BACKGROUND: The more accurate we can assess human physical behaviour in free-living conditions the better we can understand its relationship with health and wellbeing. Thigh-worn accelerometry can be used to identify basic activity types as well as different postures with high accuracy. User-friendly software without the need for specialized programming may support the adoption of this method. This study aims to evaluate the classification accuracy of two novel no-code classification methods, namely SENS motion and ActiPASS.
METHODS: A sample of 38 healthy adults (30.8 ± 9.6 years; 53% female) wore the SENS motion accelerometer (12.5 Hz; ±4 g) on their thigh during various physical activities. Participants completed standardized activities with varying intensities in the laboratory. Activities included walking, running, cycling, sitting, standing, and lying down. Subsequently, participants performed unrestricted free-living activities outside of the laboratory while being video-recorded with a chest-mounted camera. Videos were annotated using a predefined labelling scheme and annotations served as a reference for the free-living condition. Classification output from the SENS motion software and ActiPASS software was compared to reference labels.
RESULTS: A total of 63.6 h of activity data were analysed. We observed a high level of agreement between the two classification algorithms and their respective references in both conditions. In the free-living condition, Cohen\'s kappa coefficients were 0.86 for SENS and 0.92 for ActiPASS. The mean balanced accuracy ranged from 0.81 (cycling) to 0.99 (running) for SENS and from 0.92 (walking) to 0.99 (sedentary) for ActiPASS across all activity types.
CONCLUSIONS: The study shows that two available no-code classification methods can be used to accurately identify basic physical activity types and postures. Our results highlight the accuracy of both methods based on relatively low sampling frequency data. The classification methods showed differences in performance, with lower sensitivity observed in free-living cycling (SENS) and slow treadmill walking (ActiPASS). Both methods use different sets of activity classes with varying definitions, which may explain the observed differences. Our results support the use of the SENS motion system and both no-code classification methods.

BACKGROUND: The accuracy of movement determination software in current activity trackers is insufficient for scientific applications, which are also not open-source.
OBJECTIVE: To address this issue, we developed an accurate, trainable, and open-source smartphone-based activity-tracking toolbox that consists of an Android app (HumanActivityRecorder) and 2 different deep learning algorithms that can be adapted to new behaviors.
METHODS: We employed a semisupervised deep learning approach to identify the different classes of activity based on accelerometry and gyroscope data, using both our own data and open competition data.
RESULTS: Our approach is robust against variation in sampling rate and sensor dimensional input and achieved an accuracy of around 87% in classifying 6 different behaviors on both our own recorded data and the MotionSense data. However, if the dimension-adaptive neural architecture model is tested on our own data, the accuracy drops to 26%, which demonstrates the superiority of our algorithm, which performs at 63% on the MotionSense data used to train the dimension-adaptive neural architecture model.
CONCLUSIONS: HumanActivityRecorder is a versatile, retrainable, open-source, and accurate toolbox that is continually tested on new data. This enables researchers to adapt to the behavior being measured and achieve repeatability in scientific studies.

UNASSIGNED: Hip-worn accelerometers are commonly used, but data processed using the 100 counts per minute cut point do not accurately measure sitting patterns. We developed and validated a model to accurately classify sitting and sitting patterns using hip-worn accelerometer data from a wide age range of older adults.
UNASSIGNED: Deep learning models were trained with 30-Hz triaxial hip-worn accelerometer data as inputs and activPAL sitting/nonsitting events as ground truth. Data from 981 adults aged 35-99 years from cohorts in two continents were used to train the model, which we call CHAP-Adult (Convolutional Neural Network Hip Accelerometer Posture-Adult). Validation was conducted among 419 randomly selected adults not included in model training.
UNASSIGNED: Mean errors (activPAL - CHAP-Adult) and 95% limits of agreement were: sedentary time -10.5 (-63.0, 42.0) min/day, breaks in sedentary time 1.9 (-9.2, 12.9) breaks/day, mean bout duration -0.6 (-4.0, 2.7) min, usual bout duration -1.4 (-8.3, 5.4) min, alpha .00 (-.04, .04), and time in ≥30-min bouts -15.1 (-84.3, 54.1) min/day. Respective mean (and absolute) percent errors were: -2.0% (4.0%), -4.7% (12.2%), 4.1% (11.6%), -4.4% (9.6%), 0.0% (1.4%), and 5.4% (9.6%). Pearson\'s correlations were: .96, .92, .86, .92, .78, and .96. Error was generally consistent across age, gender, and body mass index groups with the largest deviations observed for those with body mass index ≥30 kg/m2.
UNASSIGNED: Overall, these strong validation results indicate CHAP-Adult represents a significant advancement in the ambulatory measurement of sitting and sitting patterns using hip-worn accelerometers. Pending external validation, it could be widely applied to data from around the world to extend understanding of the epidemiology and health consequences of sitting.

