Improved design of dual sign algorithm for adaptive identification

Optimum design of the hard-switching dual sign algorithm for adaptive echo cancellation

“…For the original HS-DSA and the two DSAs, the step size of the first SA is K1=1.599~10-~ and that of the second SA is Kz=59xKl. The moment of switching for the original HS-DSA is at the iteration [13] For the present example, R(0) and the corresponding Na should be 33.25 dB and 189 respectively, rather than 35 dB and 232 used in [13]. From Fig.10 it can be seen that the convergence performance of the optimized HS-DSA is clearly much better than those of the original HS-DSA and the DSA2.…”

“…We demonstrate in Fig.10 the simulation results of the convergence curves for the optimized HS-DSA, the original HS-DSA, and two DSAs with different transition thresholds (yt=0.1265 for DSAl and yt=0.0454 for DSA2) [13]. The DSA2 is based on the analysis results given in [13], while the DSAl is not. For the original HS-DSA and the two DSAs, the step size of the first SA is K1=1.599~10-~ and that of the second SA is Kz=59xKl.…”

“…As compared to the DSA1, the optimized HS-DSA obtains a smaller residual error with a similar convergence rate. It should be noted that the DSAl may have poor convergence performance for other cases, since it is simply an arbitrary design given in [13].…”

“…For the second adaptation interval of Fig.2, the initial residual echo to uncancelled signal ratio is With the numerical evaluation results of R(n)min (see Fig.5) and the closed-form formula of Nd(n)min (See (13) and (15)), we can optimize the threshold n for the fastest convergence by plotting Nt(n)=n+Nd(n)min (the total number of iterations required by the HS-DSA) against n, from which the minimum will give the optimum threshold nopt. The results for N=5, ua2=1, R,j=-17.2 dB, and several values of R(0) are shown in Fig.6.…”

“…Two SA variants called the dual sign algorithm DSA) and the hard-switching dual sign algorithm HS-DSA) have been presented in [12] and [13]. The DSA can be regarded as two SAS with two different step and noise cancellation [4].…”