Blind Symbol Rate Estimation Via Wavelet-Based Deep Learning for Noncoherent FSK Signals

초록

In non-cooperative contexts, accurate symbol rate estimation is essential for demodulation and information recovery. This paper proposes a deep learning-based algorithm for the blind estimation of the symbol rate in noncoherent frequency-shift keying signals. In the proposed approach, the received signals are first subjected to wavelet transform-based preprocessing with frequency normalization, from which discriminative features are extracted. The extracted features are subsequently utilized to train a deep learning model for symbol rate estimation. Simulation results demonstrate that the proposed algorithm outperforms a conventional wavelet transform-based algorithm and provides reliable estimation performance under various signal-to-noise ratios, unseen symbol rates, and frequency-shift conditions.

키워드