Support vector machine algorithm-based wearable device in sports rehabilitation training for people with disabilities

The aim was to investigate the application effect of wearable devices based on the support vector machine (SVM) algorithm in sports rehabilitation training for people with disabilities. A total of 159 people with disabilities from Nanchang were assigned to either a control group (routine training, n = 82) or an observation group (routine training plus SVM algorithm-based wearable device, n = 77). WHODAS, WHOQOL-BREF, and activities of daily living (ADL) scales were employed to assess the functional status and quality of life (QoL) changes before and after intervention, and the training compliance and satisfaction were compared. The following parameters were compared, including the range of motion (ROM), gait, and trunk parameters between the two groups, along with the classification performance of three algorithms-standard support vector machine (SVM), particle swarm optimization SVM (PSO-SVM), and artificial bee colony optimization SVM (ABC-SVM)-in motion pattern recognition. No statistical distinctions in WHODAS, WHOQOL-BREF, and ADL scores were noted at baseline (P > 0.05). A substantial improvement was noted in all indicators post-intervention (P < 0.05). The observation group had greater score reductions in the cognition, mobility, self-care, and social participation dimensions of WHODAS, a more obvious increase in scores in the physical, psychological, independence, and environment dimensions of WHOQOL-BREF, and a larger increase in ADL scores (P < 0.05). Before the intervention, there were no statistically significant differences (P > 0.05) between the two groups in hip joint ROM, knee joint ROM, step length, stride width, or walking speed. After the intervention, both groups showed significant increases in hip and knee joint ROM (P < 0.05), with the observation group demonstrating a more pronounced increase (P < 0.05). Step length, stride width, and walking speed also showed significant improvement (P < 0.05), with the observation group showing greater improvement (P < 0.05). Before the intervention, there were no statistically significant differences (P > 0.05) between the two groups in C7 lateral deviation, trunk tilt angle, shoulder tilt angle, pelvic tilt angle, thoracic kyphosis angle, or lumbar lordosis angle. After the intervention, all these parameters showed significant improvement in both groups (P < 0.05), with the observation group showing more marked improvement (P < 0.05). Training compliance and satisfaction in the observation group were significantly higher than in the control group (P < 0.05). The ABC-SVM algorithm demonstrated the best classification performance, significantly outperforming both PSO-SVM and standard SVM (P < 0.05). Wearable devices based on the SVM algorithm can more effectively improve the functional status, QoL, and ADL of people with disabilities. The SVM model combined with intelligent optimization algorithms can improve the accuracy of motion recognition and enhance the intelligence level of the rehabilitation system.