Proportionate Kalman Filter for Model-Based Channel Tracking in Underwater Acoustic Communications

• Published in: OCEANS 2021: San Diego – Porto pp. 1 - 5
• Main Authors: Wang, Yuxing, Cao, Hongli, Tao, Jun, Yang, Le, Qiao, Yongjie
• Format: Conference Proceeding
• Language: English
• Published: MTS 20.09.2021
• Subjects: AR model; Channel estimation; Degradation; Kalman filters; model-based channel tracking; Oceans; Performance gain; proportionate Kalman filter; Sparse matrices; sparse UWA channel; Steady-state
• DOI: 10.23919/OCEANS44145.2021.9705832

Model-based channel tracking methods have been widely investigated for underwater acoustic (UWA) communications. Among them, the state-space model-based approach is able to effectively handle the temporal evolution of the channel using e.g., an autoregressive (AR) transition model. However, existing state-space model-based channel tracking techniques generally do not take advantage of the sparsity in the channel taps and may suffer from performance degradation under sparse UWA channels. In this paper, we focus on enhancing a developed channel tracking technique that utilizes the AR mode as the state transition model, through taking into consideration the channel sparsity. Specifically, we integrate proportionate updating into the standard Kalman filter (KF) adopted for the channel estimation and tracking. This leads to the proposed proportionate Kalman filter (PKF), where a proportionate matrix is applied to the Kalman gain to make each channel tap have a unique updating step size proportional to its magnitude. As a result, improved convergence speed and steady-state performance are obtained. Both simulation and experimental results verify the performance gain of the proposed PKF under sparse UWA channels.