摘要
根据有限高斯混合模型可以逼近任意概率分布密度函数的思想,提出了一种基于高斯混合模型的非平稳信号粒子滤波时频分析算法。本方法兼顾了算法在频率缓变时的估计精度和频率突变时的动态性能,并结合一种简化的TVAR模型,通过降低估计量维度,较大幅度地改善了计算性能,满足了对非平稳信号进行在线时频分析的要求。实测数据的时频分析试验证明了本方法的优良效果。
According to that finite Gaussian mixture model could approximate any probability density function,a time-frequency estimation algorithm using particle filter with Gaussian mixture model is proposed for non-stationary signals,taking into account of both the algorithm＇s estimation accuracy and dynamic performance.Combined with a simplified TVAR model,the computing performance of proposed algorithm was further improved,because the frequency could be estimated directly and the estimating dimensions were quite reduced.Experimental result from the measured signal is presented to demonstrates that the proposed method has great precision,quick response and real-time character.
出处
《测控技术》
CSCD
北大核心
2011年第3期83-86,共4页
Measurement & Control Technology
基金
航空科学基金项目(05I53062)
关键词
粒子滤波
时频估计
高斯混合模型
particle filter
time-frequency estimation
Gaussian mixture model
