Gesture recognition using Wi-Fi sensing on Commercial Off-The-Shelf (COTS) hardware

This dataset is related to a DNN-based Gesture Recognition algorithm. The data is generated in a scenario comprising a single laptop device equipped with a WiFi-7 card, and 5 users. During the data generation the users in turn do 5 different gestures multiple times. The dataset is generated from 55 different sessions, 50 of those corresponding to a lab environment and 5 to a café/public space, yielding a total of 722 gesture instances. The café subset provides an out-of-environment evaluation for cross location generalisation. The gestures include: forward/backward hand wave, up/down hand move, pulse gesture (push forward and back), clockwise circular motion and side-to-side wave. From these scenarios the raw CSI is recorded and time stamped.