Compressive Sampling for Recognition of Frequency-Hopping Spread Spectrum Signals

Abstract

In this paper, a method for recognition of frequency-hopping spread spectrum (FHSS) signals from compressive measurements is proposed. Conventionally, the detection method based on energy or signal variance does not represent the signal type. Other energy signals may also be falsely detected as FHSS signals, especially in the case of dynamic spectrum sharing. Besides, full FHSS spectrum scanning also makes data storage and processing challenges. In this paper, we propose a compressive detection mechanism based on pattern recognition to alleviate the stress of wide-band signal processing and detection category ambiguity. A partial discrete Fourier transform (DFT) matrix is designed based on the maximum mutual information between the received signal and the compressive measurements. Besides, to improve detection performance, a Toeplitz matrix developed as a filter. The DFT matrix and Toeplitz matrix together construct the measurement kernel. We extract the frequency domain features of received signals and detect the FHSS signals by a KNN classifier. The simulation results demonstrate that our compressive detection system can effectively detect FHSS signals and is insensitive to other signals. Even at low SNR, high detection probability and low false alarm probability can be obtained.