Reduced-Complexity Estimation of FM Instantaneous Parameters via Deep-Learning

Signal frequency estimation is a fundamental problem in signal processing. Deep learning method to solve this problem. This paper used five deep methods and three datasets including different singles Single Tone (ST), Linear- Frequency-Modulated (LFM), Quadratic-Frequency-Modulated (QFM). affected by Additive White Gaussian (AWG) noise Symmetric alpha Stable (SαS) noise. Geometric SNR (GSNR) determine the impulsiveness of SαS mixture. are Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), Bi-Direction (BiLSTM), Convolution Neural Network (1D-CNN & 2D-CNN). When compared classifier with few layers get on high accuracy complexity reduces for Instantaneous Frequency (IF) estimation, Linear Chirp Rate (LCR) Quadratic (QCR) estimation. IF ST signals, LCR LFM IF, LCR, QCR QFM signals. The dataset GRU 58.09, LSTM 46.61, BiLSTM 45.95, 1D-CNN 51.48, 2D-CNN 54.13. 82.89, 66.28, 20%, 74.79, 98.26. 78.76, 67.8, 69.91, 75.8, 98.2. results show that better than other parameter signals