| 摘要 |
PROBLEM TO BE SOLVED: To obtain an adaptive filter with fast convergence speed and with few residual errors after convergence by multiplying the absolute value of an averaged differential signal by a norm so as to obtain a product and control ling a step size. SOLUTION: At first, a vector sgn(en +vn )a<(n)> and the absolute value|en +vn|are calculated. Then, the vector q<(n)> is obtained by leakage cumulative adders 10-12 and the norm q<(n)2> is calculated by a norm calculator 40 by using the vector. In the meantime, a scholar quantity|σ|<(n)> is obtained by the leakage cumulative adder 32. Then, the norm ((q<(n)> ))<2> is multiplied by the scholar quantity|σ|<(n)> so as to calculate the product ((q<(n)> ))<2>|σ|<(n)> , the value is multiplied by a coefficientβand the step sizeαc <(n)> is obtained and outputted. Here, en expresses a differential signal group, vn expresses a noise group, (n) expresses time, an expresses an input signal group andαc <(n)> expresses the step size in the time (n).
|
