A Segmentation Technique Based on Standard Deviation in Body Sensor Networks

Abstract

Pervasive health monitoring utilizing wearable wireless sensor nodes can greatly enhance the quality of care individuals receive. Such systems, while in terms of signal processing mostly depend on pattern recognition schemes, must operate independently of human interaction for extended periods. The lack of a general-purpose computationally inexpensive algorithm capable of segmenting sensor readings into discrete actions and nonactions has hindered the development of these systems. We examine a segmentation scheme based on standard deviation metric. We provide experimental verification of the method.