Non-cooperative Receiving Schemes For Frequency Hopping Signals

Frequency hopping (FH) communication system is widely used not only in military area but also in civil area for its excellent characters, such as anti-jamming, anti-fading and high security, along with which FH signals analysis and processing became hot both in and abroad. In this paper, some of receiving technology in noncooperative mode of frequency hopping signals is mainly studied, including methods for parameters estimation and schemes for detection and classification. The parallel calculation of the classification algorithm is finally realized on multiple DSPs. The main tasks completed are lined out as follows.1. Some kinds of time-frequency analysis are studied.The advantages and disadvantages of them are analyzed and compared, and short time Fourier transformation (STFT) is selected as the precondition of frequency hopping signals processing for its high efficiency and cross-term immunity.2. Blind estimation of parameters of FH signal is mainly researched. Firstly, a method based on cost function is designed for hop rate estimation which improves the performance of the methods based on time-frequency peak curve when few hops are included within oberserving time. Then, in order to enhance the performance under low SNR condition of the method based on time-frequency backbone curve, modification is made by substituting maximum positions after smoothing for original backbone curve. Having obtained the hop rate, the hop timing (epoch) and hop pattern are finally acquired by enlarging the length of smoothing window, and the performance is good compared with the folding method.3. Blind detection and classification of FH signals in complex electromagnetic environment with all kinds of interferences are discussed. First of all, one dimentional CASH-CFAR is amended and expanded to the two dimentional case for the purpose of making frequency hopping signals well-marked by restraining long term signals and wideband signals. Then, opening and closing operation of morphological technology in image processing are taken into account for smoothing signals and eliminating noises. Finally, utilizing paramenters of each segment acquired from fast FFT direction finding and segementation, two schemes to classify FH signals are proposed. One is based on clustering, and the other is based on time correlation of segments. For the clustering scheme, KHM algorithm and number of clusters estimation methods are modified to achieve less sensitive performance, which is well-used in FH signals classification. For the time correlation methods, the definition of time correlation is revised and the histogram is generalized. The presented modification can solve the problem of being the same frequency with adjacent hops and correctly classify FH signals in complex electromagnetic environment without knowledge of channel switching time, and meanwhile avoiding resorting process.4. Parallel calculation of classification algorithm of FH signals is carried out on multiple DSPs.Characteristics of DSP chips TS201S used in this paper are firstly shown. Then tasks assignment for 4 TS201S is detailed with corresponding flow charts. Finally, the classification algorithm is debugged successfully and is found to meet system demands.