基于DCT的实值离散Gabor变换最优窗宽选择

Optimal Window Width Selection for Real-valued Discrete Gabor Transform Based on DCT

文中针对Gabor变换中最优窗函数宽度选择的问题,提出了以提高Gabor表示的聚集性和时频分辨率为目的的基于DCT核的实值离散Gabor变换最优窗函数宽度选择算法。对香农熵的取值范围进行了研究,使其更适合度量时频分布的聚集性,进而根据熵度量实现了与信号非平稳性相适应的最优窗函数宽度的选择。实验结果表明,该算法对单分量及多分量信号都能有效地选择最优窗函数宽度,同时能够获得聚集性好、时频分辨率高的Gabor表示,并且具有很好的抗噪能力。

Aiming at the problem of optimal window width selection in Gabor transform,to improve the concentration and time-frequency resolution of Gabor representation,an algorithm for the optimal window width selection of the DCT-kernel-based real-valued discrete Gabor transforms is proposed. The value range of Shannon entropy is studied to be more adequate for measuring concentration of time-frequency distribution. Moreover,based on the entropy,an optimal window width adapted to signal non-stationary can be searched. Ex-perimental results show that the proposed algorithm gives the best Gabor representation with high concentration and time-frequency reso-lution for mono-component signal or multi-component signal. Additionally,the proposed algorithm also performs well under low signal noise ratio.