Gesture-Based Adaptive Haptic Guidance: A Comparison of Discriminative and Generative Modeling Approaches

Abstract

This paper investigates the incorporation of hidden conditional random fields (HCRF) as a discriminative statistical modeling technique into adaptive haptic guidance (HG) for physical human-robot interaction (pHRI). In this gesture-based HG approach, the knowledge and experience of experts are modeled to improve the unpredictable motions of novice trainees in a virtual minimally invasive surgery (MIS) training task. The HCRF models are developed for automatic gesture recognition and segmentation as well as generating guidance forces. The forces are adaptively calculated in real time with respect to gestural similarities among user motions and the gesture models. The HCRF-based approach is compared with a hidden Markov model-based (HMM-based) method for capturing the gestures of the user and providing adaptive HG. The experimental results show that the HCRF, as a discriminative method, can outperform HMM, as a generative method, in terms of user performance.