Activity-aware Sensor Cycling for Human Activity Monitoring in Smart Homes
Smart homes are one of the Internet of Things domains intended to support and aid the residents through various smart services. These services require accurate context inferences using daily activity patterns and environmental properties. To satisfy such a need with battery-powered sensors, various duty cycling schemes were introduced. In this letter, we propose an activity-aware sensor cycling approach that makes the best tradeoff for duty cycle adjustments by exploiting the predictable behavior of residents, thereby significantly improving the activity detection accuracy at a marginal increase of the energy consumption. Evaluation results demonstrate that it achieves up to 99% accuracy of activity detection and extends the network lifetime by supporting balanced energy consumption among sensors.
Many smart applications hinge on accurate context inference to tailor their services, which mandate extensive and continuous use of WSNs to collect streams of sensory data without missing any important activities . However, it is well known that battery-powered sensors suffer from energy depletion problems when they are extensively and continuously used. As WSNs have to be functional for a long time, a number of schemes, including duty cycling, have been introduced to achieve accurate activity detection while spending the smallest amount of energy. To counter this deficiency, event-based sensor duty cycling schemes were proposed . Their main idea is to appoint a sentry sensor that monitors an occurrence of an event-ofinterest while others are asleep and, upon detection, activates sleeping sensors to let them participate in event monitoring. However, there are two technical challenges that have to be addressed. First, the location of sentry must be carefully selected to maximize the detection accuracy. For instance, if we select a sensor attached to an oven door as a sentry while the resident is preparing a cold meal without using an oven, sensors in the kitchen will not be activated and ongoing events will be missed. Second, the expected lifetime of sentry sensors is much shorter than others since they must be active at all time. Consequently, sentries require frequent maintenance efforts, which could annoy the residents.
To accurately monitor the activities of residents, we first evenly divide a day into N time windows, each spanning _w minutes. For the kth time window wk, each sensor si is labeled as belonging to either active sensors ASk, sentry sensors TSk, or sleeping sensors SSk using Algorithm 1. Please see Section III for details of how to select an appropriate _w. The role of each sensor group is as follows. Sensors in ASk are the ones with high probability of detecting residents’ activities during wk, and hence, sensing parts must always be functional while RF is turned on only if an event is detected. In case of TSk, only sensing parts are used to handle unexpected activities. By contrast, sensors in SSk are unlikely to detect any activity during wk, thus their sensing parts are turned off to save energy while RF is in the low power listening mode. To compare the sentry-only and ASC schemes, we chose 2 sets of sentry sensors for each smart home and 3 different ASC control parameter combinations as shown in Table I, and illustrated their activity detection accuracy in Fig. 4a. According to our analysis, the accuracy of sentry-only scheme varies between 24.3_88.1%, which is poorer than ASC case 3 that achieves the accuracy of 96.1% or higher. This result demonstrates that ASC reliably maximizes the activity detection accuracy whereas the sentry-only scheme relies heavily on the (manual) choice of sentry sensors.
In this letter, we present an activity-aware sensor cycling solution tailored to smart homes, which significantly increases the accuracy and reliability of activity detection by exploiting the inherent correlations residing in the residents’ behavioral patterns. Our evaluation results demonstrate that it achieves up to 99% accuracy of activity detection guaranteeing the average accuracy of at least 80%, and extends the network lifetime by supporting fair consumption of energy among sensors. While these performance benefits make ASC attractive for various IoT applications, we will enhance ASC with the capability of self-configuring the control parameters to account for the differences in activity patterns among individuals (e.g., businessmen and elderly).
 D. J. Cook, “Learning setting-generalized activity models for smart spaces”, IEEE Intelligent Systems, vol. 2010, no. 99, pp. 1, 2010.
. G. Anastasi, M. Conti, M. Di Francesco, and A. Passarella, “Energy conservation in wireless sensor networks: A survey”, Ad Hoc Networks, vol. 7, no. 3, pp. 537-568, 2009.
 B. Jiang, B. Ravindran, and H. Cho, “Probability-based prediction and sleep scheduling for energy-efficient target tracking in sensor networks”, IEEE Trans. on Mobile Computing, vol. 12, no. 4, pp. 735-747, 2013.
 M. Magno, D. Boyle, D. Brunelli, E. Popovici, and L. Benini, “Ensuring survivability of resource-intensive sensor networks through ultra-low power overlays”, IEEE Trans. on Industrial Informatics, vol. 10, no. 2, pp. 946-956, 2014.
 Y. Rubner, C. Tomasi, and L. J. Guibas, “The earth mover’s distance as a metric for image retrieval”, International Journal of Computer Vision, vol. 40, no. 2, pp. 99-121, 2000.
 D. J. Cook, A. S. Crandall, B. L. Thomas, and N. C. Krishnan, “CASAS: A smart home in a box,” Computer, vol. 45, no 7, 2013.
 M. Tapia, S. Intille, and K. Larson, “Activity recognition in the home using simple and ubiquitous sensors”, International Conference on Pervasive Computing, Springer Berlin Heidelberg, 2004.
 P. Dutta, M. Grimmer, A. Arora, S. Bibyk, and D. Culler, “Design of a wireless sensor network platform for detecting rare, random, and ephemeral events”, Proc. ACM/IEEE IPSN’05, Los Angeles, CA, 2005.