Automated selection of appropriate advertisements for digital signage by analyzing crowd demographics

Digital signage is a relatively newer form of advertising which provides flexibility of changing advertising contents dynamically. The core challenge in the design of an effective digital signage system is selecting the appropriate advertising contents keeping in view the target audience. In this context, this paper presents a framework for the automated selection of appropriate advertisements for digital signage systems using computer vision techniques. The advertisement is selected based on the audience type which is characterized by the number of people, their gender and their stereotypic outlook. We use a fused histogram of gradient (HOG) based approach for people and gender detection thus saving the computational time of separate feature set computations. We also propose a novel rule based mechanism to collect crowd profile information based on individuals' dressing and shape of me shoulders. The experiments have been performed to test the individual modules and the integrated system. The experimental results show the applicability of our framework in real world scenarios.