Research and implementation of sign language recognition method based on Kinect

Abstract

Sign language is a kind of important communication gesture to be studied in the human-computer interaction filed. Kinect is a 3D somatosensory camera launched by Microsoft, which can capture the color, depth and skeleton frames and is helpful to the gesture recognition research. In this paper, a method using the HOG and SVM algorithms with the Kinect software libraries to recognize sign language by recognizing the hand position, hand shape and hand action features is proposed. In order to realize this method, a special 3D sign language dataset contains 72 words is collected with Kinect, and experiments are conducted to evaluate the method. It is shown in the experimental results that the use of the HOG and SVM algorithms significantly increases the recognition accuracy of the Kinect, and is insensitive to background and other factors. The average recognition rate is up to 89.8%, which means the Kinect-based recognition method proposed in this paper can effectively and efficiently recognize sign language, and it has a great significance to the research and promotion of the sign language recognition technology.