This paper describes an intelligent system ABHIVYAKTI, which would be pervasive in nature and based on the Computer Vision. It would be very easy in use and deployment. Elder and sick people who are not able to talk or walk, they are dependent on other human beings and need continuous monitoring, while our system provides flexibility to the sick or elder person to announce his or her need to their caretaker by just showing a particular gesture with the developed system, if the caretaker is not nearby. This system will use fingertip detection techniques for acquiring gesture and Artificial Neural Networks (ANNs) will be used for gesture recognition.
ABHIVYA KTI: it's a word of Hindi Language, used for expressing feelings orally. Old or sick people, who are not able to express their feelings by words or cannot walk, will be the users of this system. During the monitoring of patients, it is possible that nobody is in the room and patients want to eat something or want to call someone, then they can use their hand to make gesture with the developed system and can get what they want. It is assumed that the users are able to move their hand, they are not fully paraly zed. Here in the system, focus of research is on human to machine interaction, in which machine would be able to do action according to the predefined syntax of the gesture made by user. ABHIVYA KTI will have an interface which would include a small s mart camera, where users have to show their hand in front of camera. This hand gesture would be interpreted by the system, whether it is valid gesture syntax or not. If the gesture is not included in the rules list, system will not take any action and will give a message of wrong input. If it is valid then the system will work according to predefined action and user will be informed as action done.
ABHIVYA KTI will work on the principles of Co mputer Vision, where it will use image processing for acquiring gesture and preprocessing. In the proposed system research focus is on 2D systems only because 3D systems are much complex to setup and need at least 2 stereo cameras working synchronously. Also the complexity of modeling of images parameter co mputation is very high [30], so they are not often used for hand gesture recognition. Also in 3D, it is difficult to measure whether a hand movement was gesture or unintentional [15]. II. RELATED WORK Mitra [19] defines gesture recognition a process where user made gesture and receiver recognized them. Many Researchers have done excellent work in this area. Ahn [1] have developed augmented interface table using infrared cameras for pervasive environment. Chaudhary [4] has described designing for intelligent systems in his work. A contour following algorith m has been shown in the figure 1, which extract foreground image fro m the captured image and further detects hand and face on skin based detection. In gesture recognition color based methods are applicable because of their characteristic color footprint of human skin. The color footprint is usually more distinctive and less sensitive than in the standard (camera capture) RGB color space. Most of color segmentation techniques rely on histogram matching or emp loy a simple look-up table approach [13][24] based on the training data for the skin and possibly it’s surrounding areas. The major drawback of color based localization techniques is the variability of the skin color footprint in different lighting conditions. This frequently results in undetected skin regions or falsely detected non skin textures. The problem can be somewhat alleviated by considering only the regions of a certain size or at certain spatial position.
Another common solution to the problem is the use of restricted backgrounds and clothing like dark gloves or wearing a strip on wrist [16][17][18] [20]. Wu [31] has been extracted hand region fro m the scene using segmentation techniques. Vezhnevets [28] describes many useful methods for skin modeling and detection. Skin color detection and boundary extraction are two important part of gesture extraction form the image. Gesture recognition is the phase in which the data analyzed fro m the visual images of gestures is recognized as a specific gesture. In this step we assume that gesture image has been extracted fro m the image captured and now it is target data for gesture recognition. Graph matching is widely used for object mapping in images, but it faces problems in dependency on segmentations [23]. Identification of a hand gesture can be done in many ways depending on the problem to be solved [20].
The interpretations of gestures require that dynamic or static configurations of human hand be measurable by the machine. First attempts to solve this problem resulted in mechanical devices called as glove-based devices that directly measure hand joint angles and spatial position [3][10] [29]. In this system requires the user to wear a cumbersome device and carry a load of cables that connect the device to a computer. This hinders the ease and naturalness with which user can interact with computer controlled environment. Potentially, any awkwardness in using gloves and other devices can be overcome by videobased noncontact interaction techniques identifying gestures. Most of the static models are meant to accurately describe the visual appearance of the gesturer’s hand as they appear to a human observer. To perform recognition of those gestures, some type of parameter clustering technique stemming fro m vector quantization (VQ) is usually used. Briefly, in vector quantization, an n-dimensional space is partitioned into convex sets usin
This content is AI-processed based on open access ArXiv data.
