Demodulation Scheme Against Phase Noise Using an Ensemble Clustering Approach

Most wireless systems involve time-varying multipath channels, where the negative effect of the phase noise cannot be ignored. The phase noise can be introduced by imperfect phase-locked loop circuitry, imperfect channel estimation, or both. This paper considers the demodulation problem of wireless systems with phase noise. We propose an ensemble clustering algorithm to address the limitations of the existing demodulation schemes. The proposed ensemble clustering algorithm is named as the Ensemble Clustering algorithm using the Matrix Factorization and Information Theory (MFIT-EC). The MFIT-EC algorithm improves the performance of existing ensemble clustering algorithms, which neglect different clustering effects of different base-clustering algorithms or different effects of different clusters of the same base-clustering algorithm. Based on the MFIT-EC algorithm, we design a demodulation scheme in a coherent wireless system with phase noise, referred to as the MFIT-EC demodulation scheme. Specifically, we first utilize several base-clustering algorithms to obtain different base-clustering results. Second, we use a weighted ensemble mechanism to allocate different weights to different base-clustering results, and calculate the certainty of the clusters of each base-clustering algorithm to obtain a better and more robust demodulation performance. We implement the proposed approach and conduct extensive performance comparisons through simulations, which show that the proposed approach has better performance than the prior approaches in terms of both clustering performance and demodulation performance.