The Video Streams Prediction Based on Adaptive Kalman Model

In order to improve the rate of bandwidth utilization and achieve dynamic bandwidth allocation, in this paper, the video stream predication model of Kalman filter is improved according to linear prediction of future frames. The Kalman filtering owns a minimum mean square error estimate when the observed variables and noise are jointly Gaussian noise. Noise reduction is applied, and then the technology of scene change is added. The method proposed strengthened the correlation of data, and improved the prediction accuracy. Finally, packet loss was predicted according to network conditions. Experimental results show that the predictive efficiency have been greatly improved through the improvement of Kalman filter for video stream model.