Abstract
In the thesis, a fully automatic computerized system was proposed to monitor and track the close eyeball motion based on the designed image capture module and its peripheral circuits followed by the application of image processing method. The system is composed of three main parts- CMOS image acquisition module, wireless communication module which operates under 2.4~2.524GHz and image processing unit on PC host. The CMOS image sensor with 120*160 pixels square resolution continuously writes an sensed image onto a 32K*8bits SRAM. A MCU in the transmitting end acts as an intermediate between the SRAM and the wireless transmitting module with the data rate up to 1Mbps. Due to the capacity of sampled image and protocol, the baud rate of transmission module is suggested to be higher than 800 kbps. The MCU in the receiving end clocks out the incoming packages from the wireless receiving module and sends these data to the host PC terminal by means of the RS232 channel. Finally, the developed algorithm on PC end is accomplished to judge close eyeball motion. In the PC terminal, the noise reduction method is first applied to an incoming image, the two dimensional Gaussian low pass filter is adopted herein, then, extracting some types of features from the defined Region of Interest (ROI) which is located after some preprocessing procedures. Finally, the optical flux analysis with support vector machine neural network training is performed. The system has the ability to judge whether a captured video is in a moving or in a freezing status within a time interval. A video is composed of twenty frames. The correct rates of the outputs of neural network which was trained by the various combinations of the features are compared as well.