm can achieve faster convergence speed and lower steady-state error. |
