Machine Learning‐Assisted Soft Sensors: From Material Design to Motion Recognition

ABSTRACT The development of soft electronics provides reliable performance under deformation, excellent adaptability to irregular surfaces, and superior comfort for wear. Their deployments offer solutions to address the challenges of harsh signal acquisition conditions, poor wearing experience, insufficient continuous tracking, and low accuracy in traditional motion recognition. Recently, extensive research has been conducted on enhancing materials, structures, and multimodal sensor systems or networks for information extraction. However, motions typically produce large streams of data that require labor‐intensive processing steps to extract and interpret. The rising artificial intelligence has demonstrated high efficiency and accuracy in processing multi‐dimensional data. Using machine learning (ML) algorithms to analyze data from wearable sensors has shown significant progress, enabling smart and intelligent sensing applications. This review examines the working mechanisms, materials, and structures of soft sensors for motion recognition and discusses how their physical design shapes signal quality and algorithm selection. The applications of soft sensors aided by ML algorithms in various motion recognition tasks are reviewed, including smart healthcare monitoring, intelligent soft robotics, and human‐machine interaction. Finally, the review discusses the challenges and future development of ML‐assisted soft sensors, focusing on how long‐term reliability, system integration, material and structural design can be improved by ML strategies.