eye-tracking in video streaming Compressed by PE-VQ method in CNN and BUN

marzieh shabdiz¹ ali azar bar² hossein azgomi³ mohamadreza yamaghani⁴

1) Marzieh shabdiz1, Department of Computer Engineering , Azad University of
2) Ali Azarbar, Department of Computer Engineering,Azad University of
3) Hossein Azgomi, Department of Computer Engineering, Azad University of
4) Mohamadreza yamaghani, Department of Computer Engineering, Azad University of Lahijan,

Publication : 2nd. International Congress on science & Engineering - paris(parisconf.com)
Abstract :
ABSTRACT: This paper presents an eye-tracking study in video streaming and haw to learning angle of the rotate head and track motion of eyes for persons. In the near future automobiles might be able to behave responsively toward drivers’ eye gaze, eye movements or the pupil dilation.so, detecting the state of eyes for several persons in the movie need to a signed region of these, and compare them. Also, for tracking in video preserve of similar location and body for people is unesseccary. thus, we applied the encoded code for compression video, and sort array with special filters. Then the compact point of eyes with replacing of core level with the original image. then we used comprising pre-training several eye images for many people, for remove of redundancy of bits in connectivity weight in the hidden layer. The performance of a lossy compression algorithm is evaluated based on two conflicting parameters, namely, compression ratio and image quality which is usually measured by PSNR values. These algorithms use local features to better handle scale changes, rotation, and occlusion. In this paper, a new lossy compression method denoted as PE-VQ method is proposed which employs prediction error and vector quantization (VQ) concepts. An optimum codebook is generated by using a combination of The performance of the proposed PE-VQ method is evaluated in terms of compression ratio (CR) and PSNR values. It is also shown that higher PSNR values can be obtained by applying VQ on prediction errors rather than on the original image pixels. Finally, we create network binarization for building graph modeling of an image in the video for the detect edge of faces and node of eyes in several people and several time.
Keywords : Keywords: Image compression Artificial neural network Prediction errors Vector quantization convolutional neural network