Real-Time Recognition of Physical Activities and Their Intensities Using Wireless Accelerometers and a Heart Rate Monitor

Abstract

In this paper, we present a real-time algorithm for automatic recognition of not only physical activities, but also, in some cases, their intensities, using five triaxial wireless accelerometers and a wireless heart rate monitor. The algorithm has been evaluated using datasets consisting of 30 physical gymnasium activities collected from a total of 21 people at two different labs. On these activities, we have obtained a recognition accuracy performance of 94.6% using subject-dependent training and 56.3% using subjectindependent training. The addition of heart rate data improves subject-dependent recognition accuracy only by 1.2% and subject-independent recognition only by 2.1%. When recognizing activity type without differentiating intensity levels, we obtain a subjectindependent performance of 80.6%. We discuss why heart rate data has such little discriminatory power.

Statistical Machine Learning for Automatic Assessment of Physical Activity Intensity Using Multi-axial Accelerometry and Heart Rate

“…The accuracy achieved by our approach is remarkable (86.97% for fc-means) and outperforms other PAI classifiers using combined accelerometry and HR, like Munguia Tapia [7] (58.40% accuracy in subject-independent classification). However, it must be noted that our proposal is explicitly intensity-oriented; while theirs was activity type-oriented (with certain types allowing several intensities), so that their number of possible classes was considerably higher.…”

“…To solve this issue, several authors developed activity-specific EE estimation schemes [6][7][8]. Conversely, we intend to explore PAI-specific estimators based on the algorithms presented here.…”

Behavior Analysis Based on Coordinates of Body Tags

“…Five tri-axial accelerometers distinguished 30 physical activities of various intensities with the accuracy of 94.9 % with person-dependent training and 56.3 % with personindependent training [11]. We used person-independent training, which resulted in accuracy above 90 %, although the number of activities in our experiments was admittedly lower.…”

kNN Sampling for Personalised Human Activity Recognition

“…However, comparative evaluation with personalised models, trained using subject-dependent examples, show this to produce more accurate predictions [3,8,16]. In [16], a general model was compared with a personalised model using a c4.5 decision tree classifier. The general model produced an accuracy of 56.3% while the personalised model produced an accuracy of 94.6% using the same classification algorithm, which is an increase of 39.3%.…”

“…Previous works have shown using subject-dependent data to result in superior performance [3,8,16]. The relatively poorer performance of subject-independent models can be attributed to variations in activity patterns, gait or posture between different individuals [11].…”