"Wrist-worn Activity Recognition Based on Smart Watches"
Wrist-worn Activity Recognition Based on Smart Watches
Sprache des Titels:
The recognition of daily activities of human behavior by the application of inertial sensor technology got increased attention after customized sensor platforms became widely available. Since this time, various activity recognition application scenarios have been researched and revised with variable success. Most existing work has in common that it is always applied within a very restricted domain, mainly within laboratory setups. By this restriction usually the quality of recognition results increase within their domain of application but limits system accuracy within practical scenarios due to the unsupervised and unguided applicability in everyday living environments. The utilization of mobile and embedded devices and appliances opened the potential of performing human activity recognition tasks within different real life scenarios. Emerging systems heavily use inertial sensor technology to perform recognition of specialized activities for example within healthcare, fitness or work life scenarios. Again, they work considerably well within the areas they are explored and generally fail outside their application domains. In this work a system is presented that specifies and implements an activity recognition system that supports the practical application within daily living situations. By specifying a generalized model approach it is possible to establish (semi-)automatical and personalized activity of daily living recognition routines. They can be utilized for real world appliance control or background personal information management (e.g., do not deliver notifications when the user is in certain situations)