Using Accelerometer and GPS Data for Real-Life Physical Activity Type Detection.

This paper aims to examine the role of global positioning system (GPS) sensor data in real-life physical activity (PA) type detection. Thirty-three young participants wore devices including GPS and accelerometer sensors on five body positions and performed daily PAs in two protocols, namely semi-structured and real-life. One general random forest (RF) model integrating data from all sensors and five individual RF models using data from each sensor position were trained using semi-structured (Scenario 1) and combined (semi-structured + real-life) data (Scenario 2). The results showed that in general, adding GPS features (speed and elevation difference) to accelerometer data improves classification performance particularly for detecting non-level and level walking. Assessing the transferability of the models on real-life data showed that models from Scenario 2 are strongly transferable, particularly when adding GPS data to the training data. Comparing individual models indicated that knee-models provide comparable classification performance (above 80%) to general models in both scenarios. In conclusion, adding GPS data improves real-life PA type classification performance if combined data are used for training the model. Moreover, the knee-model provides the minimal device configuration with reliable accuracy for detecting real-life PA types.

Assessing the Transferability of Physical Activity Type Detection Models: Influence of Age Group Is Underappreciated.

Increasing the amount of physical activity (PA) in older adults that have shifted to a sedentary lifestyle is a determining factor in decreasing health and social costs. It is, therefore, imperative to develop objective methods that accurately detect daily PA types and provide detailed PA guidance for healthy aging. Most of the existing techniques have been applied in the younger generation or validated in the laboratory. To what extent, these methods are transferable to real-life and older adults are a question that this paper aims to answer. Sixty-three participants, including 33 younger and 30 older healthy adults, participated in our study. Each participant wore five devices mounted on the left and right hips, right knee, chest, and left pocket and collected accelerometer and GPS data in both semi-structured and real-life environments. Using this dataset, we developed machine-learning models to detect PA types walking, non-level walking, jogging/running, sitting, standing, and lying. Besides, we examined the accuracy of the models within-and between-age groups applying different scenarios and validation approaches. The within-age models showed convincing classification results. The findings indicate that due to age-related behavioral differences, there are more confusion errors between walking, non-level walking, and running in older adults' results. Using semi-structured training data, the younger adults' models outperformed older adults' models. However, using real-life training data alone or in combination with semi-structured data generated better results for older adults who had high real-life data quality. Assessing the transferability of the models to older adults showed that the models trained with younger adults' data were only weakly transferable. However, training the models with a combined dataset of both age groups led to reliable transferability of results to the data of the older subgroup. We show that age-related behavioral differences can alter the PA classification performance. We demonstrate that PA type detection models that rely on combined datasets of young and older adults are strongly transferable to real-life and older adults' data. Our results yield significant time and cost savings for future PA studies by reducing the overall volume of training data required.

The Key Factors in Physical Activity Type Detection Using Real-Life Data: A Systematic Review.

Background: Physical activity (PA) is paramount for human health and well-being. However, there is a lack of information regarding the types of PA and the way they can exert an influence on functional and mental health as well as quality of life. Studies have measured and classified PA type in controlled conditions, but only provided limited insight into the validity of classifiers under real-life conditions. The advantage of utilizing the type dimension and the significance of real-life study designs for PA monitoring brought us to conduct a systematic literature review on PA type detection (PATD) under real-life conditions focused on three main criteria: methods for detecting PA types, using accelerometer data collected by portable devices, and real-life settings. Method: The search of the databases, Web of Science, Scopus, PsycINFO, and PubMed, identified 1,170 publications. After screening of titles, abstracts and full texts using the above selection criteria, 21 publications were included in this review. Results: This review is organized according to the three key elements constituting the PATD process using real-life datasets, including data collection, preprocessing, and PATD methods. Recommendations regarding these key elements are proposed, particularly regarding two important PA classes, i.e., posture and motion activities. Existing studies generally reported high to near-perfect classification accuracies. However, the data collection protocols and performance reporting schemes used varied significantly between studies, hindering a transparent performance comparison across methods. Conclusion: Generally, considerably less studies focused on PA types, compared to other measures of PA assessment, such as PA intensity, and even less focused on real-life settings. To reliably differentiate the basic postures and motion activities in real life, two 3D accelerometers (thigh and hip) sampling at 20 Hz were found to provide the minimal sensor configuration. Decision trees are the most common classifier used in practical applications with real-life data. Despite the significant progress made over the past year in assessing PA in real-life settings, it remains difficult, if not impossible, to compare the performance of the various proposed methods. Thus, there is an urgent need for labeled, fully documented, and openly available reference datasets including a common evaluation framework.