Using gesture recognition to control PowerPoint using the Microsoft Kinect

Abstract

This thesis describes the design and implementation of a speech and gesture recognition system used to control a PowerPoint presentation using the Microsoft Kinect. This system focuses on the identification of natural gestures that occur during a PowerPoint presentation, making the user experience as fluid as possible. The system uses an HMM to classify the performed gestures in conjunction with an SVM to perform real-time segmentation of gestures. The fusion of these two models allows the system to classify gestures in real time as they are being performed instead of waiting until completion. The incorporation of speech commands gives the user an additional level of precision and control over the system. This system can navigate through a PowerPoint presentation and has a limited control over slide animations.