Haptic rehabilitation exercises performance evaluation using automated inference systems

Stroke has been a major source of disability that strikes humans in different age, causing chronic disability in adults. A common type of this disability is hand stroke which deprives patients from controlling their hands and fingers. Rehabilitation of stroke is usually done by functional Occupational Therapy (OT) in which the patient is trained to perform everyday exercises. On the other hand, haptics and virtual environments research offer the opportunity to improve the traditional methods of therapy. During a certain rehabilitation exercise, at least one therapist has to subjectively monitor the patient's performance in order to assess his/her improvement. This paper aims to introduce an automated Inference System that utilizes haptic data to quantize the patient's performance. For this purpose, two systems were implemented: a Fuzzy Inference System (FIS) and an Adaptive Neuro-Fuzzy Inference System (ANFIS). The two systems were validated with sample input/output datasets. However, when testing them with real subjects' data, they gave irrational evaluations. Retracing the whole process has lead us to the conclusion that the CyberForce system is incapable of providing normative data for evaluating the patient performance due to calibration and consistency issues.