Flexible touch and gesture recognition system for curved surfaces with machine learning for assistive applications

Touch is a fundamental mode of human-machine interaction and ability to monitor tactile pressure, recognize gestures and location of touch are crucial for touch-based technologies. However, achieving reliable touch sensing on curved surfaces remains challenging as flexing often disrupts the stability of sensor outputs and diminishes sensitivity, especially in dynamic environments. This study presents the development of a flexible multi-element touch sensing patch that can monitor its bending state as well as detect pressure with a sensitivity of 0.827 kPa(-1). The patch is fabricated using resistive strain sensors, screen printed onto a PET sheet with a foam backing. Evaluation electronics were integrated to ensure stable, noise-free signal acquisition, and output was processed with machine learning (ML) algorithms to classify gestures such as single and double finger taps, swipes, and touch locations, with 93 % accuracy, on both flat and curved surfaces. Based on the identified gesture, the system enables users to type text or control external devices with minimal physical effort. Its scalable fabrication, high sensitivity, mechanical resilience and seamless ML integration establishes it as a powerful and efficient tool for assistive technologies, designed to support individuals with limited speech and mobility, such as those with quadriplegia or paralysis.