Rouhollah bibak - Department of Electrical and Computer Eng., Tarbiat Modares University, Tehran, Iran
Abumoslem jannesari - Department of Electrical and Computer Eng., Tarbiat Modares University, Tehran, Iran
Mojtaba Lotfizad - Department of Electrical and Computer Eng., Tarbiat Modares University, Tehran, Iran

In this paper, an amendment to the APA algorithm is proposed. This modification is based on clipping the input signal to three levels which is then used in updating the filter coefficients. The input signal is quantized to three levels 0, 1 and -1. In addition to a reduction in the computational complexity, this scheme increases the convergence rate for a particular range of the threshold. This algorithm showed a performance similar to that of the conventional APA algorithm against the impulsive noise, and also for the case with colored input signal in the system identification configuration. Another contribution of this paper is to further reduce the computational complexity by applying the clipped input signal to the unknown system as well as to both the filtering and the weight update sections of the adaptive filter. In this case, although the convergence rate somewhat decreases, it leads to more reduction in the computational complexity.

adaptive filters; APA algorithm; CAPA algorithm; computational complexity; convergence rate; impulsive noise; colored input