LFM signal detection based on STFT and frequency-domain GOSBOS-CFAR in low SNR

According to the continuous time-frequency characteristic of linear frequency modulated (LFM) signals, a detection method based on Short-Time Fourier Transform (STFT) and a new Constant False Alarm Rate (CFAR) detector is proposed. The input signals are short-time Fourier transformed to achieve coherent integration of frequency-shift sample sequences with complex envelopes, modular square of which are to be CFAR detected in frequency domain. On the basis of GO, the proposed Greatest Of and Statistics-Based Order Statistics (GOSBOS) algorithm exploits the a priori information provided by statistics of former frames for threshold adaptation. The scheme takes advantage of both OS´s superior detection performance and GO´s noticeable ability to control false alarms under non-homogeneous background. Simulation results show its fine detection performance in -15dB signal to noise ratio (SNR) condition, which can well meet the need of electronic reconnaissance.