Knitted Sensors: Gesture Recognition

Developments in touch-sensitive textiles have enabled novel interactive techniques and applications. Fabric sensors have been produced in a variety of ways, bringing together advancements from design, manufacturing, engineering, and, more recently, machine learning. The sensors introduced in this work can be manufactured at scale using digital weft knitting, with little human intervention in the textile production process. They use capacitive touch sensing, and are constructed as one planar layer of knitted fabric. The textiles’ sensitive areas are created from a single conductive yarn and require a limited number of connections to external hardware, which increases their robustness and usability. However, these minimalistically-designed sensors shift the complexity of enabling interactivity from the hardware to computational models, since output from the system is reduced. Prior work on similar sensors relying on the same design principles and process has explored touch location identification. This work advances the capabilities of such sensors by creating the foundation for an interactive gesture recognition system. It uses a novel sensor design and relies on combining neural network architectures to accurately classify relatively complex gestures performed on it, unfolding many possibilities for future applications.




Denisa Qori McDonald, Richard James Vallett, Lev Saunders, Genevieve Dion, and Ali Shokoufandeh. Recognizing Complex Gestures on Minimalistic Knitted Sensors: Foundations for a Real–World Interactive System. Unpublished Manuscript. Available upon request.