摘要： 在已有的FDR滤波算法基础上，提出一种分层FDR阈值滤波算法。该算法对小波分解后最底层的低频系数和各层的高频系数分别进行多假设检验，并确定出多个阈值，分别进行阈值化处理。对heavysine, bumps, doppler, blocks信号进行仿真实验，结果表明，该分层FDR阈值滤波算法在信噪比、相对误差等方面均优于通用阈值算法、FDR算法及自由分布式FDR算法，具有较高的实用价值。

关键词: 滤波, 小波, 阈值, 假设检验, 错判率, 分层

Abstract: Based on the FDR threshold filtering algorithm, a layered FDR threshold filtering algorithm is put forward. Through respectively making multiple hypotheses testing to the low frequency coefficients at the last level and the high frequency coefficients at other levels, multiple threshold values are obtained. The gained threshold values are taken as the criterion at the individual level. Simulation results for signals of heavysine, bumps, doppler, blocks show the proposed threshold algorithm is better than universal threshold, FDR and free distributed FDR threshold filtering algorithm in signal to noise ratio, relative error and so on. It has higher practical value.

Key words: filtering, wavelet, threshold, hypotheses test, false discovery rate, layered

中图分类号: 

• TP911.7