Activity Classification Model Research Articles

A novel hybrid deep learning approach with GWO–WOA optimization technique for human activity recognition

The effectiveness of Human Activity Recognition (HAR) models can be largely attributed to the components derived from domain expertise. The classification system swiftly and effectively categorizes human physical activity by utilizing a comprehensive collection of variables. To construct HAR models and categorize different activities, deep learning algorithms have recently seen increased application in academic research for autonomously extracting features from raw sensory data. The primary goal of this study is to develop a strong and accurate activity classification model by first extracting features from an input dataset and then using deep learning models. In this research, we look at how well the model does in the field of HAR. Analyzing time-series data from smartphones and wearable sensors, algorithms based on deep learning (DL) have been used to forecast various human actions. Applying DL-based methods to time-series data presents a variety of challenges, notwithstanding their usage for activity identification. If the suggested approach is put into practice, the concerns mentioned before may be alleviated. To construct HAR classification techniques using data collected from wearable sensors, this study presents two Hybrid Learning Algorithms (HLA). This research employs Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) methods to teach the model sequential and spatial patterns. The process of selecting characteristics has been enhanced through the use of optimization techniques. Together, the Whale Optimization Algorithm and the Grey Wolf Optimizer form a Hybrid Optimization approach, which is introduced in this study. By combining and capitalizing the benefits of both approaches, the hybrid technique enhances the optimization process as a whole. This research may also assist individuals who have undergone amputations or experienced strokes in selecting an appropriate model and developing a personalized gait reference trajectory.

Composite activity type and stride-specific energy expenditure estimation model for thigh-worn accelerometry

BackgroundAccurately measuring energy expenditure during physical activity outside of the laboratory is challenging, especially on a large scale. Thigh-worn accelerometers have gained popularity due to the possibility to accurately detect physical activity types. The use of machine learning techniques for activity classification and energy expenditure prediction may improve accuracy over current methods. Here, we developed a novel composite energy expenditure estimation model by combining an activity classification model with a stride specific energy expenditure model for walking, running, and cycling.MethodsWe first trained a supervised deep learning activity classification model using pooled data from available adult accelerometer datasets. The composite energy expenditure model was then developed and validated using additional data based on a sample of 69 healthy adult participants (49% female; age = 25.2 ± 5.8 years) who completed a standardised activity protocol with indirect calorimetry as the reference measure.ResultsThe activity classification model showed an overall accuracy of 99.7% across all five activity types during validation. The composite model for estimating energy expenditure achieved a mean absolute percentage error of 10.9%. For running, walking, and cycling, the composite model achieved a mean absolute percentage error of 6.6%, 7.9% and 16.1%, respectively.ConclusionsThe integration of thigh-worn accelerometers with machine learning models provides a highly accurate method for classifying physical activity types and estimating energy expenditure. Our novel composite model approach improves the accuracy of energy expenditure measurements and supports better monitoring and assessment methods in non-laboratory settings.

Developmental trends in young children’s device-measured physical activity and sedentary behaviour

BackgroundKnowledge of developmental trends in meeting age-specific 24-hour movement behaviour guidelines is lacking. This study describes developmental trends in device-measured physical activity and sedentary time over a three-year period among Western Australian children aged two to seven years, including differences between boys and girls. The proportion of children meeting age-specific physical activity guidelines before and after they transition to full-time school was also examined.MethodsData from waves 1 and 2 of the Play Spaces and Environments for Children’s Physical Activity (PLAYCE) cohort study were used (analysis n = 1217). Physical activity and sedentary time were measured by accelerometry at ages two to five (preschool, wave 1) and ages five to seven (commenced full-time school, wave 2). Accelerometer data were processed using a validated machine-learning physical activity classification model. Daily time spent in sedentary behaviour, energetic play (moderate-to-vigorous physical activity (MVPA)), total physical activity, and meeting physical activity guidelines were analysed using linear and generalised linear mixed-effects models with age by sex interaction terms.ResultsAll movement behaviours changed significantly with increasing age, and trends were similar in boys and girls. Total daily physical activity increased from age two to five then declined to age seven. Mean daily total physical activity exceeded 180 min/day from ages two to five. Daily energetic play increased significantly from age two to seven, however, was below 60 min/day at all ages except for seven-year-old boys. Daily sedentary time decreased to age five then increased to age seven but remained lower than at age two. All two-year-olds met their age-specific physical activity guideline, decreasing to 5% of girls and 6% of boys at age four. At age seven, 46% of boys and 35% of girls met their age-specific physical activity guideline.ConclusionsYoung children’s energetic play and total physical activity increased with age, but few children aged three to seven met the energetic play (MVPA) guideline. Interventions should focus on increasing children’s energetic play in early childhood. Clearer guidance and strategies are needed to support young children as they change developmentally and as they transition from one age-specific movement guideline to the next.

