jault / handwritingsmartwatch Goto Github PK
View Code? Open in Web Editor NEWThis project improves on building a system in which hand-written letters can be recognized with adequate accuracy from solely the sensors available on commercial smartwatches. Prior works have found varying success at the tasks ranging from 90% in a lab setting to an abysmal 17% in an unrestrained environment. Combining stroke-based features with gyroscope and accelerometer gathered data gives a more robust feature set for recognition of handwritten characters doubling performance. Heavily based on Raveen Wijewickrama, Anindya Maiti, and Murtuza Jadliwala. 2019. deWristified: handwriting inference using wrist-based motion sensors revisited. In Proceedings of the 12th Conference on Security and Privacy in Wireless and Mobile Networks.