Infrastructure support for augmented memory

Abstract

During the past few decades, we have witnessed the development of an increasing number of products and research projects in the area of lifelogging. These range from early works by Steve Mann and Microsoft's MyLifeBits project up to wearable cameras like the Autographer and Narrative Clip. These and other previous approaches focused on sensors worn by the user rather than sensors in the environment. However, as the author already discussed in [CMD14] wearable sensors have several shortcomings. These include blurry pictures due to the user's movements, sensors occluded by clothes and failing devices due to drained batteries or insufficient storage. In this thesis we pursue a complementary approach. We designed and implemented a middleware for memory augmentations that can aggregate sensor data from a diverse set of sources. We focus on privacy protection and scalability due to the system's ubiquity and comprehensive use of sensors. The system therefore features a location based access control system that does not make users traceable. We lowered bandwidth requirements via distributed storing of sensor data and duplicate avoidance. In parallel, we designed an Android library to process location based information. The Android library can be used with the system designed in this bachelor thesis, as well as the display appropriation framework Tacita developed by researchers at Lancaster University. We demonstrate the system's scalability by showing that response times increase linearly with the scale at which the system operates. Our measurements thereby included the simulation of up to 10000 recordings and 2000 sensors on one of the system's nodes. Finally, we investigated the system's financial feasibility by estimating its operational costs.