Smart Wristband for Gesture Recognition

This paper aim to design a smart wristband for gesture recognition. Tendon movements around the wrist were measured by FSR sensors as input variables to classify different gestures. Polydimethylsiloxane material (PDMS) was applied to encapsulate FSR sensors, so that the wristband is flexible and suitable for people with different wrist sizes. Subsequently, the sensor data was transmitted to the computer via Bluetooth low energy (BLE) technology. MATLAB was used to train a classifier with ensemble subspace discrimination algorithm. After that, the received signal was processed by this trained classifier and made prediction. The accuracy is about 99.4%. Additionally, the paper explored how predict accuracy would be impacted when twisting the wrist. The result showed that a gesture in different angles was classified as different gestures. Overall, the wristband is rechargeable, portable and can accurately recognize over 6 gestures.