Associations between natural blue and green space and preschool children's movement behaviours

Abstract Natural outdoor environments provide important settings for children to engage in physical activity. It is unclear if these relationships also exist in preschool aged children. We examined associations between amount and type of neighbourhood vegetation and blue space, proximity to the beach and preschooler's device‐measured movement behaviours. Accelerometer data from 1081 preschool children were processed into physical activity and sedentary behaviours using a machine learned random forest physical activity classification model specifically developed for children under five. High‐resolution aerial imagery data and Geographic Information Systems were used to identify total percentage of neighbourhood vegetation and vegetation components (grass, shrubs and trees) and presence and total percentage of blue space (ocean, riverways and lakes and swamps) for 500, 1600 and 5000 m service areas around children's residences. Distance to the closest beach, patrolled beach, and dog beach were also calculated. Amount of neighbourhood vegetation was inversely associated with boys sedentary time in the 500 m (β = −0.91 min/day, p = 0.006, 95% CI [−1.55, −0.33]) and 1600 m (β = −1.30 min/day, p = 0.001, 95% CI [−2.08, −0.53]) service area, and positively associated with walking time in the 1600 m (β = 0.12 min/day, p = 0.045, 95% CI [0.02, 0.22]) service area. The presence of local riverways was positively associated with girls walking time in the 1600 m (β = 2.34 min/day, p = 0.009, 95% CI [0.59, 4.09]), and 5000 m (β = 1.67 min/day, p = 0.019, 95% CI [0.27, 3.06]) buffer. Distance to dog beach was positively associated with boys sedentary behaviours. Presence of lake and swamps were negatively associated with boys walking, and presence of riverways was negatively associated with girls moderate‐to‐vigorous activities and games and boys running in the 5000 m service area. These findings suggest higher amounts of neighbourhood vegetation and certain types of blue space such as riverways and dog beaches could facilitate positive movement behaviours in young children. Our findings that natural outdoor environments are associated with preschool children's movement behaviours add to evidence which could contribute to policy implications for urban planning and the provision and protection of natural environments within cities. Read the free Plain Language Summary for this article on the Journal blog.

Performance evaluation of an occupant metabolic rate estimation algorithm using activity classification and object detection models

To create a comfortable indoor environment, the metabolic rate (MET), which affects the thermal sensation of occupants, needs to be reflected in real-time. Recently, methods employing computer vision techniques classify activities based on the pose of the body in images. However, these methods face challenges in determining the MET depending on the objects used, even with the same pose. Therefore, the objective of this study is to develop a MET estimation algorithm that can estimate various METs by integrating a pose-based activity classification model and an object detection model. To achieve this, an object detection model capable of detecting and classifying six regularly used objects indoors was developed, and a performance evaluation was conducted. The MET estimation algorithm was assessed through the implementation of a thermal control system, validating its applicability in experimental settings. As a result, the object detection model exhibited a real-time classification accuracy of 89%. Additionally, when evaluating the mode value over 15-s intervals, it demonstrated a classification accuracy of 100%. The algorithm exhibited a real-time estimation accuracy of 83% for the six METs and examining the mode value for 15-s intervals, it demonstrated a classification accuracy of 99%. This study thus confirmed the control capability of the proposed MET estimation algorithm and its potential for the estimation of various METs. The developed method can be used for the real-time estimation of occupant thermal comfort in indoor comfort-based control systems, contributing to the realization of a comfortable environment for occupants that protects their well-being.

