Software Defined Radio-based Human Activity Recognition: A Comprehensive Review

The Human Activity Recognition (HAR) systems have been recognized as one of central important components in numerous inevitable everyday-life applications, e.g. healthcare, security, search and rescue. In order to overcome the wearable sensor burden, RF-based HAR has become a promising technology candidate for many applications to save on privacy concern. The fundamental concept of this systems is achieved by the fact that human movement will affect the RF propagation path and characteristics, which will lead to reflected signals with distinctive fingerprint for different activities. In this research, a comprehensive review of RF-based HAR with the utilization of Software Defined (SDR) technology is presented for practical implementation by transmitting the RF signals towards the human and receiving signals from various scenarios through the utilization of several Universal Software Radio Peripheral (USRP) platforms. This research gives details about the types of HAR classifiers, human activities, type of the utilized waveforms and systems architectures. The High classification accuracies for different tasks have been achieved with the use of various deep learning, machine learning and model-based techniques. The efficacy of SDR technology has been demonstrated in real-time practical HAR systems.