CModel: An Informer-Based Model for Robust Molecular Communication Signal Detection

Molecular communication signal detection faces numerous challenges, including complex environments, multi-source noise, and signal drift. Traditional methods rely on precise mathematical models, which are constrained by drift speed and signal-to-noise ratio. To address these issues, this paper proposes an innovative detection model based on the Informer architecture, named ComModel (CModel). This framework integrates probSparse Attention, Cross Attention, and convolutional layers to enhance detection accuracy and adaptability to various environmental conditions. Experimental results demonstrate that CModel consistently outperforms traditional deep neural networks and Transformer-based models, especially in complex scenarios with varying drift speeds and noise levels. As the drift speed increases, CModel maintains superior stability and exhibits lower bit error rates, particularly at medium and high drift speeds. Moreover, CModel shows excellent performance in environments with significant noise. Overall, CModel demonstrates robust and reliable signal detection capabilities in multi-noise environments.