Multimedia classification using ANN approach

Abstract

Digital multimedia data in the form of speech, text and fax is being used extensively. Segregation of such multimedia data is required in various applications. While communication of such multimedia data, the speech, text and fax data are encoded with CVSD coding, Murray code and Huffman code respectively. The analysis and classification of such encoded multimedia from unorganized and unstructured data is an important problem for informationmanagement and retrieval. In this paper we proposed an ANN based approach to classify text, speech and fax data. The normalized frequency of binary features of varying length and PCA criterion is considered to select effective features. We use selected features in Back-propagation learning of MLP network for multimedia data classification. The proposed method classifies data efficiently with good accuracy. The classification score achieved for encoded plain data is of the order of 91, 93 and 90% for speech, text and fax respectively. Also for 30% distorted data, the classification score obtained is of the order of 78, 80 and 72% for speech, text and fax respectively.