Določanje položaja kožnih formacij pri taktilni komunikaciji v navideznih vmesnikih : doktorska disertacija

slovenian

doktorska dela

2003

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

People use gestures to communicate with each other and so are gestures as a method of user interaction with a machine more intuitive and offer more natural way of communication in comparison to existing primary input devices as mouse, keyboard and similar. Mentioned method is also more functional, because gesture plays a significant role in human communication, and so the recognition of gesture is central to developing user friendly environments in domain of virtual and augmented reality. The real-time processing plays an important role in achievement of more natural and physically intuitive way of user machine interaction. The main scope of this dissertation is a development of computer vision user interface and a certain user environment that would fulfill more natural user machine interaction in virtual or augmented reality applications and their usage within contemporary mobile telecommunication solutions. One of the main problems of current mobile devices is a restricted size of the showing area and the methods of data input as a manner of device navigation. To achieve the stated objective we have defined procedures for color filtering, motion analysis, gesture recognition and suggested a model of user environment within applications of virtual or augmented reality. As in such cases the skin color can be a very comprehensive feature to achieve the goal we have used color tracking where we have developed a new method for color and spatial filtering in purpose to indicate skin features and constitute skin formations within the level of the binary mask stream. A new structure of hardware accelerated digital nonlinear parametrical filter is shown that features real-time searching for skin colored regions in a two-dimensional image stream where the parametrical design of the filter also offers color detection at comprehensive sense. Parametrical scheme of the shown digital filter enables automatic simultaneous adaptation of parameters due to changes in scene luminance. Regarding a parametrical scheme of the color filter there is also a possibility of automatic adaptation of the color filter to the different types of skin. The next part of this dissertation aims to extract two-dimensional motion across names in binary mask sequence in later classification of underlying three-dimensional motion samples. We have developed a method for extraction and classification of two-dimensional motion within binary mask image sequence with use of a logical operator. An evaluation of the purposed method is given where we have discuss the problems underlying spatial covering between skin colored subjects and objects within the captured scene. To achieve more natural way of interaction within the suggested user environment of the virtual or augmented reality applications we have use stereoscopic computer vision that makes it possible to define positions in all three space extensions. A procedure for classification of underlying three-dimensional motion samples is presented where those samples are given as potential pairs of active areas and their appurtenant epipolar areas on the left and right dynamic binary masks. The classification of the potential pairs is based on stereoscopic disparity conditions of the parallel camera configuration and solving the stereo correspondence problem. Solving the stereo correspondence problem for active and appurtenant epipolar areas includes two decision criterions. As we have used time-sequencing video stream capturing we have shown the problems that are caused by motion parallax. To define space coordinates of the navigation point modeling of the finger tip on the user hand is included where the user hand is used as an avigation subject within the user interface. To control the influence of the scene background that occurs with motion of the image sensors we included a new method that uses accelerometers where influence parameters are modeled with scene complexity estimation. For the purpose of user gesture analysis and respectively defining expression states of the navigation subject in the captured scene we have suggested a time analysis of the navigational point spatial position. Distinction between gestures of navigation and confirmation upon the desired machine activity is recognized through time-stable motion and apparent short duration anomalies within the time depended spatial function of the navigation point. The contemporary mobile telecommunication technologies are addressed with a newly defined user environment that includes stereoscopic head-mounted displays, stereoscopic setup of the cameras and an additional accelerometer. Within the user environment we have suggested a user area that is seen by a user at the given time and is spatially aligned with the navigation area of the tactile interface. The user area can be moved within the user environment by tracking the orientation of the user's head. In the last chapter we have suggested a special purpose hardware design that gains all needed algorithms to achieve the stated goal in real time and features newly designed hardware accelerated color filtering.

07.08.2023