AI-Enhanced Gait Re-Education: Overcoming Linguistic and Cultural Barriers in Rehabilitation

Abstract

The success of motor rehabilitation is closely related to the quality of communication between the therapist and the patient. Cognitive impairment is common among acute stroke survivors and is frequently associated with slowed information processing. Under such conditions, verbal instructions often fail to be conveyed in a form that the patient can effectively translate into action. This paper presents two complementary electronic systems developed to provide language-independent feedback for gait re-education. The first is a low-cost device built on an Arduino MKR1000 platform that combines a triaxial accelerometer and three FSR pressure sensors to provide real-time visual feedback during walking, with no AI on board. The second one uses an STM32 microcontroller with six synchronised accelerometers to record multi-segment gait data for offline processing using Gait Flow Analyser, which provides graphical summaries, descriptive statistics, and an optional AI module that detects deviations from a learned normal gait baseline. Preliminary clinical use with 12 patients over six months suggested that visual cues can prompt movement correction without verbal direction. The STM32 system has additionally been evaluated in a parallel investigation on a cohort of 30 participants (15 neurological participants and 15 healthy controls), thereby providing broader empirical context for the descriptive findings reported here. The two systems are not alternatives. Real-time feedback supports the therapeutic session, whereas offline AI-assisted analysis provides the therapist with an objective assessment of gait asymmetry and its evolution over time. Together, they establish a rehabilitation framework in which therapeutic guidance relies less on verbal communication and more on directly interpretable visual feedback.