K-Means Clustering-Aided Non-Coherent Detection for Molecular Communications

TL;DR

This paper introduces a novel non-coherent detection method for molecular communication that employs K-means clustering to estimate thresholds and improve bit-error-ratio performance without requiring channel state information.

Contribution

It proposes a K-means clustering-based approach for non-coherent detection in molecular communications, enabling threshold estimation and bit retrieval without explicit channel knowledge.

Findings

Clustering-based thresholds improve detection accuracy.

Iterative clustering reduces bit-error-ratio.

Simulation confirms effectiveness of the proposed method.

Abstract

In this paper, we consider non-coherent detection for molecular communication systems in the presence of inter-symbol-interference. In particular, we study non-coherent detectors based on memory-bits-based thresholds in order to achieve low bit-error-ratio (BER) transmission. The main challenge of realizing detectors based on memory-bits-based thresholds is to obtain the channel state information based only on the received signals. We tackle this issue by reformulating the thresholds through intermediate variables, which can be obtained by clustering multi-dimensional data from the received signals, and by using the K-means clustering algorithm. In addition to estimating the thresholds, we show that the transmitted bits can be retrieved from the clustered data. To reduce clustering errors, we propose iterative clustering methods from one-dimensional to multi-dimensional data, which are shown to reduce the BER. Simulation results are presented to verify the effectiveness of the proposed methods.