A detailed occupant activity classification model in a residential environment using building monitoring data: Considering occupant characteristics

In this study, we developed a detailed occupant activity classification model to detect occupant activities using various non-intrusive monitoring data in buildings (i.e., indoor environmental data, energy consumption data, motion·door sensor data, and state of indoor climate control devices). To collect building monitoring data, we conducted experiments with five subjects (Person A to E) in an experimental testbed. In total, ten occupant activities (i.e., Sleeping, Eating, Personal hygiene and care, Working, Cooking, Cleaning up after meals, Indoor cleaning, Exercising, Resting, and Away) were selected to predict occupant activity experimentally. Three classification algorithms, including random forest (RF), K-nearest neighbor, and support vector machine, were employed to develop the detailed occupancy activity classification models. In this study, we found that the key variables influencing each occupant activity classification varied significantly for each individual. The lifestyle and characteristics of each occupant were reflected in the building monitoring data. Accordingly, the performance of the model varied depending on each occupant’s behavior. In the case of Person A, all classification algorithms displayed a high performance when only environmental datasets were considered; the accuracy was 0.98, and the f1-score was also above 0.98. Person B’s classification model required various building monitoring data to ensure optimal performance; the highest accuracy and f1-score was 0.86 using RF. In contrast, Person D’s classification model showed a low performance when all monitoring datasets were used; the accuracy was 0.43, and the f1-score was 0.41 using RF. These results support that occupants’ characteristics affect the feature variables in an activity classification model, and these feature variables determine the performance of the model of each occupant.

Radar‐based human activity recognition using denoising techniques to enhance classification accuracy

AbstractRadar‐based human activity recognition is considered as a competitive solution for the elderly care health monitoring problem, compared to alternative techniques such as cameras and wearable devices. However, raw radar signals are often contaminated with noise, clutter, and other artifacts that significantly impact recognition performance, which highlights the importance of prepossessing techniques that enhance radar data quality and improve classification model accuracy. In this study, two different human activity classification models incorporated with pre‐processing techniques have been proposed. The authors introduce wavelet denoising methods into a cyclostationarity‐based classification model, resulting in a substantial improvement in classification accuracy. To address the limitations of conventional pre‐processing techniques, a deep neural network model called Double Phase Cascaded Denoising and Classification Network (DPDCNet) is proposed, which performs end‐to‐end signal‐level classification and achieves state‐of‐the‐art accuracy. The proposed models significantly reduce false detections and would enable robust activity monitoring for older individuals with radar signals, thereby bringing the system closer to a practical implementation for deployment.

Deep learning-based classification model for GPR151 activator activity prediction

BackgroundGPR151 is a kind of protein belonging to G protein-coupled receptor family that is closely associated with a variety of physiological and pathological processes.The potential use of GPR151 as a therapeutic target for the management of metabolic disorders has been demonstrated in several studies, highlighting the demand to explore its activators further. Activity prediction serves as a vital preliminary step in drug discovery, which is both costly and time-consuming. Thus, the development of reliable activity classification model has become an essential way in the process of drug discovery, aiming to enhance the efficiency of virtual screening.ResultsWe propose a learning-based method based on feature extractor and deep neural network to predict the activity of GPR151 activators. We first introduce a new molecular feature extraction algorithm which utilizes the idea of bag-of-words model in natural language to densify the sparse fingerprint vector. Mol2vec method is also used to extract diverse features. Then, we construct three classical feature selection algorithms and three types of deep learning model to enhance the representational capacity of molecules and predict activity label by five different classifiers. We conduct experiments using our own dataset of GPR151 activators. The results demonstrate high classification accuracy and stability, with the optimal model Mol2vec-CNN significantly improving performance across multiple classifiers. The svm classifier achieves the best accuracy of 0.92 and F1 score of 0.76 which indicates promising applications for our method in the field of activity prediction.ConclusionThe results suggest that the experimental design of this study is appropriate and well-conceived. The deep learning-based feature extraction algorithm established in this study outperforms traditional feature selection algorithm for activity prediction. The model developed can be effectively utilized in the pre-screening stage of drug virtual screening.

