Locomotion Traces Data Mining for Supporting Frail People with Cognitive Impairment

The rapid increase in the senior population is posing serious challenges to national healthcare systems. Hence, innovative tools are needed to early detect health issues, including cognitive decline. Several clinical studies show that it is possible to identify cognitive impairment based on the locomotion patterns of older people. Thus, this thesis at first focused on providing a systematic literature review of locomotion data mining systems for supporting Neuro-Degenerative Diseases (NDD) diagnosis, identifying locomotion anomaly indicators and movement patterns for discovering low-level locomotion indicators, sensor data acquisition, and processing methods, as well as NDD detection algorithms considering their pros and cons. Then, we investigated the use of sensor data and Deep Learning (DL) to recognize abnormal movement patterns in instrumented smart-homes. In order to get rid of the noise introduced by indoor constraints and activity execution, we introduced novel visual feature extraction methods for locomotion data. Our solutions rely on locomotion traces segmentation, image-based extraction of salient features from locomotion segments, and vision-based DL. Furthermore, we proposed a data augmentation strategy to increase the volume of collected data and generalize the solution to different smart-homes with different layouts. We carried out extensive experiments with a large real-world dataset acquired in a smart-home test-bed from older people, including people with cognitive diseases. Experimental comparisons show that our system outperforms state-of-the-art methods.