Intent-Aware Motion Analysis with Deep Learning: Methods, Applications, and Future Directions

Abstract

Recent advances in intelligent sensors coupled with next-generation interface technologies have made motion intention recognition a foundational technology for assistive systems—including prosthetic controllers, exoskeleton walkers, and rehabilitation devices. Focusing on deep learning applications, this review examines current developments in motion intention recognition across different countries, covering both signal analysis and behavioral pattern identification. By comparing the advantages and disadvantages of different signal types (such as electromyographic signals, muscle pressure signals, inertial sensor signals, etc.) and recognition algorithms (traditional machine learning and deep neural networks), the significant advantages of deep learning in improving recognition accuracy and adaptability are revealed. Additionally, by reviewing recent representative achievements, the practical deployment of motion intention recognition technology in lower-limb exoskeletons and prosthetic systems is discussed. Finally, the main challenges in current research, such as signal heterogeneity, individual user differences, and model generalization ability, are summarized, and future development directions based on multimodal fusion and transfer learning are envisioned.