Based On The Frequency Overlapping Signals In The The Cyclostationary Single Channel Analysis

Wireless communication in the future will encounter serious challenges from spectrum resource excessively ocuupied inevitably. In many practical environments such as radio communications, telemetry, sonar and speech contexts, the observed signals from the co-channel are usually time-frequency overlapped multi-signal. As a result, signal analyses in such situations will become a very problem demanding prompt solution. Single-antenna receiver has prime advantages of lower cost and better performance. This paper demonstrates Cyclostationary theory and model for time- frequency overlapped signal in co-channel, and researchs three aspects including parameter estimation,modulation recognition,blind separation technology.This paper introduces cyclic spectrum and its SSCA algorithm implementation, analyses the multi-path effect on impact on it, proposes an algorithm based on cyclic spectral features to solve the problems in modulation classification for time-frequency overlapped two-signal in the combination of BPSK, QPSK, OQPSK and MSK signals automatically. Each signal's features from the cyclic spectrum are extracted, and a multi-pole classifier is proposed to classify multi-signals. Both theoretical analysis and simulation via Matlab verify that the algorithm is reasonable and efficient.This paper introduces cyclostationary based parameter estimation algorithm. Cyclostationary reflects signals'potential periodicity, such as symbol rate and carrier frequency. A novel approach is proposed based on 4-th cyclic cumulant. Firstly search cycle frequency bins until one signal information achieved, a new cyclic cumulant field is reconstructed, then search maximum bin in the new field as two signal carrier frequency. Experiments indicate that the method is reliable and all-purpose.To solve the problem in modulation recognition for two-signal combining QPSK with 16QAM, this paper proposes a novel algorithm using the combination of diffient- orders cyclic cumulants as robust classification features. Designing SVM classifier can map the 2-D features to higher dimension super-space. When the input data is sufficient, the algorithm can classify the above two-signal effectively in low SNR.In the end, this paper introduces two FRESH filter structures and theory. Especially blind adaptive FRESH filter which only needs some cyclic frequencies knowledge can separates spectrum-overlapped signals effectively blindly. Exbanding to time-frequency overlapped signal in co-channel, this method has good performance.