基于粒子滤波的单通道正弦调频混合信号分离与参数估计*

摘要

A signal separation and parameter extraction method based on particle filtering for single channel sinusoidal fre-quency modulated (SFM) signals is put forward. By assuming that the frequency of SFM signals mixture is continuous, a phase-difference de-aliasing arithmetic based on particle filtering is proposed. And the dimension of state space is reduced by using phase-difference between source signals. A likelihood function model suitable for high dimensional state space is proposed. Particles weight is accurately measured by comparing error between estimated values and true values of particles with fixed length. The problem of particle diversity reduction in the static parameters situation is solved by the introduction of Markov-chain Monte Carlo (MCMC) transfer after re-sampling, and the speed of particle filter iteration convergence is also effectively improved. Single channel SFM signal parameters are extracted and signals are separated by reconstructing signals only with the prior knowledge of modulation type. Finally, the simulation results indicate that this method can separate the multi-component signal sources and estimate the parameters effectively.%以单通道正弦调频(SFM)混合信号为研究对象，提出了基于粒子滤波的正弦调频混合信号分离与参数提取方法。针对正弦调频混合信号频率无跳变的特征，提出了一种基于粒子滤波的相位差解混叠算法，并通过源信号相位差解决了本算法中粒子滤波高维状态空间降维问题，提出了一种适合高维状态空间的似然函数模型，比较固定长度粒子估计值和真实值误差，进而准确衡量粒子权重。通过在重采样后引入MCMC转移，解决了静止参数下粒子多样性降低问题，有效提高粒子滤波迭代收敛速度。从而在先验知识仅已知信号调制方式的情况下，完成对单通道正弦调频混合信号的参数提取，并通过重构信号完成正弦调频混合信号分离。最后通过仿真分析发现，该方法能够有效的实现正弦调频混合信号的分离与参数估计。