Hand Gesture Recognition


Aim of this work is to use depth data for hand gesture recognition. Currently our work is based on the Kinect depth sensor data. We explore 3D hand region features for recognition of hand gestures that includes motion in 3 dimensions

One of the main challenges of depth-based hand gesture recognition is the decremnt performance when the distance of the location of the hand to the sensor becomes large. We have developed a distance-invariant depth-based 3D hand gesture-recognition scheme. The proposed method involves normalisation of the hand region, 3D projection of depth data, morphological filtering and adaptive interpolation-based pre-processing. More about this can be found in Our ICIP 2013 paper. The presentation can be found here.


M. Assad and C. Abhayaratne, Kinect Depth Stream Pre-Processing for Hand Gesture Recognition, in Proc. ICIP 2013.