Commercial Block Detection in Broadcast News Videos

Abstract

Automatic identification and extraction of commercial blocks in telecast news videos find a lot of applications in the domain of broadcast monitoring. Existing works in this domain have used channel specific assumptions, machine learning techniques and frequentist approaches for detecting commercial video segments. We note that in the Indian context, several channel specific assumptions do not hold and often news and commercials have comparable frequencies of occurrence. This motivates us to use the machine learning techniques for classifying commercials in news videos. Our main contribution lies in the proposal of two features which are shown to outperform the existing audio-visual featuresfirst, the MFCC bag of words (BoW) as audio track feature and second, overlaid text distribution as video shot feature. The shot feature space is further extended by appending contextual features which are categorized by SVM based classifiers. Additionally, we have used a post-processing stage to suppress the false positives. We have experimented with 54 hours of video acquired from three different Indian English based news channels and have obtained a F-measure of around 97%.

Automatic loudness control in short-form content for broadcasting

“…Commercial block detectors that use audiovisual features may build more complex decision regions (Vyas et al, 2014). One of the challenges of this work was to determine whether satisfactory results could be achieved with a limited number of audio features, in terms of developing a classifier that does not introduce significant delay.…”

“…, that correspond to 150 h of recorded channels (Vyas et al, 2014). Samples were arbitrarily divided into two parts, twothirds for training and one-third for validation.…”

Linear Maximum Margin Classifier for Learning from Uncertain Data

“…The proposed algorithm is evaluated in the problem of detection of advertisements in TV news videos using the publicly available and very large dataset of [47]. This dataset comprises 120 hours of TV news broadcasts from CNN, CNNIBN, NDTV, and TIMES NOW (approximately 22k, 33k, 17k, and 39k videos, respectively).…”

“…Since the information about the input variance is provided just for the 24 of the 4125 features, there is no natural way of estimating a single variance value, i.e., an isotropic covariance matrix, for each training example. Table 4 shows the performance of the proposed linear SVM-GSU (LSVM-GSU) in terms of F1 score in comparison to LSVM, similarly to [47]. As discussed above, since methods that model the uncertainty isotropically (such as [14], [18], [27]), are not applicable in this dataset, we experimented on this dataset using only the proposed algorithm and the standard linear SVM.…”

“…As discussed above, since methods that model the uncertainty isotropically (such as [14], [18], [27]), are not applicable in this dataset, we experimented on this dataset using only the proposed algorithm and the standard linear SVM. Following the protocol of [47], we did cross-dataset training and testing. For optimizing the λ parameter of both LSVM and LSVM-GSU we used a line search on a 3-fold cross-validation procedure.…”