Wireless Vibrotactile Trainer for Balance Rehabilitation

Abstract

During balance rehabilitation, physical therapists typically provide verbal instruction and/or physically reposition a patient to demonstrate proper postural position and movements. We have developed a wireless device that enables an expert (such as a physical therapist) to map his/her movements to a trainee in a hands-free fashion. The trainee is subsequently able to mimic the motion of the expert by interpreting positional cues presented via vibrotactile feedback to the relevant body segments. This device will potentially enable a therapist to aid multiple patients simultaneously and/or remotely, or enable a trainee (such as an athlete or student) to replicate expert movements. The device comprises an Expert Module (EM) and Trainee Module (TM). Both the EM and TM are composed of six degree-of-freedom inertial measurement units, microcontrollers, and batteries. The TM also has an array of vibrating actuators that provides the user with vibrotactile biofeedback. The expert dons the EM, and his/her relevant body position is computed by an algorithm based on an extended Kalman filter that provides asymptotic state estimation. The captured body position information is transmitted wirelessly to the trainee, and directional instructions regarding the desired motion/position are displayed via vibrotactile feedback. The trainee is instructed to move in the direction of the vibration sensation until the vibration is eliminated. While prior work has demonstrated the use of vibrotactile stimulation for improved motor learning, this portable and wireless device is suitable for use outside of a laboratory environment. Five healthy young blindfolded subjects were instructed to mimic recorded expert anterior-posterior trunk tilt motion using the aforementioned device in a series of proof-of-concept studies designed to investigate the effects of changing the feedback activation threshold and varying the nature of the feedback. To characterize the efficacy of the system, we performed a cross correlation of expert and trainee trunk tilt angle while varying the threshold angle difference at which vibrotactile feedback was applied. Preliminary results showed that subjects performed best at 0.5 and 0.75 degree thresholds among those tested (0.5, 0.75, 1.0, 1.25, 1.5). The normalized mean cross correlations for the 0.5 and 0.75 threshold conditions were 0.96 and 0.97 respectively, while the mean differences between expert and trainee trunk tilt angles were 1.1 and 1.2 degrees respectively. Further studies at 0.5 and 0.75 threshold conditions confirmed that proportional plus derivative feedback of the angle difference resulted in superior performance compared to proportional or derivative feedback alone. Repetition of the task was not significant suggesting that trainees could immediately use the device to accurately replicate expert anterior-posterior trunk tilt movements.