Towards a telehealth infrastructure supported by machine learning on edge/fog for Parkinson's movement screening

Approximately 10 million people worldwide live with Parkinson's disease (PD), a progressive and incurable neurological movement disorder. Symptomatic treatment is available for PD but requires patients to make periodic clinic visits (2–3 times per year) for symptom assessment. Advanced telehealth technologies can enhance clinical care for PD but warrant an Internet-of-Things-based (IoT) infrastructure that can enable symptom monitoring in out-of-clinic settings, such as homes. In this paper, we present an edge and Fog device-based IoT framework and a machine learning-based telehealth infrastructure that can detect and classify hand movement tasks based on a clinical test (UPDRS) for remote symptom assessment. We used a pair of smart gloves integrated with finger flex sensors, an inertial measurement unit (IMU), and a wireless embedded system (i.e., an edge device) to record the hand movements. The edge device (ESP32) detected the activity on the edge node and transmitted the data to the Fog node for classification. The Fog node (Raspberry pi) hosted the Machine Learning (ML) based activity classification models to classify UPDRS-based hand movement tasks. In this paper, we present the development of edge-fog-supported ML infrastructure. To develop ML models, we utilized the data of 9 participants (five healthy and four people with PD) who performed hand movements tasks while wearing the smart gloves. We developed and tested different classification models such as K-nearest neighbors (KNN), Support Vector Machine (SVM), and Decision Tree (DT) to identify which one fits best for such an edge-fog infrastructure for remote symptom assessment. Results showed that the SVM model outperformed others by giving 94% training, 94% testing, and 93% validation accuracy with a mean inference time of 560 μs on the Fog node. Our preliminary results are promising to further our research in deploying the edge-fog-driven ML-based telehealth infrastructure in real-world settings. • Development of Edge-Fog-supported telehealth IoT infrastructure. • Development and assessment of Machine Learning classifiers for hand movement activity classification. • Deployment and evaluation of the ML classification models on the Fog infrastructure. • Demonstration of how an edge-fog computing architecture can be useful for telehealth applications, including movement assessment, symptom tracking, and symptom management.

Hybrid consensus and k-nearest neighbours (kNN) strategies to classify dual BRD4/PLK1 inhibitors

ABSTRACT A novel decision-making procedure is proposed here for the first time to identify active/inactive and selective/non-selective dual inhibitors using consensus approaches and pools of k-nearest neighbours (kNN) classifications instead of individual models. Dual BRD4/PLK1 inhibition with adequate selectivity is a potential therapeutic strategy for targeting tumour cells in high-risk patients. We report the unique way to identify both active and selective dual BRD4/PLK1 inhibitors using consensus and kNN strategies together with two sources of receptor-based and ligand-based information which are the ranked binding energies of residues and important molecular features, respectively. The results of consensus approaches were compared with the results of individual kNN models. The chemical space similarity was measured using three different distance functions to increase the reliability. All activity and selectivity classification models were validated using cross-validation and y-randomization tests. The outcomes show that consensus approaches can increase the reliability and accuracy of active/inactive or selective/non-selective detections up to 90%. Consensus approaches also reached more balanced values of sensitivity and specificity compared to the individual kNN models because of the compensation in the integration of diverse sources of information.

Human Behavior Recognition Model Based on Feature and Classifier Selection.

