RF-GRA: random forests-based RFID-driven gesture recognition algorithm for aphasia patients

Purpose With the growing demand for intelligence and precision in the medical industry, this paper aims to propose RF-GRA, a noncontact gesture recognition algorithm based on RFID technology, to address the expressive needs of aphasia patients in rehabilitation and communication. Design/methodology/approach The proposed RF-GRA system is a noncontact gesture recognition framework for aphasia patients: it uses RFID phase data for precise gesture tracking, adopts a hierarchical random forest for high-accuracy identification, optimizes real-time processing via coarse/fine-grained segmentation and combines signal processing (phase unwinding, smoothing) with machine learning to boost robustness. Its performance was tested by measuring recognition accuracy, response speed and anti-interference capability. Findings The RF-GRA system achieves 96.67% gesture recognition accuracy with 400ms average response time (meeting clinical real-time needs). It helps aphasia patients express needs and supports hand function/cognitive rehabilitation, maintains 93.50% accuracy under environmental interference and its noncontact design preserves patient privacy. Originality/value This work presents the first RFID-based gesture recognition system specifically designed for aphasia patients, offering superior accuracy and real-time performance compared to conventional vision-based methods. The algorithm significantly enhances communication efficiency while maintaining patient privacy, contributing to intelligent medical rehabilitation solutions.