自适应旋转投影分解法

Adaptive Oriented Orthogonal Projective Decomposition

本文提出了一种新的时一频分解方法──自适应旋转投影分解法（ＡＯＰ法）．在表征信号空间的线性调频高斯信号集上，我们针对原始信号自适应地搜索出一组与信号匹配最好的基函数序列．以此用尽可能少的基函数来重构信号子空间．根据分解系数，得到信号的时－频能量分布．由于调频高斯信号时频会聚性能极佳，又能灵活高效地匹配各类信号，该算法不论从分辨率、效率还是描述能力等方面都具有良好的性能．将它用于语音压缩也取得了很好的结果．

In this paper,we present a new algorithm,called adaptive oriented orthogonal projective decomposition (AOP),to decompose a signal into a sum of Gauss functions. The elementary functions used in AOP are the dilations,modulations and translations of a normalized Gauss function. They can be adjusted to best match the original signal. We derive a Time-Frequency energy distribution, by adding the Wigner-Ville distribution of the selected elementary functions. Since the Gauss functions are well localized in the time-frequency domain,the AOP distribution has high resolution,no crossterm and no negative energy. It's a powerful tool of time-frequency analysis.