Personalized Arm Gesture Recognition Using the HMM-Based Signature Verification Engine

摘要

Gesture-based interfaces can significantly improve access to computer technologies for people with disabilities. However, the gestures of the disabled can be very unstable, which means that standard gesture recognition solutions may occur not applicable in this case. In this study, a personalized arm gesture recognition system is presented. It uses automatic handwritten signature verification (ASV) techniques which deal with the instability of human motion processes by default. The proposed gesture-based HCI framework involves the Microsoft Kinect sensor, the data acquisition module, and the formerly tested HMM offline signature verification engine. The evaluation process included intentional, non-intentional and randomly distorted gestures converted into images. The confusion matrices and receiver-operation characteristic (ROC) analysis were used to evaluate the accuracy of the system. The performed tests showed that the applied ASV software could effectively recognize unstable gestures, even when the evaluation set included both random and distorted gesture patterns.