Audio-Video Detection and Fusion of Broad Casting Information

Abstract

In the last few decade of multimedia information systems, audio-video data has become an glowing part in many digital computer applications. Audio-video classification has been becoming a focus in the research of audio-video processing and pattern recognition. Automatic audio-video classification is very useful to audio-video indexing, content-based audio-video retrieval and on-line audio-video distribution such as online audio-video shopping, but it is a challenge to extract the most similar and salient themes from huge data of audio-video. In this paper, we propose effective algorithms to automatically segmentation and classify audio-video clips into one of  Six classes: advertisement, cartoon, songs, serial,  movie and news. For these categories a number of acoustic and visual features that include Mel Frequency Cepstral Coefficients, Color Histogram are extracted to characterize the audio and video data. The autoassociative neural network model (AANN) is used to capture the distribution of the acoustic and visual feature vectors. The AANN model captures the distribution of the acoustic and visual features of a class, and the back propagation learning algorithm is used to adjust the weights of the network to minimize the mean square error for each feature vector.

Keywords: - Audio and Video detection, Audio and Video fusion, Mel Frequency Cepstral Coefficient, Color Histogram, Autoassociative Neural Network Model(AANN)