Predicting lying, sitting and walking at different intensities using smartphone accelerometers at three different wear locations: hands, pant pockets, backpack

Khataeipour, Seyed Javad; Anaraki, Javad Rahimipour; Bozorgi, Arastoo; Rayner, Machel; Basset, Fabien A.; Fuller, Daniel

doi:10.1136/bmjsem-2021-001242

Cited by 3 publications

(1 citation statement)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A summary of the demographics (gender, age, height, and weight) is presented in Table 1 . We did not include the participants’ demographics as attributes in our dataset because, in another study [ 36 ], we found that these attributes did not significantly influence the performance of the machine learning models.…”

Section: Methodsmentioning

confidence: 99%

Exploring Orientation Invariant Heuristic Features with Variant Window Length of 1D-CNN-LSTM in Human Activity Recognition

et al. 2022

Self Cite

View full text Add to dashboard Cite

Many studies have explored divergent deep neural networks in human activity recognition (HAR) using a single accelerometer sensor. Multiple types of deep neural networks, such as convolutional neural networks (CNN), long short-term memory (LSTM), or their hybridization (CNN-LSTM), have been implemented. However, the sensor orientation problem poses challenges in HAR, and the length of windows as inputs for the deep neural networks has mostly been adopted arbitrarily. This paper explores the effect of window lengths with orientation invariant heuristic features on the performance of 1D-CNN-LSTM in recognizing six human activities; sitting, lying, walking and running at three different speeds using data from an accelerometer sensor encapsulated into a smartphone. Forty-two participants performed the six mentioned activities by keeping smartphones in their pants pockets with arbitrary orientation. We conducted an inter-participant evaluation using 1D-CNN-LSTM architecture. We found that the average accuracy of the classifier was saturated to 80 ± 8.07% for window lengths greater than 65 using only four selected simple orientation invariant heuristic features. In addition, precision, recall and F1-measure in recognizing stationary activities such as sitting and lying decreased with increment of window length, whereas we encountered an increment in recognizing the non-stationary activities.

show abstract

Section: Methodsmentioning

confidence: 99%

Exploring Orientation Invariant Heuristic Features with Variant Window Length of 1D-CNN-LSTM in Human Activity Recognition

et al. 2022

Self Cite

View full text Add to dashboard Cite

show abstract

Quantifying the Uncertainty of Human Activity Recognition Using a Bayesian Machine Learning Method: A Prediction Study

Mamiya,

Fuller

2023

Preprint

View full text Add to dashboard Cite

BackgroundMachine learning methods accurately predict physical activity outcomes using accelerometer data generated by wearable devices, thus allowing the investigation of the impact of built environment on population physical activity. While traditional machine learning methods do not provide prediction uncertainty, a new method, Bayesian Additive Regression Trees (BART) can quantify such uncertainty as posterior predictive distribution. We evaluated the performance of BART in predicting physical activity status.MethodsWe applied multinomial BART and the benchmark method, random forest, to accelerometer data in 25,424 time points, which were generated by wearable devices attached to 37 participants. We evaluated prediction accuracies and confusion matrix using leave-one-out cross-validation.ResultsBART and random forest demonstrated comparable accuracies in prediction.ConclusionsBART is a relatively novel ML method and will advance the incorporation of predicted physical activity status into built environment research. Future research includes the evaluation of the association between the built environment and predicted physical activity with and without accounting for prediction uncertainty.

show abstract

Comparing Accelerometer Processing Metrics and Hyperparameter Optimization for Physical Activity Classification Accuracy Using Machine Learning Methods

Musa,

Barua,

Stanley

et al. 2024

Journal for the Measurement of Physical Behaviour

View full text Add to dashboard Cite

Background: Physical activity (PA) is a crucial factor in maintaining good health and preventing chronic diseases. However, accurately measuring PA is challenging. Euclidean Norm Minus One (ENMO), ActiGraph Counts, and Monitor-Independent Movement Summary (MIMS) units are processing metrics used to classify PA through accelerometry, but they employ different methods to calculate activity levels. This study aimed to compare ENMO, ActiGraph Counts, and MIMS accelerometer metrics using machine learning algorithms. Methods: Data from a smartphone accelerometer were collected from 50 participants who held the smartphone in their right hand while completing six activities. The data were used to generate ENMO, ActiGraph Counts, and MIMS acceleration metrics. Random Forest, K-Nearest Neighbor, and Support Vector Machine algorithms were applied to the data to classify PA into different levels of activity intensity and types. The algorithms’ performance was evaluated using various metrics such as accuracy, precision, and recall. Results: The results showed that both the Random Forest and K-Nearest Neighbor algorithms performed well, achieving above 80% accuracy in classifying PA into different intensity levels and types. Both the ENMO and MIMS metrics proved more accurate than ActiGraph Counts in classifying moderate to vigorous PA. Conclusions: This study provides evidence that both ENMO and MIMS metrics can accurately measure PA with accelerometry, and machine learning algorithms can classify the activity into different intensity levels. These metrics and methods are valuable tools for monitoring PA and understanding the relationship between PA and health outcomes.

show abstract

Predicting lying, sitting and walking at different intensities using smartphone accelerometers at three different wear locations: hands, pant pockets, backpack

Cited by 3 publications

References 32 publications

Exploring Orientation Invariant Heuristic Features with Variant Window Length of 1D-CNN-LSTM in Human Activity Recognition

Exploring Orientation Invariant Heuristic Features with Variant Window Length of 1D-CNN-LSTM in Human Activity Recognition

Quantifying the Uncertainty of Human Activity Recognition Using a Bayesian Machine Learning Method: A Prediction Study

Comparing Accelerometer Processing Metrics and Hyperparameter Optimization for Physical Activity Classification Accuracy Using Machine Learning Methods

Contact Info

Product

Resources

About