A New Technique of Modeling Acoustic Echo and its Implementation Using FPGA

Abstract

We propose an integrated acoustic echo cancellation solution based on using multiple of small adaptive filters rather than using one long adaptive filter. A new approach is proposed using the concept of decomposing the long adaptive filter into low order multiple sub-filters in which the error signals are independent on each other. The independency of the error signals exhibits the parallelism technique. A novel class of efficient and robust adaptive algorithms. It exhibits fast convergence, superior tracking capabilities of the signal statistics. The proposed algorithm is also compared with multiple sub-filters approach used for acoustic echo cancellation as the technique of decomposition of error. It is generally found that adaptive LMS algorithm with lower order has faster convergence. In most of the cases, the eigen-value spread of the auto correlation decreases as the order of the filter decreases except for white input. We discuss also the complexity of our proposed design to show how our design has a good performance. Modern (FPGAs) include the resources needed to design efficient filtering structures. The modeling of the acoustic echo path was represented by using three sub-adaptive filters of order=10 with fixed step size =0.05/3 for each adaptive filter. We use sinusoidal input signal with additive white Gaussian noise (AWGN) which has different signal-to-noise ratio (SNRs) to examine our approach. The steady state error of our proposed technique is still high as the technique of decomposition of error. This steady state error is small with respect to using one long adaptive filter and this will be obvious in our simulation results. This paper addresses also the problems of blind source separation (BSS). In blind source separation, signals from multiple sources arrive simultaneously at a sensor array, so that each sensor output contains a mixture of source signals. Sets of sensor outputs are processed to recover the source signals from the mixed observations. The term blind refers to the fact that specific source signal values and accurate parameter values of a mixing model are not known a priori. Application domains for the material in this paper include communications, biomedical, and sensor array signal processing. Simulations are often needed when the performance of new methods is evaluated. If the method is designed to be blind or robust, simulation studies must cover the whole range of potential random input. It follows that there is a need for advanced tools of data generation. The purpose of this thesis is to introduce a technique for the generation of correlated multivariate random data with non-Gaussian marginal distributions for blind source separation technique.The output random variables are obtained as linear combinations of independent components. The covariance matrix and the first five moments of the output variables may be freely chosen. Moreover, the output variables may be filtered in order to add autocorrelation. The overall system is composed of echo with different sources of sounds, first we separate all these sources using blind source separation then we use our proposed multiple sub adaptive filters approach for cancellation the unwanted sources.