The Particle Filter Based On The Multi-component Fm From The Reconnaissance Signal Separation And Parameter Extraction

Single-channel signal separation problem is widespread in radio detection, separation of speech signals, physiological signals denoising and other fields. Since the receiver is only one sensor,independent component analysis and other signal separation algorithms based on array sensor are not applicable.Particle filter algorithm highly adapt to nonlinear systems and non-gaussian noise environment with simple filtering method,it can solve nonlinear and complex problems in multi-parameter estimation. Therefore, the application of particle filter algorithm in single-channel signal separation problems have great significance.In this dissertation, the basic theory of particle filter are introduced,problems exist in the application of particle filter for mixed-signal separation are detailed analyzed.Then a single-channel multi-component SFM mixed-signal parameter estimation method is introduced.An idea of reduce the dimension of state space by using phase difference between source signals, recovering phase difference of the source signals from orthogonal signal, using particle filter to estimate some parameters related to the phase difference is proposed.The above method reduce the dimension of state space,and solve the particle filter problem in high-dimensional state space. A particle filter likelihood function model under high-dimensional space is presented.Particles weight are accurately measured by comparing error between particles estimates and actual values.Particle diversity reduce problem in static parameters situation is solved by the introduction of MCMC transfer after re-sampling, and particle filter iteration convergence speed is effectively improved. Based on the estimated parameters from phase difference, remaining parameters are estimated by particle filter from time-domain waveform, thus all parameters of multi-component SFM mixed-signal are extract.Then the method of estimate parameter of single-channel multi-component chirp mixed-signal and single-channel chirp mixed with SFM signal are introduced. The problem of reducing the state space dimension in particle filter during estimate parameter from chirp and SFM mixed-signal phase difference is solved.The impact of symbol ambiguity during recovery of phase difference to the parameter estimates and how to fix it are analyzed. Finally, a hardware simulation platform for single-channel mixed-signal parameter estimation algorithm is established, ideas of each module are introduced. Implement particle filter parameter estimation algorithms in this platform,and achieve good results.