Multispectral hand recognition using the Kinect v2 sensor

Multispectral data from inexpensive, yet accurate, sensors has become readily available within the last several years and opened many possibilities for contactless biometrics applications. The Kinect v2 provides depth, RGB, and Near-Infrared (NIR) data and can be used for recognition of individuals using extracted hand regions in all three spectra. Initially, the depth data is used to extract the hand region for use as a mask to extract the hand region in the depth, RGB, and Near-Infrared (NIR) spectra. These extracted regions then have Principal Component Analysis (PCA) applied to them before passing through classification. K-Nearest-Neighbors (KNN) and Support Vector Machines (SVM) are compared for classification. In testing it was found that on average the RGB and NIR data provided a recognition rate of approximately 75%-80% for either KNN or SVM classification and at different amounts of principal components for PCA.