Support Vector Machine-Based Wireless Channel Classification for Adaptive AFC in LTE Downlink

Abstract

An adaptive control scheme of digital loop filter for automatic frequency control (AFC) is proposed to improve LTE downlink throughput in high-speed train (HST) channel. A mobile device in HST channel suffers from performance degradation due to the abrupt change in Doppler shift. In this case, a higher loop gain is desirable to quickly track the Doppler shift. However, in general, there is a fundamental trade-off between locking time and phase jitter performance in AFC. Therefore, it is crucial to identify the channel condition for adjusting AFC loop gain. In this paper, a wireless channel classification problem is formulated and the optimal classifier is derived by using support vector machine (SVM) algorithm. In particular, HST and non-HST channel environments are classified by exploiting the statistical distribution of the phase estimates. Experimental results show that the SVM-based adaptive loop gain control improves the throughput in HST channel.