A Comparison of Feature Extraction Methods for the Classification of Dynamic Activities From Accelerometer Data

By Stephen J. Preece, John Yannis Goulermas, Laurence P. J. Kenney, David Howard. IEEE Transactions on Biomedical Engineering, 2009

570 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition, Physical Therapy, Sports Therapy and Rehabilitation

View Paper: https://www.semanticscholar.org/paper/8b0145634e19ca5453b15b962c0c282537aa9eb4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Computer Vision and Pattern Recognition

In Physical Therapy, Sports Therapy and Rehabilitation

Citations in the corpus

A Survey on Ambient-Assisted Living Tools for Older Adults (2013, unknown): “coefficients, or wavelet features [62], [63].”
Activity Recognition Using Inertial Sensing for Healthcare, Wellbeing and Sports Applications: A Survey (2010, background): “They are mainly used to detect the transition between different activities[70].”
Optimal Placement of Accelerometers for the Detection of Everyday Activities (2013, background): “[10] addressed the area of dynamic activity recognition and the specific challenge of extracting relevant features from the accelerometer signal.”
Comparison of Data Preprocessing Approaches for Applying Deep Learning to Human Activity Recognition in the Context of Industry 4.0 (2018, background): “This will cause a delay in response for real-time applications [23] and restrict the application scenarios. recognition accuracy, but the In most of the existing studies, segments of 1 to 10 s are considered for HAR [29]. time will be increased and more time will be sufﬁcient data.”
Smart Devices are Different: Assessing and MitigatingMobile Sensing Heterogeneities for Activity Recognition (2015, background): “selected activity classes (or a subset of them) have been investigated prominently in a large number of HAR publications [10, 11, 14, 27, 29, 41, 47].”
Classification Accuracies of Physical Activities Using Smartphone Motion Sensors (2012, methodology): “As discussed above, different window sizes have been used in the literature, including 1 second [3,9,15], 2 seconds [20], and 10 seconds [11].”
Comparative study on classifying human activities with miniature inertial and magnetic sensors (2010, methodology): “Because the initial set of features was quite large (1170) and not all features were equally useful in discriminating between the activities, we investigated different feature selection and reduction methods [55].”
Human Activity Recognition Using Inertial/Magnetic Sensor Units (2010, result): “Furthermore, time-domain features are selected more often than frequency-domain features, as also confirmed in a previous study [31].”
Physical Human Activity Recognition Using Wearable Sensors (2015, background): “For more information the reader is referred to [40,45].”
On-body localization of wearable devices: An investigation of position-aware activity recognition (2016, background): “Nevertheless, some researchers have compared different groups of features and also stated that frequency-based features improve the accuracy of the recognition [16].”
The Promise of mHealth: Daily Activity Monitoring and Outcome Assessments by Wearable Sensors (2011, background): “Once a person’s template has been defined, further data collections during everyday activities are evaluated by the MDAWN system for periods of activity with features that match it.(5,103,104) Sensor-derived algorithms can potentially be trained to recognize most real-world daily activity patterns, ”
Gait Analysis Using Wearable Sensors (2012, methodology): “Various feature extraction methods for the classification of dynamic activities from accelerometer data were compared based on two datasets of activities collected from 20 subjects [94].”
Window Size Impact in Human Activity Recognition (2014, background): “The remaining activities ( i.e. , 9–17, 19–22, 24, 25, 27, and 30), here classiﬁed as less-or non-energetic, are best recognized for W max perf between 1.25 and 5.75 s, while the minimum window size W min size ranges between 0.25 and 2 s.”
Fusion of Smartphone Motion Sensors for Physical Activity Recognition (2014, unknown): “We used the sliding window approach with an overlap of 50% (1 s overlap here), based on the reporting of previous studies [3,34,35].”