With the rapid development of the computer and sensor field, inertial sensor data have been widely used in human activity recognition. At present, most relevant studies divide human activities into basic actions and transitional actions, in which basic actions are classified by unified features, while transitional actions usually use context information to determine the category. For the existing single method that cannot well realize human activity recognition, this paper proposes a human activity classification and recognition model based on smartphone inertial sensor data. The model fully considers the feature differences of different properties of actions, uses a fixed sliding window to segment the human activity data of inertial sensors with different attributes and, finally, extracts the features and recognizes them on different classifiers. The experimental results show that dynamic and transitional actions could obtain the best recognition performance on support vector machines, while static actions could obtain better classification effects on ensemble classifiers; as for feature selection, the frequency-domain feature used in dynamic action had a high recognition rate, up to 99.35%. When time-domain features were used for static and transitional actions, higher recognition rates were obtained, 98.40% and 91.98%, respectively.

Children's Activity Classification for Domestic Risk Scenarios Using Environmental Sound and a Bayesian Network.

Children’s healthcare is a relevant issue, especially the prevention of domestic accidents, since it has even been defined as a global health problem. Children’s activity classification generally uses sensors embedded in children’s clothing, which can lead to erroneous measurements for possible damage or mishandling. Having a non-invasive data source for a children’s activity classification model provides reliability to the monitoring system where it is applied. This work proposes the use of environmental sound as a data source for the generation of children’s activity classification models, implementing feature selection methods and classification techniques based on Bayesian networks, focused on the recognition of potentially triggering activities of domestic accidents, applicable in child monitoring systems. Two feature selection techniques were used: the Akaike criterion and genetic algorithms. Likewise, models were generated using three classifiers: naive Bayes, semi-naive Bayes and tree-augmented naive Bayes. The generated models, combining the methods of feature selection and the classifiers used, present accuracy of greater than 97% for most of them, with which we can conclude the efficiency of the proposal of the present work in the recognition of potentially detonating activities of domestic accidents.

Balancing Data on Deep Learning-Based Proteochemometric Activity Classification.

In silico analysis of biological activity data has become an essential technique in pharmaceutical development. Specifically, the so-called proteochemometric models aim to share information between targets in machine learning ligand–target activity prediction models. However, bioactivity data sets used in proteochemometric modeling are usually imbalanced, which could potentially affect the performance of the models. In this work, we explored the effect of different balancing strategies in deep learning proteochemometric target–compound activity classification models while controlling for the compound series bias through clustering. These strategies were (1) no_resampling, (2) resampling_after_clustering, (3) resampling_before_clustering, and (4) semi_resampling. These schemas were evaluated in kinases, GPCRs, nuclear receptors, and proteases from BindingDB. We observed that the predicted proportion of positives was driven by the actual data balance in the test set. Additionally, it was confirmed that data balance had an impact on the performance estimates of the proteochemometric model. We recommend a combination of data augmentation and clustering in the training set (semi_resampling) to mitigate the data imbalance effect in a realistic scenario. The code of this analysis is publicly available at https://github.com/b2slab/imbalance_pcm_benchmark.

Sensor-data augmentation for human activity recognition with time-warping and data masking

Human activity recognition (HAR) using an accelerometer can provide valuable information for understanding user context. Therefore, several studies have been conducted using deep learning to increase the recognition rate of activity classification. However, the existing dataset that is publicly available for HAR tasks contains limited data. Previous works have applied data augmentation methods that simply transform the entire accelerometer-signal dataset. However, the label of the augmented signal cannot be easily recognized by humans, and the augmentation methods cannot ensure that the label of the signal is preserved. Therefore, we propose a novel data augmentation method that reflects the characteristics of the sensor signal and can preserve the label of the augmented signal by generating partially occluded data of the accelerometer signals. To generate the augmented data, we apply time-warping, which deforms the time-series data in the time direction. We handle jittering effects and subsequently apply data masking to drop out a part of the input signals. We compare the performance of the proposed augmentation method with that of conventional methods by using two public datasets and an activity recognition model based on convolutional neural networks. The experimental results show that the proposed augmentation method improves the recognition rate of the activity classification model, regardless of the dataset. Additionally, the proposed method shows superior performance over conventional methods on the two datasets.

Deep artificial neural network based on environmental sound data for the generation of a children activity classification model.

