期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2016
卷号：4
期号：1
页码：522
DOI：10.15680/IJIRCCE.2016.0401114
出版社：S&S Publications
摘要：People have emphasis on retrieval o f videos on internet with specifi c category and it is infeasible to find v ideo of interest. It becomes diffi cult to c lassify video with users demand due to l imited research in video classifi cation area. Furth er it affects the interest level of the users. There is need to have easier metho d for use rs to access video of interest. Due to the problems such as misdetection, increased computational loads on sy stem and poor video quality number of methods becomes unsuccessful , computationally expensive and hard to implement. The proposed system g ives the soluti on to the current problem using Recurrent Neural Net work. Recently all video classifi cation benchmarks p erformed for clip level prediction but the proposed system worked for clip level prediction to global video level prediction usi ng Recurr ent Neural Network. Diff erent patter ns are generated for each class for classifi cation. Hue, Saturation, Value color model is used to extract color features from each frame. Recurrent Multilayer Perceptron Neural Network is used to classify videos with the color features model and pattern generated for classes. Implementation results show that the proposed system increases the performance of the pro posed system by increasing peak signal - to - noise and reduces the Mean Absolute Error, Mea n Percentage Error and Relative Standard Error and hence perform better
关键词：Video Classification; Recurrent Neural Network; Recurrent Multilayer Perceptron; Heu-Saturation-Value (HSV) Color Model; Peak Signal to Noise Ratio (PSNR); Mean Absolute Error (MAE); Mean Percentage Error (MPE); Relative Standard Error(RSE)
