On Wireless Channel Classification Based on CP-OFDM System

The channel feature identification plays a key role in booting the transmission efficiency through dynamically adjusting the transceiver parameters. In this paper, we propose a novel machine learning-based channel identification method for cyclic-prefix orthogonal frequency division multiplexing (CP-OFDM) systems. We first extract the features of the channel state information (CSI) in the time domain. On this basis, learning based classification methods, including K-nearest neighbor (KNN), support vector machine (SVM) and neural network (NN), are used to classify the different channel types. To further reduce the computational complexity, we leverage data extraction and data truncation. Our simulation results show that the proposed method can achieve a favorable performance in terms of classification accuracy (CA). For example, the KNN algorithm can achieve CA
$\geq 90\%$
when SNR is −10dB, and KNN and SVM algorithm can achieve CA
$\geq 90\%$
when SNR is −8dB. In addition, we also introduce the multi-frame gain combination method to further improve the classification performance in the low SNR regions.