Personalized emotion-aware video streaming for the elderly

Abstract

We consider the problem of video therapy services for the elderly based on their current emotional status. Given long hours watching TV in the elder population, most of the existing TV services are not geared for them. The elderly cannot tolerate complexity and negativity due to decline in cognitive abilities. In addition, the program is not adapted to the user’s current emotional status. As a result, existing TV services can not achieve optimal performance across a broad set of user types and context. To provide content tailored to individual needs, and interests of the elderly, caregivers have to select an appropriate program manually. However, this can not scale well due to shortage of caregivers and high monetary cost. We present the personalized emotion-aware video streaming system, a redesign of conventional TV system to provide appropriate program flexibly, efficiently and responsively. Our proposed architecture adds video affective profiling, real-time emotion detection and Markov decision process based video program generation to the streaming service to this end. We present a complete implementation of our design. Trace-driven simulation has shown the effectiveness of our system.