Hybrid feature selection for gesture recognition using support vector machines

This paper presents an approach for a multi-cue based two-dimensional gesture recognition that combines two different forms of cues, namely shape cues and motion cues, in a support vector machine (SVM) framework. Each gesture is comprised of trajectories in an image sequence and thus can be represented as a time series in a Principal Component Analysis (PCA) reduced dimensional space. A novel multicue feature extraction technique is proposed that performs well under large variations of hand shapes and movements. A class of SVMs applicable to sequential-pattern recognition is employed by incorporating a hybrid distance measure into the kernel function. The hybrid distance measure accounts for both hand shape and movement. The performance of the proposed method is evaluated utilizing hand gestures, captured from continuous tactile gesture streams, in recognition experiments. The proposed multi-cue approach is implemented and tested utilizing 9 different gestures performed by 25 subjects...

• Type: Conference
• Year: 2008
• Where: ICASSP
• Authors: Yu Yuan, Kenneth Barner
• Keywords: Gesture, Hybrid Distance Measure, ICASSP 2008, Signal Processing, Two-dimensional Gesture Recognition