Fast Algorithms Of Time Frequency Atom Decomposition For Radar Emitter Signals

Time-frequency atom decomposition(TFAD) algorithm is a new method in signal processing, which was developed after Fourier transform, Gabor transform and wavelet transform. TFAD decomposes a signal into a linear expansion of waveforms that are selected from a redundant dictionary of functions. TFAD algorithm is general procedures to compute adaptive signal representations. Because of the redundant dictionary and the greedy decomposition strategy, the input signal is decomposed into the sum of dictionary elements which is best adapted to approximate parts of the signal. The signal can be flexible decomposed into waveforms whose time-frequency properties are adapted to its local structures. And a signal energy distribution is derived in the time-frequency plane, and does not include interference terms, unlike Wigner and Cohen class distributions.The main problem of TFAD is its high computational complexity which restricts the application of radar emitter signal processing. The only way to improve the efficiency of the signal processing is to explore better and more efficient time-frequency atom decomposition methods. Until now, relatively limited work have been done on the method's application, what is used to processing radar emitter signal. Aiming at solving the same problem, three fast TFAD algorithms are presented in this dissertation. The main work and research are as follows.1. The principle of TFAD algorithm is introduced. And then the shortcomings of conventional signal processing methods are analyzed, when they are applied to the non-stationary signals.The performances of the Chirp atom and Gabor atom. The experimental results show that the TFAD based on Chirp atom has a high time-frequency resolution of reconstructed signal. And using TFAD to processg non-stationary random signals are better than using traditional methods.2. In order to decrease the high computational complexity, a novel fast TFAD algorithm based on quantum-inspired genetic algorithm (QGA) is proposed. Making full use of QGA's advantages such as good global search capability, rapid convergence and short computing time. The method builds dictionaries by using chirp atoms which can best match the original signal and uses QGA to reduce time complexity. Experiments conducted on radar emitter signals show that the introduced method not only can release the computational burden but also can obtains time-frequency concentration of reconstructed signal, compared with the traditional TFAD algorithms.3. Based on particle swarm optimization algorithm(PSO), a high efficiency fast TFAD algorithm for radar emitter signal is presented. The method builds time-frequency atom dictionaries by using chirp atoms which can best match the original signal, and uses PSO to reduce time complexity. Experiments conducted on radar emitter signals show that the introduced method can decrease the computational burden. This method can effectively control noise and interference terms as well.4. To enhance the good global search capability of PSO, a fast TFDA based on chaos particle swarm optimization (CPSO) algorithm. The method builds time frequency atom dictionaries by using chirp atoms which can best match the original signal, and the initial position of the particle is evaluated by chaos. At the same time, the chaos "catastrophic" operation can drag the search out of local extremum by introducing the strategies of chaos catastrophe. Experiments conducted on radar emitter signals show that the introduced method have good performance on controling noise and interference terms, especially in multiple signals processing.This paper was supported by the National Natural Science Foundation of China (Grand No. 60702026) and Youth Science and Technology Foundation of Sichuan (09ZQ026-040).