No abstract

Recent studies in Human Activity Recognition (HAR) have shown that Deep Learning methods are able to outperform classical Machine Learning algorithms. One popular Deep Learning architecture in HAR is the DeepConvLSTM. In this paper we propose to alter the DeepConvLSTM architecture to employ a 1-layered instead of a 2-layered LSTM. We validate our architecture change on 5 publicly available HAR datasets by comparing the predictive performance with and without the change employing varying hidden units within the LSTM layer(s). Results show that across all datasets, our architecture consistently improves on the original one: Recognition performance increases up to 11.7% for the F1-score, and our architecture significantly decreases the amount of learnable parameters. This improvement over DeepConvLSTM decreases training time by as much as 48%. Our results stand in contrast to the belief that one needs at least a 2-layered LSTM when dealing with sequential data. Based on our results we argue that said claim might not be applicable to sensor-based HAR. CCS CONCEPTS• Computing methodologies → Neural networks; • Humancentered computing → Ubiquitous and mobile computing design and evaluation methods.

No abstract

Background One relevant strategy to prevent the onset and progression of type 2 diabetes mellitus (T2DM) focuses on increasing physical activity. The use of activity trackers by patients could enable objective measurement of their regular physical activity in daily life and promote physical activity through the use of a tracker-based intervention. This trial aims to answer three research questions: (1) Is the use of activity trackers suitable for longitudinal assessment of physical activity in everyday life? (2) Does the use of a tracker-based intervention lead to sustainable improvements in the physical activity of healthy individuals and in people with T2DM? (3) Does the accompanying digital motivational intervention lead to sustainable improvements in physical activity for participants using the tracker-based device? Methods The planned study is a randomized controlled trial focused on 1,642 participants with and without T2DM for 9 months with regard to their physical activity behavior. Subjects allocated to an intervention group will wear an activity tracker. Half of the subjects in the intervention group will also receive an additional digital motivational intervention. Subjects allocated to the control group will not receive any intervention. The primary outcome is the amount of moderate and vigorous physical activity in minutes and the number of steps per week measured continuously with the activity tracker and assessed by questionnaires at four time points. Secondary endpoints are medical parameters measured at the same four time points. The collected data will be analyzed using inferential statistics and explorative data-mining techniques. Discussion The trial uses an interdisciplinary approach with a team including sports psychologists, sports scientists, health scientists, health care professionals, physicians, and computer scientists. It also involves the processing and analysis of large amounts of data collected with activity trackers. These factors represent particular strengths as well as challenges in the study. Trial Registration The trial is registered at the World Health Organization International Clinical Trials Registry Platform via the German Clinical Studies Trial Register (DRKS), DRKS00027064. Registered on 11 November 2021.