Hybrid CNN-BiLSTM for ISI Mitigation in Molecular Communication for Nanosensors with Imperfect Transmitter

Inter-symbol interference (ISI) mitigation is a critical challenge in diffusion-based molecular communication systems for nanosensors due to the inherent channel memory at the nanoscale. In the presence of an imperfect transmitter, the emitted molecules include interference components, further complicating signal decoding at the receiver. Traditional ISI mitigation techniques often rely on precise channel state information, which is time-varying and computationally demanding. To address this issue, we propose a hybrid deep learning model, C-BiLSTM, which integrates convolutional neural networks (CNN) for local feature extraction and bidirectional long short-term memory (BiLSTM) networks for capturing long-range dependencies. The model is trained and decoded using received time-window sequences to enhance detection performance. Simulation results demonstrate that C-BiLSTM achieves a lower bit error rate (BER) than CNN alone and approaches the optimal detection performance. These findings highlight the potential of deep learning-based approaches for ISI mitigation, offering a promising solution for enhancing reliability in molecular communication systems.