Visual information-based activity recognition and fall detection for assisted living and ehealthcare

TL;DR: This chapter mainly focuses on describing visual information-based daily activity recognition and anomaly detection through using low-resolution visual sensors, and provides further support to the robustness of manifold-based methods.

Abstract: Ambient intelligence for assisted living and healthcare has drawn increasing interest due to population aging across many countries. Challenges remain in developing robust methods for effective assisted living systems under complex real scenarios. Understanding/recognition of human activities is one of the fundamental issues in a human-centric smart environment, where visual data provides rich information on human behaviors including their interaction with other objects and surroundings. Real-time or near real-time visual information-based approaches offer effective analysis without the risk of invading the privacy, where videos are discarded after extracting features. This chapter mainly focuses on describing visual information-based daily activity recognition and anomaly detection through using low-resolution visual sensors. First, current state-of-the-art methods on visual activity recognition are briefly reviewed. Detailed descriptions are then given on three robust methods that exploit smooth manifolds. Manifold-based methods are attractive as human activity and context features can be efficiently represented by using low-dimensional smooth manifolds. Finally, experimental results and performance of several methods are given and compared, which provide further support to the robustness of manifold-based methods for visual activity recognition and anomaly detection. Information on some publicly available datasets is also included to facilitate the use of benchmark datasets for testing in the near future.