MyoSpring: 3D Printing Mechanomyographic Sensors for Subtle Finger Gesture Recognition

Abstract

We introduce MyoSpring, a novel technique to fabricate customised mechanomyographic sensors with 3D printed springs. MyoSpring averts from a traditional gone-size-fits-all' approach and facilitates the development of functionally customisable wrist-worn sensors, which can be used to recognise subtle finger gestures with high accuracy. We automate the design process of MyoSpring sensors with an intuitive Graphical User Interface enabling novices to quickly design and fabricate customised sensors. We conducted a user study with 15 participants and 12 gestures, including partial and full flexion finger gestures, to verify the functionality of MyoSpring. Through the evaluation, we show that 1) MyoSpring can achieve high average accuracy of finger gesture sensing of 94.11% (SD = 9.73%), 2) mechanical customisation allows MyoSpring to significantly improve the average accuracy of overall gesture sensing and 3) MyoSpring better supports partial finger flexion gesture sensing, achieving an accuracy of 91.44% (SD = 11.76%) for partial gestures.