Research On Adaptive Echo Cancellation Based On Affine Projection Algorithm

With the rapid development of communication and network technology,people have higher requirements for the quality of communication,especially for voice communication quality.A recent investigation has found that the presence of echo is one of the main factors that affect the quality of voice communication,that are generated acoustically by the coupling between the loudspeaker and the microphone via the room impulse response.The echo which are produced during the call are harmful to the speech intelligibility,and it has a great influence on the normal communication.At present,the echo canceller based on adaptive filtering technology is a widely used approach to solve this problem.The basic principle of echo canceller is to use the adaptive algorithm to estimate the echo path impulse response,and then use the estimated impulse response to filter the echo from the desired signal.The least mean square error algorithm is the one of the earliest application of adaptive filter.It has the advantages of simple calculation and easy to apply.However,the least mean square algorithm has poor convergence performance for the colored signal,especially the speech signal.To solve this problem,Ozeki extends the least mean square error algorithm,and proposes the affine projection algorithm which updates the weight of the adaptive filter by multiple regression factors.Because the affine projection algorithm has the characteristics of decorrelation and minimum perturbation,it can maintain good convergence for colored signal,and has faster convergence speed and lower computational complexity.However,similar to other adaptive filtering algorithms,affine projection algorithm needs to determine the design parameters in advance for the compromise between convergence rate and steady-state error.In this paper,to solve the contradiction between convergence speed and steady-state error and improve the overall convergence ability of adaptive filter,three kinds of variable parameter affine projection algorithm were put forward.These three algorithms adjust the regularization factor,step-size and convex combined filter parameter respectively.The main theoretical basis of several methods proposed in this paper is the principle of maximum mean square error descent,that is,by adaptively adjusting parameters within that convergence process,it is expected to be obtained the minimum mean square error.Compared with the traditional affine projection algorithm,the proposed method not only effectively improves the convergence performance of the adaptive filter,but also has stronger robustness and has higher values on application.Echo cancellation simulation results show that the proposed several variable parameter affine projection methods in this paper,has better performance in terms of convergence rate and steady state error,and improve the overall performance of echo canceller effectively.