Enhanced Deep Transfer Learning Model based on Spatial-Temporal driven Scalograms for Precise Decoding of Motor Intent in Stroke Survivors

Motor function loss greatly impacts post-stroke survivors while performing activities of daily living. In the recent years, intelligent rehabilitation robotics have been proposed to enable the patients recover their lost limb functions. Besides, a large proportion of these robots function in passive mode that only allow users to navigate trajectories that rarely align with their limb movement intent, thus precluding full functional recovery. A potential solution would be to explore utilizing an efficient Transfer Learning based Convolutional Neural Network (TL-CNN) to decode multiple classes of post-stroke patients’ motion intentions towards realizing dexterously active robotic training during rehabilitation. In this regard, we propose and examined for the first time, the use of Spatial-Temporal Descriptor based Continuous Wavelet Transform (STD-CWT) as input to TL-CNN to optimally decode limb movement intent patterns of stroke patients to provide adequate input for active motor training in rehabilitation robots. Importantly, we examined the proposed (STD-CWT) method on three distinct wavelets including the Morse, Amor, and Bump, and compared their decoding outcomes with those of the commonly adopted CWT technique under similar experimental conditions. Our method was validated using electromyogram signals of five stroke survivors who performed up to twenty-two distinct limb motions. The obtained results showed that the proposed technique recorded a significantly higher decoding (p<0.05) and converges faster compared to the commonly adopted method. The proposed method equally recorded obvious class separability for individual movement classes across the stroke patients. Findings from this study suggest that the STD-CWT Scalograms would provide potential inputs for robust decoding of motor intent that may facilitate intuitively active motor training in stroke rehabilitation robots.

© 20XX IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.