| 摘要 |
PROBLEM TO BE SOLVED: To reduce a residual error after convergence without reducing the convergence speed of an adaptive filter by controlling a step size at the time of obtaining an N-dimensional tap weight vector so that it becomes almost equal to the value minimum in the square average error at that time. SOLUTION: For the N-dimensional tap weight vector c<(n)> at time (n), which is realized as a non-cyclic filter, a probability grade sine algorithm given by an expression I is used as an input signal series whereαc <(n)> is set to be the step size at time (n), en to be an error signal series,νn to be a noise series, sgn to be a polarity function and a<(n)> to be time (n). A step size controller controlling the step sizeαc <(n)> in the expression I so that the square average errorε<(n)> at time (n), which is given by an expression II, becomes almost minimum is provided. Thus, the residual error after convergence is reduced and convergence can be speeded up.
|
