Gesture Media: wave at an ESP32-S3 to play, pause and skip tracks

Gesture Media is DustinYe's media controller, with code on GitHub, built from two M5Stack parts: an M5Stack AtomS3R (ESP32-S3 with a 128x128 screen and a front button) and an M5Stack Unit Gesture with the PAJ7620U2 infrared sensor, joined by the Grove cable. It shows up as a standard BLE HID consumer-control device, so a computer, phone or tablet sees ordinary media keys, with no relay PC, cloud service or companion app. The gestures: open your palm and close it into a fist toward the sensor to play or pause, make a slap-style movement left or right for previous or next track, and draw circles clockwise or counter-clockwise for one volume step per turn. Pressing the screen pauses gesture control, a double-click switches to another host, and a 3-second hold clears the stored Bluetooth bonds. Under the hood the C++/Arduino firmware polls the sensor about every 4 ms, filters the trajectory, and runs separate sweep, grip and circle trackers before sending a key. The author is careful about limits: the PAJ7620U2 can't count fingers or truly tell a fist from an open palm, a sideways fist can still trigger a track change in the published v24 firmware, thresholds depend on how you mount it (sensor face vertical, start about 8 to 10 cm away), and no accuracy or latency figures are claimed. The repo pins M5Stack's ESP32 Arduino core 3.3.9 and the M5AtomS3R board, with PowerShell setup, build and upload scripts for Arduino CLI on Windows, a build guide for the IDE, and a Python replay of recorded sensor data for testing. It was checked on Windows, a Xiaomi phone and an iPad at different stages. It's a two-part, plug-together build: on the M5Stack store on October 20 the AtomS3R dev kit was $17.50 and the Unit Gesture $12.95. Note that the repo has no project-wide license yet.