Classification of Human Motion Data Based on Inertial Measurement Units in Sports: A Scoping Review

Abstract

Inertial measurement units (IMU) are widely used in sports applications to digitise human motion by measuring acceleration and rotational velocity in three-dimensional space. A common machine learning problem is the classification of human motion primitives from IMU data. In order to investigate the classification methods used in the existing literature and to analyse whether and how the time-dependent data structure is considered in the classification process of motion data analysis in sports, a scoping review was conducted. Based on a keyword search, articles from 2010 to 2021 were extracted, and 93 articles were relevant for data extraction. Over- and undersampling of data and data augmentation techniques were rarely used. The classification methods applied can be divided into three main branches: classic machine learning and deep learning models, threshold-based approaches, and dynamic time warping. The most often applied algorithms were support vector machines (SVM), followed by neural networks and k-nearest neighbours. In comparative works, when more than one classifier was applied, random forests, neural networks, boosting models and SVM were found to be the methods that achieved the highest accuracy. If the time-dependent data structure was taken into account, it was incorporated either within the models, for example, by using long-short-term memory models or within the feature calculation step by using rolling windows with an overlap, which was the most common method of considering the time dependency of the IMU data.

Surrogate modelling of heartbeat events for improved J-peak detection in BCG using deep learning

“…For example, the calculation of heart rate variability metrics relies on interbeat intervals (IBI) and reflects a person’s physiological state and health ( Shaffer and Ginsberg, 2017 ). In addition, most sleep stage classification algorithms rely on IBIs as input, i.e., interpolation of the temporal differences between successive heartbeats ( Kranzinger et al, 2023 ).…”

“…The dataset comprises both the BCG and ECG, which were collected from 11 participants over a period of approximately 8 h during 17 nights of sleep. The data were collected as part of the Virtual Sleep Lab project, as detailed in Kranzinger et al (2023) . The electrocardiogram (ECG) was recorded using the BrainAmp Standard Amplifier (Brain Products GmbH, Germany), a laboratory-standard device known for high-quality recordings.…”

“…These drawbacks limit the sample sizes of PSG studies and render them unsuitable for longitudinal studies. Recent trends in sleep research indicate that high-accuracy estimations of sleep stage fluctuations can be derived from variations in signals such as inter-beat interval (IBI) time-series ( kranzinger et al, 2023 ) or alterations in respiratory effort over time when analyzed with machine learning models.…”

“…Given the suitability of ballistocardiography for unobtrusive long-term sleep measurement, and the availability of improved machine learning algorithms and computational resources, there is an increasing amount of research focused on heartbeat extraction from BCG. Research has shown that J-peaks can be used to predict the subject’s sleep stages and therefore sleep quality ( Kranzinger et al, 2023 ). This work proposes a novel approach to model heartbeat events as a continuous signal, thereby improving the accuracy of heartbeat extraction within a supervised deep learning framework.…”