Environment-Robust Device-Free Human Activity Recognition With Channel-State-Information Enhancement and One-Shot Learning

Deep Learning plays an increasingly important role in device-free WiFi Sensing for human activity recognition (HAR). Despite its strong potential, significant challenges exist and are associated with the fact that one may require a large amount of samples training, trained network cannot be easily adapted to new environment. To address these challenges, we develop novel scheme using matching enhanced channel state information (MatNet-eCSI) facilitate one-shot learning HAR. We propose CSI correlation feature extraction (CCFE) method improve condense activity-related input signals. It can also significantly reduce computational complexity by decreasing dimensions training strategy which effectively utilizes data set from previously seen environments (PSE). In least, realize only sample each testing environment PSE. Numerous experiments conducted results demonstrate our proposed outperforms state-of-the-art HAR methods, achieving higher accuracy less time.