Research Of The Classification Algorithm Based On Nonparametric Bayesian

With the rapid development of network technology and computer technology, the data volume of video and image information is exploding. Finding the videos and images which users needed from mass and complex data is becoming a difficult problem. For this purpose, a nonparametric Bayesian method based on semantic information classification has been proposed by the scientists.In general, the semantic information extraction from video and image is all based on the content in video and image. Images classification needs to extract the semantic features. The main work in this paper is listed as follow:Firstly, a novel algorithm based on Multi-Variable LDA has been proposed. With two different and independent type of semantic information, the algorithm can dig them at the same time. The algorithm solved the problem that existing method just can dig semantic information from one variable. The experimental re-sult shows that the method digs the semantic information and classifies the words efficiently. The method is better than the original LDA algorithm in recognition.Secondly, an algorithm based on Bayesian Mixture model has been proposed to solve the problem that some images cannot extract the hidden content of se-mantic features. The algorithm extracts the locations of discriminant features with Bayesian Mixture model based on inverse regression, then reduces the dimensions of images data, and classifies the processed images data into different topics by DLA. The experimental result shows that the algorithm can efficiently use the locations which have more discriminant information for the classification to enhance the ac-curacy.With the experimental results, the methods can extract the semantic features from different types of data to improve the classification effect. With images data, the methods can locate the discriminant features to improve the extraction effect, so that the classification efficient can be improved.