An Embedded AI System for Predicting and Correcting the Sensor-Orientation of an Electronic Travel Aid During Use by a Visually Impaired Person

People with visual impairment frequently rely on Electronic Travel Aids for navigation, with sensor-equipped guide canes being a prevalent choice. These ETAs alert users of nearby obstacles using directive sensing technology, effectively covering the navigation corridor of the person motion, thus facilitating obstacle avoidance. However, the efficacy of these devices depends upon the user’s ability to maintain them in the correct orientation, a requirement that may not always be met due to various factors, including lack of training and excess cognitive load on the user in mobility, resulting in injuries. This paper proposes the development of an advanced AI-based embedded system, designed to be integrated into a traditional white guide cane. This proposed system predicts the instantaneous orientation of the ETA from raw proprioceptive measurement of ETA’s instantaneous velocity and acceleration. If the estimated angle lies beyond a nominal range, audio or vibration feedback is provided proportional to the error.