Alex Smith

Gesture tracking for Multi Factor Authenticationl

Developed a real-time system to track human gestures performed using a mobile device and verified them with minimal false-positive rates to increase the adaptability of multi-factor authentication. We employ various signal processing techniques to cross-correlate and fuse signals in IMU, inaudible acoustics and radio waves domain and also transmit data in inaudible frequencies. I particularly lead the algorithm design of these gesture recognition models. We employed both machine learning and template matching based techniques parallelly to improve the robustness of the system. There are many challenges you face when you try making a real-world product such as device variability, sensor drifts, processing delays. We managed to resolve these challenges and achieve high accuracy and low false-positives. You can find a blogpost by Silence Laboratories highlighting the perils of current TFA methods here.