Bipart: Learning Block Structure for Activity Detection

Abstract

Physical activity consists complex behavior, typically structured in bouts which can consist of one continuous movement (e.g. exercise) or many sporadic movements (e.g. household chores). Each bout can be represented as a block of feature vectors corresponding to the same activity type. This paper introduces a general distance metric technique to use this block representation to first predict activity type, and then uses the predicted activity to estimate energy expenditure within a novel framework. This distance metric, dubbed Bipart, learns block-level information from both training and test sets, combining both to form a projection space which materializes block-level constraints. Thus, Bipart provides a space which can improve the bout classification performance of all classifiers. We also propose an energy expenditure estimation framework which leverages activity classification in order to improve estimates. Comprehensive experiments on waist-mounted accelerometer data, comparing Bipart against many similar methods as well as other classifiers, demonstrate the superior activity recognition of Bipart, especially in low-information experimental settings.

Estimating physical activity in youth using an ankle accelerometer

“…For example, Trost et al (Trost, Wong, Pfeiffer, & Zheng, 2012), used an artificial neural network for estimating MET y in youth wearing an ActiGraph GT1M while performing 12 structured activities and found a RMSE of 0.9 MET y . In contrast, Mu et al (Mu, Lo, Ding, Amaral, & Crouter, 2014) compared several machine learning approaches, including Bipart and an artificial neural network, for predicting MET y in youth wearing an ActiGraph GT3X while performing 18 structured activities and found a RMSE of 1.37 MET y and 1.39 MET y . However, these machine learning models have not been evaluated in youth in a free-living setting where all models typically perform worse than in a lab-based setting so it is unclear if using machine learning approaches are superior for decreasing group estimates in individual error when predicting MET y with accelerometers.…”

υ-Support vector machine based on discriminant sparse neighborhood preserving embedding

“…By substituting (18,19,20,21) into (17), the constraint optimization problem (11) is reformulated as the dual problem in (13,14,15,16). The proof is now completed.…”

“…LPP not only can preserve the local manifold structure of the dataset, but also can obtain a linear projection on new test samples for dimensionality reduction. Local distance metric learning is used for classification [18,19]. Usually, the local manifold structure of the data can be captured by an adjacency graph.…”

A Survey on Energy Expenditure Estimation Using Wearable Devices

“…State-of-the-art of AR is out of the scope of this paper, but it is more advanced than EEE because there are several open datasets [31,73,138], competitions [15,54], tutorials [26] and surveys [74,127]. Different approaches have been developed and accuracy is normally in the range [80][81][82][83][84][85][86][87][88][89][90][91][92][93][94][95]% when the number of activities to recognize is less than 15 [127] depending also on the modality and number of sensors.…”