Violent Videos Classification Using Wavelet and Support Vector Machine

Video is a technology for sending electronic signals from a set of images. Video can record an event that can be viewed again for certain purposes. One of them can be used to detect the presence of violence in that recorded video. However, manual observation takes a long time and foresight. Therefore, it is necessary to automatically classify the presence or absence of violence in a video using computer vision. In this paper, we extract features from video using Discrete Wavelet Transform (DWT) and then classify them using Support Vector Machine (SVM). As a comparison, feature extraction is also carried out using Principal Component Analysis (PCA). By using the Hockey Fights and Movies datasets, training and testing were carried out with kernel linear, radial basis function, and polynomials. The experimental results show that feature extraction with DWT has a higher accuracy rate than PCA. However, the time for training and testing required by DWT is longer than PCA.