Novel sign subband adaptive filter algorithms with individual weighting factors

Abstract

a b s t r a c tThis paper presents a new sign subband adaptive filter (SSAF) algorithm with an individual-weighting-factor (IWF) for each subband, instead of a common weighting factor in the original SSAF algorithm, called the IWF-SSAF. Each individual weighting factor only depends on the corresponding subband input signal power. Compared with the SSAF algorithm, the proposed approach fully utilizes the inherent decorrelating property of subband adaptive filter for colored inputs, leading to a better convergence performance. After that, to further enhance the performance of the IWF-SSAF in a sparse system, an improved proportionate IWF-SSAF (IWF-IPSSAF) algorithm is proposed. The proposed algorithms not only inherit the good robustness of sign algorithm against impulsive interferences, but also obtain a significant improvement in the performance as compared to their counterparts (i.e., SSAF and IPSSAF), in terms of the convergence rate and tracking capability. Besides, the IWF-IPSSAF algorithm has faster convergence rate than the IWF-SSAF algorithm for sparse impulse responses. Finally, the performances of two proposed algorithms are demonstrated in the system identification and the acoustic echo cancellation with double-talk.

A Review on Noise and Denoise of Biometric Images

“…To additionally advance the performance of the IWF-SSAF, an upgraded comparable IWF-SSAF (IWF-IPSSAF) system is suggested to additionally upgrade the enactment of the IWF-SSAF. Simulation outcomes proved the superiority of the offered technique [30].…”

Novel combination schemes of individual weighting factors sign subband adaptive filter algorithm

“…19,32 The Gaussian white background noise v(n) is added to the output end of the unknown system to give a signal-to-noise rate (SNR) of 30 or 20 dB. In all the simulations, the unknown vector w o is a measured acoustic echo path with M = 512 taps except for Section 4.3.…”

“…19,32,35,36 This assumption leads to that (1) the subband noises ϑ i (k), i = 0 , 1 , . The filter bank used for decomposing the input and desired signals is paraunitary.…”

“…4,6,11 However, the convergence performance of these algorithms is severely deteriorated when the input signal is correlated (also called the colored input signal). 6,7,[16][17][18][19] To efficiently suppress the impulsive noise, the sign algorithm (SA) 16 and its variants 6,17 were proposed by optimizing the l 1 -norm cost function, but their convergence rates are very slow for the correlated input. In the SAF, the correlated input signal is partitioned into almost mutually exclusive multiple subband signals by the analysis filters; then, the decimated subband signals close to white are used to update the filter's weight vector, thus improving the convergence rate.…”

“…[13][14][15] Regrettably, since the above-mentioned algorithms are based on the l 2 -norm optimization strategies, their performance is seriously damaged by the impulsive noise, which is often encountered in practice. 6,7,[16][17][18][19] To efficiently suppress the impulsive noise, the sign algorithm (SA) 16 and its variants 6,17 were proposed by optimizing the l 1 -norm cost function, but their convergence rates are very slow for the correlated input. For this purpose, by minimizing the l 1 -norm of the subband error vector of the filter, Ni et al proposed a sign subband adaptive filter (SSAF) algorithm and a variable regularization parameter SSAF (VRP-SSAF) algorithm, 20 which not only speed up the convergence rate but also are robust against the impulsive noise.…”