Deep Learning-Enhanced Anti-Jamming Decoder for OTFS Systems: A CNN Bi-LSTM Hybrid Approach
Yangyang Li, Yuhua Xu, Guoxin Li, Songyi Liu, Wenhui Liu, Jiachen Hu · IEEE Transactions on Cognitive Communications and Networking · 2025
This paper addresses the challenge of anti-jamming in orthogonal time frequency space (OTFS) modulation systems by proposing a novel anti-jamming decoder. The design of this decoder presents significant challenges, including the integration of spatial and temporal feature extraction into a cohesive architecture and ensuring convergence stability in rapidly changing jamming environments. Our approach leverages the strengths of convolutional neural networks (CNNs) and bidirectional long short-term memory networks (Bi-LSTMs) to effectively capture both the spatial and temporal characteristics of OTFS signals. This hybrid model allows the system to adapt to varying levels of jamming dynamically and sustain reliable communication. We incorporate batch normalization and soft updates to enhance convergence stability and ensure robust training. Our solution is designed to balance computational efficiency with performance, making it suitable for deployment across various communication devices. In simulations, our decoder demonstrates significant resilience against jamming, achieving similar bit error rates (BER) to CNN decoders under nearly 30 dB of jamming, with only moderate computational requirements.