Unsupervised Clustering-based Non-Coherent Detection for Molecular Communications

As an emerging communication paradigm, molecular communication is suitable for exchanging information between nano-machines in a fluid medium (e.g., in vivo), utilizing molecules as a major carrier. Subject to the inherent long tail characteristic of channel response and the difficulties in obtaining Channel State Information (CSI), conventional coherent detection schemes prevalent in wireless scenarios are no longer attractive. To achieve reliable communication, this letter proposes a novel non-coherent detection method based on unsupervised Fuzzy C-means (FCM) clustering exploiting the transient features of molecular signal for Molecular Communications via Diffusion (MCvD). Numerical results demonstrate that the performance of bit error rate (BER) is significantly improved by the proposed method compared with other non-coherent schemes and is comparative to the optimal coherent method. Especially, in the scenario of multi-level modulation, almost all threshold-based non-coherent detection schemes perform badly except for our method.