Real-Time Recognition and Parameters Estimation of Linear Frequency Modulation Microwave Signal Based on Reservoir Computing

Abstract

Real-time waveform recognition and parameter evaluation for linear frequency modulated (LFM) pulse waveform is crucially important while challenging in microwave detection systems. To address this issue, in this work, we proposed a new artificial intelligence enabled classification method based on reservoir computing (RC). A sampled sequence, generated by random concatenation of LFM signals with different chirp rates and initial frequencies, is used to training the designed reservoir with 200 nodes. The testing result shows that the RC can recognize individual LFM signals in the sequence, and estimate the instantaneous frequency of an LFM signal within the sequence. Compared to conventional computing methods for instantaneous frequency identification such as Hilbert transform or short-time Fourier transform, RC-based approach features faster speed and great potential for hardware implementation using photonic devices.