Improving Digital Video Commercial Detectors With Genetic Algorithms

The advent of digital video offers many opportunities to add features that enhance the viewing experience. One much-discussed feature is the possibility that commercials might be automatically detected in the video stream. We report on initial experiments with a class of commercial detection algorithms and show how their performance can be enhanced by applying genetic search to the optimization of some of their internal parameters. We show how a scalar genetic algorithm can locate sets of parameters in a multi-objective space (precision and recall) that outperform the values selected by an expert engineer. While a useful observation in itself, we also argue that this approach may be a necessity as the features that distinguish commercials from other video content will certainly vary with video format, the country of broadcast and possibly over time. We present the results of optimizing a commercial detection algorithm for different data sets and parameter sets. We are convinced that GAs drastically improved our approach and enabled fast prototyping and performance tuning of commercial detection algorithms.