摘要
本文推导了最小延迟任意长度M带余弦调制小波滤波器组的完全重构条件。选择低通原型滤波器最大阻带衰减为优化的目标函数,通常的优化目标函数选用最小平方逼近的方法,本文提出了使用最佳一致逼近的方法。最后用黄金分割和牛顿迭代方法解决非线性约束优化极值问题,得到满足几乎完全重构和小波正则性条件的低通原型滤波器。
A method for designing lowest-delay arbitrary length M-band cosine-modulated wavelet perfect reconstruction filter banks is deduced. The stopband error of prototype filter is used to be the objective function for optimization. Instead of the method of least square approximation, this paper introduce the method of best uniformity approximation. We use the method of golden section and Newton iteration to solve the nonlinear constrained minimization problem. As a result, the prototype filter of near perfect reconstruction and wavelet regular condition could be obtained.
出处
《信号处理》
CSCD
2003年第3期242-246,共5页
Journal of Signal Processing
