基于优化初值选择的自适应高斯包络线性调频基信号分解

基于高斯包络线性调频基的自适应信号分解是一种分辨力高、性能优良的时频分析算法,在语音信号、地震信号、雷达信号等可以用调频类函数进行建模的信号分析中有着广阔的应用前景。该算法在四维参数空间构造和求解超越方程,得到参数的闭式解,与传统的优化算法相比大大降低了运算量。对于由高斯包络线性调频基不相交或浅相交形成的简单信号,该算法具有非常优良的分辨性能,而对由高斯包络线性调频基深相交形成的复杂信号分解存在较大误差。本文针对这一问题进行研究,指出初值选择在该算法中的重要作用,分析了得到高精度分解结果的初值条件,提出了基于优化初值选择的高斯包络线性调频基自适应信号分解算法。通过在局部信号粗时频平面中搜索最优初值,结合自适应分解中建立和求解超越方程的方法得到参数闭式解,提高了分解精度,同时降低了运算量。对仿真信号及语音信号的分解效果验证了改进后算法的有效性和准确性。

Optimized Initial-Adaptive Gaussian Chirplet Signal Decomposition

The adaptive Gaussian chirplet decomposition algorithm,which has wide prospect in speech,seismology and radar sig- nal processing,has high resolution and good performance in time-frequency analysis.It decomposes signal through curve fitting in 4-di- mentional space,by which the complexity has been reduced greatly.The algorithm has good resolution for simple signals which are com- posed of disintersected or weak intersected Gaussian chirplets,however the results are not accurate when the original signal is composed of deep intersected Ganssian chirplets.In this paper,we point out that the initial point is very important for adaptive Gaussian chirplet decomposition,the initial condition for getting high accuracy is analyzed,and an Optimized Initial-adaptive Gaussian chirplet signal de- composition algorithm is proposed.Through positioning initial point by searching optimal initial point within the short time Fourier trans- form of a part of original signal,the precision is improved and the complexity is reduced.Numerical simulation and analysis on speech signal shows the effectiveness and accuracy.