From Compression of Wearable-based Data to Effortless Indoor Positoning

Abstract

The dissertation focuses on boosting the energy efficiency of IoT and wearable devices by implementing lossy compression techniques onto sensor-based time-series data and into indoor localization paradigms. The thesis deals with lossy compression mechanisms that can be implemented for energy-e¿cient, delay-sensitive wearable data gathering, transfer, and storage. The novel DLTC compression method ensures optimal compression ratio and reconstruction error trade-off, with minimum complexity and delay. In the scope of indoor positioning, the proposed bit-level, feature-wise, and sample-wise reduction of the radio map supports accurate positioning while saving resources in data storage and transfer. The work implements a multi-dimensional compression of the radio map to boost the performance e¿ciency of the positioning system and proposes a cascade model to compensate for k-NN¿s drawback of computationally expensive prediction on voluminous datasets.