Human Activity Inference Systems for Context-aware Internet of Things Applications

Abstract

After the advent of Internet of Things (IoT) in late 1990s, everyday objects are envisioned to be tightly interconnected by equipping them with wireless sensing and actuation capabilities to monitor and manipulate target environments. Consequently, such an extensive interconnection of physical objects generates massive amount of sensory data representing the physical phenomenon. While such functionalities are the basis of IoT, the raw data itself cannot be used to make smart and autonomous actuations. Considering such a lack of intelligence, many scholars stressed the need of context-aware computings to attain comprehensive knowledge regarding environmental contexts, including human activity contexts. Considering the fact that most of IoT applications and services are designed to be human centric, we emphasized the need of Human Activity Inference Systems (HAIS) while addressing key research problems raised when HAIS is used by different IoT services. Throughout this work, we selected four interesting services, which we elaborated in the main chapters, within three IoT application domains. After going through each technical chapters, we noted that accurate context inferencing results yielded from HAIS can benefit one or more IoT architecture layer(s) by carefully controlling various system parameters depending on the human activity contexts. Contributions of each technical chapters can be briefly summarized as follows. The very first chapter started from questioning the possibilities of identifying a human activity context, namely driving. According previous studies, texting while driving significantly increases the chance of getting into an accidents, thus appropriate protection services must be applied when the user is found to be driving a vehicle. In order to extract such a human activity context, we proposed a HAIS which effectively extract, analyze, and fuse the heterogeneous sensory information on commodity smartphones. In accordance with the IoT architecture layers, this chapter contributed to improve the performance of top two layers, namely application service and information integration layers. Next chapter was initiated with an objective to compress the capabilities of a microphone sensor to fit in resource-limited IoT platforms. In fact, we proposed a HAIS, which extracts most frequently found office meeting activities using heterogeneous IoT-enabled objects installed throughout a smart office environment. we believe that our considerations can open the possibilities to quickly commercialize human activity context extraction systems to everyday users by providing them with an easy-to-install system that is applicable to various application scenarios. Under such a system, various smart office applications are be integrated to increase the human comfort levels and maximize the operational efficiency. The contributions of this chapter dealt with information integration layer and application service layer. Throughout the third chapter, we present an activity-aware sensor cycling solution tailored to smart home environments that significantly increases the accuracy and reliability of activity detection by exploiting the inherent correlations residing in the residents' behavioral patterns. The proposed solution predicts the activity patterns that are most likely to occur and, based on the prediction results, determines the role of each sensor to monitor the environment. The contributing factor with this work was that human activity context extracted by HAIS can also be used to enhance the performance of object sensing layer by carefully controlling the sensory data acquisition techniques. We note that this chapter dealt with the object sensing layer from the IoT architecture. The last chapter introduced a light-weight, energy-efficient, low-latency privacy protection scheme for smart home environments against side channel attacks. According to the studies, data encryptions cannot provide acceptable level of privacy protections against smart adversaries. In fact, we introduced the concept of cloaking activities to hide the actual human activity context while saving as much energy as possible. Contributing factors of this work dealt with privacy protection and data exchange layer to support the intelligent IoT applications and services provided through smart home environments. ⓒ 2016 DGIST