Towards quality and energy aware complex event processing

There is a rising interest to use sensors of various kinds for human activity recognition. Complex human activities are composed of temporally related simple activities and the concept of complex events is well suited to capture this relationship and to detect complex activities, i.e., complex events. This paper addresses the need for high accuracy of complex activity detection and low battery consumption of the sensors. To achieve this goal, this paper presents a framework to obtain the required accuracy of complex activity detection and minimize the cost of delivering samples from the sensors to the sink. We formalize the problem as an integer-programming problem and solve it by an approximation approach. The goal is to satisfy the user specified accuracy at a minimum cost. We compare the performance of our approximation approach with exact method for complex human activity determination. Evaluation results with a publicly available dataset demonstrate the potential impact of our approach.