Exploring LoRa and Deep Learning-Based Wireless Activity Recognition

Today’s wireless activity recognition research still needs to be practical, mainly due the limited sensing range and weak through-wall effect of current based on Wi-Fi, RFID (Radio Frequency Identification, RFID), etc. Although some recent has demonstrated that LoRa can used for long-range wide-range sensing, no pertinent studies have been conducted LoRa-based recognition. This paper proposes applying communication technology contactless We propose deep learning indoor first time a more lightweight improved TPN (Transformation Prediction Network, TPN) backbone network. At same time, using only two features signal amplitude phase as input model, experimental results demonstrate is better than original directly. The accuracy reaches 97%, which also recognition, meet engineering applications.