Error coded affine projection algorithm

Error Coded Affine Projection Like Algorithm

“…The error codification has been applied to LMS algorithm [7], and the results show a lower computational load and a faster convergence speed, also in [4] is presented the AP algorithm with coded error where the minimum mean square error is improved and the computational load decreases. The APL-I algorithm is an algorithm that does not requiere the inversion of the input signal matrix, however, this algorithm needs to calculate the filter's coefficients every iteration, therefore the proposal of codifying the error, in order to reduce the computational load.…”

“…To demonstrate t of the proposal, all experiments were some APA variants that reduce com such as the Set Membership APA which that updates their coefficients based on bound [2] and the ECAPA [4] (this chosen because it uses the same princi the error). Figure 3 shows the results of the error coded to 8 bits and using a projec where it can be observed that APL slightly faster convergence speed th algorithm, however, the most importan Table 2 and Table 3 In figure 5 can be seen the effects the error is coded with different num projection order used for all the algor and the number of bits used to codify 12.…”

“…In some applications the use of large finite impulse response adaptive filters (FIR) is necessary in order to provide a correct output signal, therefore the filter requires several hundred or thousand taps, if the filters are implemented in a digital system they will be high accurate, however, the processing speed will fall off owing to the high number of operations needed. To counteract this problem there has been proposed some algorithms which improve the convergence speed and reduce the computational load [4].…”

“…The AP algorithm presents faster convergence speed than the Least Mean Square (LMS) and the Normalized Least Mean Square (NLMS) algorithms, nevertheless its computational complexity and the misadjustment level are bigger [4]. AP algorithm contains matrix inversion which makes it more computationally complex than the LMS algorithm [5].…”