Tracking a person’s activities is relevant in a variety of contexts, from health and group-specific assessments, such as elderly care, to fitness tracking and human–computer interaction. In a clinical context, sensor-based activity tracking could help monitor patients’ progress or deterioration during their hospitalization time. However, during routine hospital care, devices could face displacements in their position and orientation caused by incorrect device application, patients’ physical peculiarities, or patients’ day-to-day free movement. These aspects can significantly reduce algorithms’ performances. In this work, we investigated how shifts in orientation could impact Human Activity Recognition (HAR) classification. To reach this purpose, we propose an HAR model based on a single three-axis accelerometer that can be located anywhere on the participant’s trunk, capable of recognizing activities from multiple movement patterns, and, thanks to data augmentation, can deal with device displacement. Developed models were trained and validated using acceleration measurements acquired in fifteen participants, and tested on twenty-four participants, of which twenty were from a different study protocol for external validation. The obtained results highlight the impact of changes in device orientation on a HAR algorithm and the potential of simple wearable sensor data augmentation for tackling this challenge. When applying small rotations (<20 degrees), the error of the baseline non-augmented model steeply increased. On the contrary, even when considering rotations ranging from 0 to 180 along the frontal axis, our model reached a f1-score of $0.85\pm 0.11$ against a baseline model f1-score equal to $0.49\pm 0.12$.

As a result of technological advancements, functional capacity assessments, such as the 6-minute walk test, can be performed remotely, at home and in the community. Current studies, however, tend to overlook the crucial aspect of data quality, often limiting their focus to idealised scenarios. Challenging conditions may arise when performing a test given the risk of collecting poor-quality GNSS signal, which can undermine the reliability of the results. This work shows the impact of applying filtering rules to avoid noisy samples in common algorithms that compute the walked distance from positioning data. Then, based on signal features, we assess the reliability of the distance estimation using logistic regression from the following two perspectives: error-based analysis, which relates to the estimated distance error, and user-based analysis, which distinguishes conventional from unconventional tests based on users’ previous annotations. We highlight the impact of features associated with walked path irregularity and direction changes to establish data quality. We evaluate features within a binary classification task and reach an F1-score of 0.93 and an area under the curve of 0.97 for the user-based classification. Identifying unreliable tests is helpful to clinicians, who receive the recorded test results accompanied by quality assessments, and to patients, who can be given the opportunity to repeat tests classified as not following the instructions.

Walking speed and distance are usually collected when performing clinical tests such as the 6-Minute Walk Test (6MWT). Wearable devices and smartphones can help bring these tests to the home environment. However, there are difficulties in obtaining measures of distance indoors, where GPS cannot be relied on. Step counting is another even simpler form of data collection that can be obtained through digital technologies. In this work, we investigate the relationship between the step count variable and the standardised 6-Minute Walk Distance (6MWD) variable. By considering 176 6MWTs from 55 participants, we found a high correlation between ground truth distance and the number of steps taken during a test (0.83). Additionally, when considering low-quality outdoor tests, using the step count becomes significantly more reliable (MAE of 22.5m) compared to a state-of-the-art algorithm (MAE of 93.8m). We conclude that step count can be considered as a valid proxy to estimate 6MWD and a candidate approach for monitoring patients' physical health in free-living conditions.

Monitoring physical activity, function and capacity through Digital Health Technology (DHT) has a high potential to benefit healthcare providers and patients. Current practices of observing physical function and capacity use standard tests such as the 6-Minute Walk Test (6MWT) or the Timed Up and Go, which collect momentary information on the patient’s health status. The use of DHT is demonstrated to enhance these types of assessments, from the instrumentation of physical tests with technology to the analysis of digital biomarkers collected during one’s daily life. Research at national and international levels investigates these topics on a large scale; however, it often lacks transparency and details in regards to used algorithms and data quality. These aspects are crucial when implementing technology for health-related purposes, where data quality and methods accuracy are fundamental for impacting clinical practices. This thesis investigates how DHT can support physical testing in ecological or community environments, answering three main research questions related to data collection and quality, algorithms, and the association between daily life physical activity and physical tests. Throughout six articles, this work investigates methods for inferring walked distance during the 6MWT in indoor and outdoor conditions, highlighting the importance of data quality collected through DHT. It shares a publicly available dataset providing inertial and localization information of patients and healthy volunteers. Lastly, it reports on insights and common procedures regarding DHT used during everyday life and its relation with physical tests. Foundations are laid for future work in this domain.