Aiming at the rehabilitation needs of patients with wrist dysfunction, this study proposes a digital wrist rehabilitation device based on multi-sensor fusion technology. A multimodal data acquisition and fusion framework was constructed by combining surface electromyographic signals and joint motion angle information. The improved WVPSO algorithm is used to optimize the LSSVM hyperparameters to achieve high-precision classification and recognition of finger-wrist movement intentions. Based on ADAMS simulation platform, a virtual prototype model of the device is established to verify its kinematic consistency and human-computer interaction adaptability. The clinical effect of the device was evaluated through a randomized controlled trial of 50 patients with chronic wrist dysfunction. After treatment, the pain, grip strength, function, and dorsal extension/palmar flexion mobility scores of the two groups improved compared with those before treatment, and the mean values of the experimental group were higher than those of the control group by 3.09, 2.41, 4.01, and 3.84 points, respectively. Physical, somatic, and emotional-social scores of the two groups improved compared with the pre-treatment, and the mean values of the experimental group were higher than those of the control group by 8.71, 9.01, 9.83, and 9.32 points, respectively, and the differences were statistically significant (P<0.05).
