基于遗传算法的EBFB小波滤波器优化设计

Optimization of EBFB Wavelet Filter Design Based on Genetic Algorithms

EBFB是基于伯恩斯坦多项式的偶数长滤波器组。针对伯恩斯坦多项式中参数的不确定性,提出了一种基于遗传算法的具有EBFB特征的小波滤波器优化设计方法。该方法在保证滤波器组具有完全重构性和消失矩特性的前提下,设定伯恩斯坦多项式中不确定性参数的个数,采用遗传算法对参数进行分阶段优化,得到任意长度的反对称双正交小波滤波器组。实验结果表明,利用此方法设计的小波滤波器具有较大的阻带衰减和良好的数据压缩性能。

Even-length Bernstein filter banks （EBFB） is a new class of even-length filter banks based on Bernstein polynomial. Due to the uncertainty of parameters of the Bernstein polynomial, a design method for optimizing wavelet filter while maintaining the property of EBFB based on genetic algorithms is presented. Under the precondition of guaranteeing the property of perfect reconstruction and vanishing moments, this method first sets the number of the uncertain parameters in Bernstein polynomial, then optimizes them by genetic algorithms, a biorthogonal filter banks with arbitrary length and anti-symmetry is thus obtained. The experimental result shows that the method proposed has more stopband attenuation and good data compressing performance.