Windowing is an established technique employed within dense sensing environments to extract relevant features from sensor data streams. Among the established approaches of Explicit, Time-based and Sensor-Event based windowing, Dynamic windowing approaches are beginning to emerge. These dynamic approaches claim to address the inherent shortcomings of the aforementioned established approaches by determining the appropriate window length for live sensor data streams in real-time, thereby offering the potential to optimize and increase the recognition of these sensor represented activities. Beyond these potential benefits, dynamic approaches can also support anomaly detection by actively uncovering new, unknown window patterns within a trained model. This paper presents findings from a study which utilizes data from a single source dataset, towards benchmarking and comparing more traditional windowing approaches against a dynamic windowing approach. The experiments conducted on a real-world smart home dataset suggest Time-based windowing is the best approach. Through evaluation of results, Dynamic windowing approaches may benefit from carefully annotated datasets.

A common challenge in smart environments is tracking individuals throughout different environments. Reliable solutions to this problem often involve cameras, which pose significant privacy issues, or trackable tags such as RFID which require that individuals be 'prepared' for the environment. In this paper an exploratory study is presented that investigates the utility of portable visible range spectroscopy hardware for the purposes identifying individuals based on the spectral pattern of their clothing. This is done by assessing the accuracy of a data-driven machine learning models when differentiating clothing. These sensors have the potential to achieve similar success to using the colour histogram from camera tracking without the privacy issues, and without the need to pre-tag individuals as with RFID's.

No abstract