Context Recognition in AmI Environments: Sensor and MMES Technology, Recognition Approaches, and Pattern Recognition Methods

Abstract

There exist a vast range of AmI architectures that essentially aim to provide the appropriate infrastructure for AmI systems. Typically, they include many sensors of diverse types, information processing systems or computing devices where modeling and reasoning occur, and actuators through which the system acts, reacts, or pre-acts in the physical world. There are many permutations of enabling technologies and computational processes of AmI, which result in many heterogeneous components (devices and systems and associated software applications) which have to interconnect and communicate seamlessly across disparate networks as part of vast architectures enabling context awareness, machine learning and reasoning, ontological representation and reasoning, and adaptation of services. The sensors are basically utilized to acquire the contextual data needed for the context recognition process-that is, observed information as input for AmI systems to analyze, model, and understand the user's context, so to undertake in a knowledgeable manner actions accordingly. Sensor technology is thus a key enabler of context awareness functionality in AmI systems. Specifically, to acquire, fuse, process, propagate, interpret, and reason about context data in the AmI space to support adaptation of services requires using dedicated sensors and signal and data processing techniques, in addition to sophisticated context recognition algorithms based on a wide variety of methods and techniques for modeling and reasoning. The challenge of incorporating context awareness functionality in the AmI service provision system lies in the complexity associated with sensing, learning, capturing, representing, processing, and managing context information.Context-aware systems are increasingly maturing and rapidly proliferating, spanning a variety of application domains, owing to recent advances in capture technologies, the diversity of recognition approaches, multi-senor fusion techniques, and sensor networks, as well as pattern recognition algorithms and representation and reasoning techniques. Numerous recognition approaches have been developed 129 and studied, and a wide variety of related projects have been carried out within various domains of context awareness. Most of early research work on context awareness focused on user's physical context, which can be inferred using different types of sensing facilities, including stereo-type cameras, RFID, and smart devices. While most attempts to use context awareness within AmI environments were centered on the physical elements of the environment or users, in recent years, research in the area of context recognition has shifted the focus to human elements of context, such as emotional states, cognitive states, physiological states, activities, and behaviors. This has led to the development and employment of different recognition methods, mainly vision-based, multisensory-based, and sensor-based context and activity recognition approaches. Furthermore, investigating methods for context recognition ...