Age-Based Ranking of YouTube Videos for Improved Parental Controls in Smart TV Environment

YouTube is a popular social media networking site that contains billions of videos. Many YouTube videos target children of different ages with offensive, inappropriate, violent, etc. Numerous one-size-fits-all countermeasures and research work have been deployed and suggested. However, these solutions are ineffective in accurately detecting inappropriate content for a diverse audience of different needs and requirements. In this research work, instead of one-size-fits-all, we consider a mutable age-based context, where different Age-Groups (AG) have different choices and needs. A novel and real-time approach have been proposed to prevent and allow the audience of varying AG towards the diverse content of YouTube, specifically in a smart TV-watching scenario. The proposed system analyses the running video through its metadata for teenagers, children, and adults. In parallel with data, the proposed model captures the viewers in real-time, detects their age, and checks the displayed video against the detected AG for appropriateness. The same system responds to parental consent and overrides its stop/play policies according to the direct input provided by parents/guardians with a diverse psychological setup, developed by their beliefs, religion, and cultural sensitivities. The proposed solution leverages Random Forest Classifier–a supervised text classification approach with 80% accuracy and Convolutional Neural Network for age determination using the Caffe model.