Children activity recognition (CAR) is a subject for which numerous works have been developed in recent years, most of them focused on monitoring and safety. Commonly, these works use as data source different types of sensors that can interfere with the natural behavior of children, since these sensors are embedded in their clothes. This article proposes the use of environmental sound data for the creation of a children activity classification model, through the development of a deep artificial neural network (ANN). Initially, the ANN architecture is proposed, specifying its parameters and defining the necessary values for the creation of the classification model. The ANN is trained and tested in two ways: using a 70–30 approach (70% of the data for training and 30% for testing) and with a k-fold cross-validation approach. According to the results obtained in the two validation processes (70–30 splitting and k-fold cross validation), the ANN with the proposed architecture achieves an accuracy of 94.51% and 94.19%, respectively, which allows to conclude that the developed model using the ANN and its proposed architecture achieves significant accuracy in the children activity classification by analyzing environmental sound.

In-silico screening of database for finding potential sweet molecules: A combined data and structure based modeling approach

In this study, we present a framework comprises of several independent modules which are built upon data based (structure activity relationship and classification model) and structure (molecular docking) based for identifying possible sweeteners from a vast database of natural molecules.A large database, Universal Natural Products Database (UNPD) consisting of 213,210 compounds was screened using the developed framework. At first, 10,184 molecules structurally similar to the known sweeteners were identified in the database. Further, 1924 molecules from these screened molecules were classified as sweet molecules. The shortlisted 1354 molecules were subjected to ADMET analysis. Finally, 60 molecules were arrived at with no toxicity and acceptable oral bioavailability as potential sweetener candidates. Further, molecular docking of these molecules on sweet taste receptor performed to obtain their binding energy, binding sites and correlation with sweetness index. The developed framework offers a convenient route for fast screening of molecules prior to synthesis and testing.

Machine Learning Models for Classifying Physical Activity in Free-Living Preschool Children.

Machine learning (ML) activity classification models trained on laboratory-based activity trials exhibit low accuracy under free-living conditions. Training new models on free-living accelerometer data, reducing the number of prediction windows comprised of multiple activity types by using shorter windows, including temporal features such as standard deviation in lag and lead windows, and using multiple sensors may improve the classification accuracy under free-living conditions. The objective of this study was to evaluate the accuracy of Random Forest (RF) activity classification models for preschool-aged children trained on free-living accelerometer data. Thirty-one children (mean age = 4.0 ± 0.9 years) completed a 20 min free-play session while wearing an accelerometer on their right hip and non-dominant wrist. Video-based direct observation was used to categorize the children’s movement behaviors into five activity classes. The models were trained using prediction windows of 1, 5, 10, and 15 s, with and without temporal features. The models were evaluated using leave-one-subject-out-cross-validation. The F-scores improved as the window size increased from 1 to 15 s (62.6%–86.4%), with only minimal improvements beyond the 10 s windows. The inclusion of temporal features increased the accuracy, mainly for the wrist classification models, by an average of 6.2 percentage points. The hip and combined hip and wrist classification models provided comparable accuracy; however, both the models outperformed the models trained on wrist data by 7.9 to 8.2 percentage points. RF activity classification models trained with free-living accelerometer data provide accurate recognition of young children’s movement behaviors under real-world conditions.

Machine Learning to Quantify Physical Activity in Children with Cerebral Palsy: Comparison of Group, Group-Personalized, and Fully-Personalized Activity Classification Models.

Pattern recognition methodologies, such as those utilizing machine learning (ML) approaches, have the potential to improve the accuracy and versatility of accelerometer-based assessments of physical activity (PA). Children with cerebral palsy (CP) exhibit significant heterogeneity in relation to impairment and activity limitations; however, studies conducted to date have implemented “one-size fits all” group (G) models. Group-personalized (GP) models specific to the Gross Motor Function Classification (GMFCS) level and fully-personalized (FP) models trained on individual data may provide more accurate assessments of PA; however, these approaches have not been investigated in children with CP. In this study, 38 children classified at GMFCS I to III completed laboratory trials and a simulated free-living protocol while wearing an ActiGraph GT3X+ on the wrist, hip, and ankle. Activities were classified as sedentary, standing utilitarian movements, or walking. In the cross-validation, FP random forest classifiers (99.0–99.3%) exhibited a significantly higher accuracy than G (80.9–94.7%) and GP classifiers (78.7–94.1%), with the largest differential observed in children at GMFCS III. When evaluated under free-living conditions, all model types exhibited significant declines in accuracy, with FP models outperforming G and GP models in GMFCS levels I and II, but not III. Future studies should evaluate the comparative accuracy of personalized models trained on free-living accelerometer data.

