Real-time gesture recognition based on motion quality analysis

This paper presents a robust and anticipative realtime gesture recognition and its motion quality analysis module. By utilizing a motion capture device, the system recognizes gestures performed by a human, where the recognition process is based on skeleton analysis and motion features computation. Gestures are collected from a single person. Skeleton joints are used to compute features which are stored in a reference database, and Principal Component Analysis (PCA) is computed to select the most important features, useful in discriminating gestures. During real-time recognition, using distance measures, real-time selected features are compared to the reference database to find the most similar gesture. Our evaluation results show that: i) recognition delay is similar to human recognition delay, ii) our module can recognize several gestures performed by different people and is morphology-independent, and iii) recognition rate is high: all gestures are recognized during gesture stroke. Results also show performance limits.