Research On The Adaptive Filtering Algorithm Under VoIP

With the continuous and rapid development of Internet, the speech data transfer through Internet gets into reality, and inexpensive IP phone brings challenge to the PSTN network. However, because of compression and encrypting, transfer delay becomes longer, the echo path becomes longer and the correspondence quality of speech signal obviously drops, and we have to add an echo-canceller at the terminal of VoIP. The major method of echo-cancel is that: using an adaptive filter to imitate the echo-path, so we can get the estimated echo, and then remove the echo from the proximal-end sound signal so as to cancel the echo. The process that adaptive filter tracks echo's way is adaptive filter's coefficient updating process. The Least Mean Square algorithm is a basic adaptive filtering algorithm.It is mainly based on"the Least Mean Square criterion", and makes the mean error to be smallest between the output signal and the expected output signal.This article uses as the starting point. So I worked it at the aspects below:Firstly, on the basic of expressing the key technology of VoIP, I described the concept of adaptive filtering, the basic element of adaptive filtering algorithm and the foundation of choosing it.Secondly, after introducing the fundamental principle of LMS algorithm, analyzing the function of the LMS adaptive filtering algorithm and SVSLMS algorithm. According to the fact that the convergent speed and steady-state error are affected by the fixed step-size in the classical LMS algorithm, an improved algorithm is presented with normalized variable step-size.Thirdly, I analyzed that the LMS adaptive filtering algorithm based on fractional Fourier Transform. Basing on the above work, I combined the LMS adaptive filtering algorithm with the changeable step size BLMS algorithm in this article, and made a new LMS adaptive filtering algorithm basing on the changeable step size in FRFT come true.Finally, there are simulation experiment and performance analysis to the new algorithms by using the MATLAB. The simulation results indicate that the new changeable step-size adaptive filtering algorithm have faster velocity of convergence and smaller steady state error confficient. Though the simulation experiment, it is proved that a new BLMS adaptive filtering algorithm with the changeable step size in FRFT had improvement on its astringency.