Preliminary Investigation of Fine-Grained Gesture Recognition With Signal Super-Resolution
Naoya Yoshimura, Takuya Maekawa, Daichi Amagata, Takahiro Hara · 2018
This study investigates the feasibility of fine-grained gesture recognition using upsampled acceleration sensor data. Because the maximum sampling rate of smartwatch devices is limited by operating systems, we simulate high resolution acceleration data using a neural network from low resolution signals in order to capture distinguishing features of gestures containing high frequency components.