Learning to Equalize: Data-Driven Frequency-Domain Signal Recovery in Molecular Communications

TL;DR

This work tackles ISI and noise in molecular communications by introducing a data-driven, frequency-domain equalizer based on LSTM (LSTM-FDE) that operates without explicit CSI. By training to model temporal correlations directly in the frequency domain, LSTM-FDE achieves superior BER performance compared to model-based FDE and other data-driven baselines, while maintaining scalable complexity suitable for long channel memories. The method demonstrates strong robustness to channel estimation errors and excels in fast transmission scenarios where ISI is severe. Its data-driven nature and favorable performance–complexity trade-off suggest significant practical impact for MC systems, including biomedical applications, with future work focusing on lightweight architectures and MIMO extensions.

Abstract

In molecular communications (MC), inter-symbol interference (ISI) and noise are key factors that degrade communication reliability. Although time-domain equalization can effectively mitigate these effects, it often entails high computational complexity concerning the channel memory. In contrast, frequency-domain equalization (FDE) offers greater computational efficiency but typically requires prior knowledge of the channel model. To address this limitation, this letter proposes FDE techniques based on long short-term memory (LSTM) neural networks, enabling temporal correlation modeling in MC channels to improve ISI and noise suppression. To eliminate the reliance on prior channel information in conventional FDE methods, a supervised training strategy is employed for channel-adaptive equalization. Simulation results demonstrate that the proposed LSTM-FDE significantly reduces the bit error rate compared to traditional FDE and feedforward neural network-based equalizers. This performance gain is attributed to the LSTM's temporal modeling capabilities, which enhance noise suppression and accelerate model convergence, while maintaining comparable computational efficiency.

Figures (6)

• Figure 1: Structure diagram of LSTM-FDE.
• Figure 2: Block diagram of the LSTM-FDE training process.
• Figure 3: BER performance comparison of various FDEs when the symbol duration $T_b = 4*t_p$.
• Figure 4: BER performance comparison of various FDEs when the symbol duration $T_b = t_p/2$.
• Figure 5: BER performance comparison of various equalizers under channel estimation errors.