Advancing flexible sensing systems: Deep learning-driven insights and future perspectives

Advancing flexible sensing technology is a key enabler for the development of a highly interconnected and intelligent society. In recent years, significant progress has been achieved in the performance of flexible sensors as well as in the efficiency of data processing algorithms. To address the rising requirements of emerging intelligent systems, extensive efforts have focused on improving sensing materials, underlying mechanisms, and device architectures. Simultaneously, advanced data analysis methods have been applied to obtain meaningful information from the intricate, high-dimensional data generated by flexible sensing platforms. As a core branch of artificial intelligence, deep learning (DL) has emerged as a powerful tool for flexible sensing applications, owing to its ability to model nonlinear relationships and handle heterogeneous data, and effectively compensate for inherent material non-idealities such as mechanical hysteresis, signal drift, and environmental interference. This review provides an overview of the fundamental principles of flexible mechanical sensors and representative deep neural network architectures, including emerging paradigms such as Transformers and self-supervised learning, and systematically summarizes recent advances in DL-assisted data analysis for flexible sensing. Key application areas, including gesture recognition, motion detection, surface topology identification, voice interaction, health monitoring, and pressure sensing, are highlighted. In conclusion, key challenges and future opportunities are evaluated, with a specific focus on addressing data imbalance, developing physics-informed algorithms, and implementing on-device privacy desensitization, to guide the progress of emerging intelligent sensing technologies.