Research On Signal Intra-pulse Modulation Analysis Algorithms Basedon Wavelet And Time-frequent Atom Dictionary

In order to distinguish multiple enemy radars, and get the parameters of each radar,we need to analysis the random overlapped signal pulse flow by signal sorting technology and separate them into single signals. Radar signals in new modulation methods continue to emerge, they have low-power,large bandwidth and good concealment and anti-jamming capability,which make sorting methods based on the classic 5 parameters(RF, TOA, PW, PA, DOA) for radar signal face enormous difficulties. Intra-pulse characteristics’(signal features of internal pulse) necessity is recognized in recent years. Using time-frequency analysis theory to acquire the Intra-pulse characteristics of the Intercepted signal is also a hot topic of researchment nowadays. Among the many kinds of signal modulations, Phase-coded signal and Chirp signal are typical LPI signals,which are the focus of this paper.The key to identifying PSK signals is mutation points detection. The paper uses the wavelet transform multi-resolution analysis(MRA) method, and Gabor transform to detect mutation points.For PSK signal identification, the key is to detect the mutation point.This paper studies the MRA algorithm and Gabor transform algorithm to detect point mutations.MRA algorithm has advantage lies in that it’s multi-scaled,so that it can use different waveforms to adapt to different mutation points. However, with the change in the carrier frequency signal, the deal result is unstable. Gabor transform has a very high time-frequency analysis capabilities,and a very stable deal result. However, in the case of an unknown signal carrier frequency, a two-dimensional Gabor transformation is needed, which takes a relatively long computing. In this paper, in order to shorten the processing time, the carrier frequency is estimated at first by FFT, and a one-dimensional Gabor transform is followed to detect mutation points.this algorithm has a very high precision,good stability and fast computing speed.The key of LFM signal identification is a intactly obtained linear feature of t-f energy distribution of the signal.In this paper,the distribution is obtained by STFT algorithm,and a linear fit operation is acted to get the center frequency and modulation slope. The method has a short processing time,however,it result in a distributed linearity feature,so it is sensitive to noise, and can’t analysis overlaped signals. TFAD algorithm can obtain the linear feature of LFM signal entirely,it has a good robustness to noise,and a capacity to analysis overlapped signals. But this algorithm needs a huge calculation,which is a big problem.In this paper,a kind of algorithm called by-dimesion-searching is proposed,which greatly improves the processing speed while ensuring a good analysis capacity,and can even get a better deal result than the classic algorithm in a same amount of time.