Digital Studio

GESTURE RECOGNITION

As we know, the vision-based technology of hand gesture recognition is an important part of human-computer interaction (HCI). In the last decades, keyboard and mouse play a significant role in human-computer interaction. However, owing to the rapid development of hardware and software, new types of HCI methods have been required. In particular, technologies such as speech recognition and gesture recognition receive great attention in the field of HCI.

Gesture is a symbol of physical behavior or emotional expression. It includes body gesture and hand gesture. It falls into two categories: static gesture and dynamic gesture. For the former, the posture of the body or the gesture of the hand denotes a sign. For the latter, the movement of the body or the hand conveys some messages. Gesture can be used as a tool of communication between computer and human. It is greatly different from the traditional hardware based methods and can accomplish human-computer interaction through gesture recognition. Gesture recognition determines the user intent through the recognition of the gesture or movement of the body or body parts. In the past decades, many researchers have strived to improve the hand gesture recognition technology. Hand gesture recognition has great value in many applications such as sign language recognition, augmented reality (virtual reality), sign language interpreters for the disabled, and robot control.

In, the authors detect the hand region from input images and then track and analyze the moving path to recognize America sign language. In, Shimada et al. propose a TV control interface using hand gesture recognition. Keskin et al. divide the hand into 21 different regions and train a SVM classifier to model the joint distribution of these regions for various hand gestures so as to classify the gestures. Zeng et al. improve the medical service through the hand gesture recognition. The HCI recognition system of the intelligent wheelchair includes five hand gestures and three compound states. Their system performs reliably in the environment of indoor and outdoor and in the condition of lighting change.

The work flow of hand gesture recognition is described as follows. First, the hand region is detected from the original images from the input devices. Then, some kinds of features are extracted to describe hand gestures. Last, the recognition of hand gestures is accomplished by measuring the similarity of the feature data. The input devices providing the original image information includes normal camera, stereo camera, and ToF (time of flight) camera. The stereo camera and ToF camera additionally provide the depth information so it is easy to segment the hand region from the background in terms of the depth map. For the normal camera, the skin color sensitive to the lighting condition and feature points are combined to robustly detect and segment the hand region. When the region of interest (ROI, the hand region in the case) is detected, features are needed to be extracted from the ROI region. Color, brightness, and gradient values are widely used features. Li and Kitani describe various features for hand region detecting including the Gabor filter response, HOG, SIFT, BRIEF, and ORB. For the recognition of hand gestures, various classifiers, for example, SVM (support vector machine), HMM (hidden Markov model), CRF (conditional random field), and adapted boosting classifier are trained to discriminate hand gestures. Although the recognition performance of these sophisticated classifiers is good, the time cost is very high.

In this paper, we present an efficient and effective method for hand gesture recognition. The hand region is detected through the background subtraction method. Then, the palm and fingers are split so as to recognize the fingers. After the fingers are recognized, the hand gesture can be classified through a simple rule classifier.

The novelty of the proposed method is listed as follows.

• The first novelty of the proposed method is that the hand gesture recognition is based on the result of finger recognition. Therefore, the recognition is accomplished by a simple and efficient rule classifier instead of the sophisticated but complicated classifiers such as SVM and CRF.
• Some previous works need the users to wear data glove to acquire hand gesture data. However, the special sensors of data glove are expensive and hinder its wide application in real life. In the work, the authors use TOF camera, that is, Kinect sensor, to capture the depth of the environment and a special tape worn across the wrist to detect hand region. Our approach only uses a normal camera to capture the vision information of the hand gesture meanwhile does not need the help of the special tape to detect hand regions.
• The third advantage of the proposed method is that it is highly efficient and fit for real-time applications.

The rest of the paper is organized as follows. In Section 2, the proposed method for hand gesture recognition is described in detail. In Section 3, the performance of our approach is evaluated on a data set of hand gestures. Then, our method is compared with a state-of-art method (FEMD) on another data set of hand gestures. Section 4 presents the conclusion and future works.