Machine Learning-Based Channel Classification and Its Application to IEEE 802.11ad Communications

We study the application of machine learning to channel classification for identifying whether a channel belongs to the Line of Sight (LOS) or Non-Line of Sight (NLOS) classes. The machine learning approach is able to work on multiple features, resulting in a much more accurate pattern identification and classification performance. We show that even in the absence of channel estimation, it is possible to classify the channel using the received preamble sequence with machine learning. This allows quicker classification and it is robust to channel estimation error, which is favorable in the low Signal to Noise Ratio (SNR) regime. The scheme is evaluated for IEEE 802.11ad systems, but the concept is also applicable to other wireless systems in general.

[1]  Gaetano Giunta,et al.  Dynamic LOS/NLOS Statistical Discrimination of Wireless Mobile Channels , 2007, 2007 IEEE 65th Vehicular Technology Conference - VTC2007-Spring.

[2]  W. Wirnitzer,et al.  60 GHz in-cabin real-time channel sounding , 2009, 2009 Fourth International Conference on Communications and Networking in China.

[3]  Ismail Güvenç,et al.  NLOS Identification and Mitigation for UWB Localization Systems , 2007, 2007 IEEE Wireless Communications and Networking Conference.

[4]  Klaus Wehrle,et al.  Machine learning-based jamming detection for IEEE 802.11: Design and experimental evaluation , 2014, Proceeding of IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks 2014.

[5]  Jingon Joung,et al.  Machine Learning-Based Antenna Selection in Wireless Communications , 2016, IEEE Communications Letters.

[6]  A. Maltsev,et al.  Statistical channel model for 60 GHz WLAN systems in conference room environment , 2010, Proceedings of the Fourth European Conference on Antennas and Propagation.

[7]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.