Machine Learning Activity Classification Models For Preschool-aged Children: The Need For Free-living Training Data.

Machine learning (ML) classification models trained on laboratory activity trials exhibit poor performance when evaluated under free-living conditions. Training models on free living data, including temporal features such as lead and lag windows, and using shorter sliding windows may improve recognition accuracy under free-living conditions. PURPOSE: To evaluate the accuracy of free-living hip and wrist Random Forest activity classifiers for pre-schoolers trained on features extracted from windows of 1s, 5s, 10s, and 15s. Performance was benchmarked against classifiers trained on laboratory-based data using a 15s window. METHODS: 31 preschool-aged children (4.0 ± 0.9 y) were video recorded during a 20-minute unstructured active play session. Participants wore an accelerometer on their right hip and non-dominant wrist. A bespoke two-stage direct observation system was used to code ground truth activity class and specific activity types occurring within each class. Data from 21 of the children were randomly selected to train the classifiers. Models were trained with and without temporal features and cross-validated in a hold sample of the remaining 10 children with overall and class-level accuracy. RESULTS: Accuracy improved as window size increased from 1 sec (73.5%-77.7%) to 10 sec (82.4%-86.0%); with only minimal improvements observed for 15s windows. Inclusion of lag and lead features increased accuracy by 1.6% to 6.6%, with the largest improvements observed for shorter duration windows (≤10s). Comparatively, the accuracy of the laboratory trained model was 56.9% and 67.5% for wrist and hip, respectively. For a 10s window, training models on free-living data and including temporal features increased recognition of sedentary from 70.6% - 74.4% to 83.3% - 90.4%; light activities and games from 57.5% - 76.9% to 88.6% - 88.8%; walking from 7.5% - 17.5% to 64.1% - 75.0%; and running from 50.0% - 77.8% to 71.4% - 85.7%. There was no improvement in recognition of mod-vig activities and games (56.3% - 62.5%). CONCLUSIONS: Unlike models trained on laboratory activity trials, ML activity classification models for pre-schoolers trained on free-living accelerometer data perform well when evaluated under true free-living conditions. Funding: Australian Research Council Discovery Project Grant: DP150100116

Workplace activity classification from shoe-based movement sensors

BackgroundHigh occupational physical activity is associated with lower health. Shoe-based movement sensors can provide an objective measurement of occupational physical activity in a lab setting but the performance of such methods in a free-living environment have not been investigated. The aim of the current study was to investigate the feasibility and accuracy of shoe sensor-based activity classification in an industrial work setting.ResultsAn initial calibration part was performed with 35 subjects who performed different workplace activities in a structured lab setting while the movement was measured by a shoe-sensor. Three different machine-learning models (random forest (RF), support vector machine and k-nearest neighbour) were trained to classify activities using the collected lab data. In a second validation part, 29 industry workers were followed at work while an observer noted their activities and the movement was captured with a shoe-based movement sensor. The performance of the trained classification models were validated using the free-living workplace data. The RF classifier consistently outperformed the other models with a substantial difference in in the free-living validation. The accuracy of the initial RF classifier was 83% in the lab setting and 43% in the free-living validation. After combining activities that was difficult to discriminate the accuracy increased to 96 and 71% in the lab and free-living setting respectively. In the free-living part, 99% of the collected samples either consisted of stationary activities or walking.ConclusionsWalking and stationary activities can be classified with high accuracy from a shoe-based movement sensor in a free-living occupational setting. The distribution of activities at the workplace should be considered when validating activity classification models in a free-living setting.