AbstractsComputer Science

This research work presents artificial neural network (ANN) algorithm approach for retrieving a video, based on a query. The queries are a combination of plain text, sound file (wave file), and an image. Based on the combinations of the query, either an existing template file will be used or newlinesubsequent processing of the queries will be done to obtain features. These features are used to train ANN algorithms to obtain a set of final weights. When the plain text is used as query, the features are the characters newlineconverted into ASCII (American Standard Code for Information Interchange) values. When the sound wave file is input as query, dynamic time warping (DTW) is applied to find out best match of the spoken word and subsequently, 10 cepstrum values are calculated which will be used as features. When the image is used as query, then features are obtained by finding number of objects, region properties of the objects matching a standard template to confirm for closed polygons and grey level cooccurrence newlinematrix (GLCM) for the extraction of texture properties of the image ( clouds, water etc). All these features are appended and used for training and testing the proposed artificial neural network (ANN) algorithms. The proposed ANN algorithms are supervised back propagation algorithm (BPA) and Radial basis function (RBF). The BPA undergoes weight updation for each training pattern and goes thorough many iterations until a mean squared error (MSE) value is reached. When the specified MSE is reached, a set of final weights are stored. In RBF training, RBF values are obtained for each training newlinepattern, based on the number of centers used, and a transformation process is carried out to obtain a set of final weights. During testing of BPA / RBF, actual retrieval of video is achieved based on the outputs obtained in the output layer of BPA / RBF. The retrieval process is achieved, by searching the available template using the outputs of the ANN.%%%References p. 114-131, Appendix p. 132-145, Synopsis included