A Weak Signal Detection Method Based on Artificial Fish Swarm Optimized Matching Pursuit

Abstract

To detect weak signals is difficult in signal processing and is very important in many areas such as non-destructive evaluation (NDE), radar etc. Sparse signal decomposition from overcomplete dictionaries are the most recent technique in the signal processing community. In this paper, this technique is utilized to cope with ultrasonic weak flaw detection problem. But its calculation is huge (NP problem). A new improved matching pursuit algorithm is proposed. The mathematical model of searching algorithms based on artificial fish swarm is established; the artificial fish swarm with the advantages of distributed parallel searching ability, strong robustness, good global astringency, and insensitive preferences are employed to search the best matching atoms. It can reduce complexity of sparse decomposition and space of memory. Experimental results shows that the amplitude, frequency and initial phase parameters of ultrasonic signal blurred by strong noise can be estimated according to the proposed algorithm, and the expected weak signal can be then reconstructed. When this method is used in the ultrasonic flaw detection, compared with the wavelet entropy and wavelet transform, the results show that the signal quality and performance parameters are improved obviously.