Although machine learning techniques have been repeatedly used for activity prediction from wearable devices, accurate classification of 24-hour activity behaviour categories from accelerometry data remains a challenge. We developed and validated a deep learning-based framework for classifying 24-hour activity behaviours from wrist-worn accelerometers.
Using an openly available dataset with free-living wrist-based raw accelerometry data from 151 participants (aged 18-91 years), we developed a deep learning framework named AccNet24 to classify 24-hour activity behaviours. First, the acceleration signal (x, y, and z-axes) was segmented into 30-second nonoverlapping windows, and signal-to-image conversion was performed for each segment. Deep features were automatically extracted from the signal images using transfer learning and transformed into a lower-dimensional feature space. These transformed features were then employed to classify the activity behaviours as sleep, sedentary behaviour, and light-intensity (LPA) and moderate-to-vigorous physical activity (MVPA) using a bidirectional long short-term memory (BiLSTM) recurrent neural network. AccNet24 was trained and validated with data from 101 and 25 randomly selected participants and tested with the remaining unseen 25 participants. We also extracted 112 hand-crafted time and frequency domain features from 30-second windows and used them as inputs to five commonly used machine learning classifiers, including random forest, support vector machines, artificial neural networks, decision tree, and naïve Bayes to classify the 24-hour activity behaviour categories.
Using the same training, validation, and test data and window size, the classification accuracy of AccNet24 outperformed the accuracy of the other five machine learning classification algorithms by 16%-30% on unseen data.
AccNet24, relying on signal-to-image conversion, deep feature extraction, and BiLSTM achieved consistently high accuracy (>95 %) in classifying the 24-hour activity behaviour categories as sleep, sedentary, LPA, and MVPA. The next generation accelerometry analytics may rely on deep learning techniques for activity prediction.

The employment of machine learning algorithms to the data provided by wearable movement sensors is one of the most common methods to detect pets\' behaviors and monitor their well-being. However, defining features that lead to highly accurate behavior classification is quite challenging. To address this problem, in this study we aim to classify six main dog activities (standing, walking, running, sitting, lying down, and resting) using high-dimensional sensor raw data. Data were received from the accelerometer and gyroscope sensors that are designed to be attached to the dog\'s smart costume. Once data are received, the module computes a quaternion value for each data point that provides handful features for classification. Next, to perform the classification, we used several supervised machine learning algorithms, such as the Gaussian naïve Bayes (GNB), Decision Tree (DT), K-nearest neighbor (KNN), and support vector machine (SVM). In order to evaluate the performance, we finally compared the proposed approach\'s F-score accuracies with the accuracy of classic approach performance, where sensors\' data are collected without computing the quaternion value and directly utilized by the model. Overall, 18 dogs equipped with harnesses participated in the experiment. The results of the experiment show a significantly enhanced classification with the proposed approach. Among all the classifiers, the GNB classification model achieved the highest accuracy for dog behavior. The behaviors are classified with F-score accuracies of 0.94, 0.86, 0.94, 0.89, 0.95, and 1, respectively. Moreover, it has been observed that the GNB classifier achieved 93% accuracy on average with the dataset consisting of quaternion values. In contrast, it was only 88% when the model used the dataset from sensors\' data.

In this work, a novel method is presented for non-contact non-invasive physical activity monitoring, which utilizes a multi-axial inertial measurement unit (IMU) to measure activity-induced structural vibrations in multiple axes. The method is demonstrated in monitoring the activity of a mouse in a husbandry cage, where activity is classified as resting, stationary activity and locomotion. In this setup, the IMU is mounted in the center of the underside of the cage floor where vibrations are measured as accelerations and angular rates in the X-, Y- and Z-axis. The ground truth of activity is provided by a camera mounted in the cage lid. This setup is used to record 27.67 h of IMU data and ground truth activity labels. A classification model is trained with 16.17 h of data which amounts to 3880 data points. Each data point contains eleven features, calculated from the X-, Y- and Z-axis accelerometer data. The method achieves over 90% accuracy in classifying activity versus non-activity. Activity is monitored continuously over more than a day and clearly depicts the nocturnal behavior of the inhabitant. The impact of this work is a powerful method to assess activity which enables automatic health evaluation and optimization of workflows for improved animal wellbeing.

The development of smartphones technologies has determined the abundant and prevalent computation. An activity recognition system using mobile sensors enables continuous monitoring of human behavior and assisted living. This paper proposes the mobile sensors-based Epidemic Watch System (EWS) leveraging the AI models to recognize a new set of activities for effective social distance monitoring, probability of infection estimation, and COVID-19 spread prevention. The research focuses on user activities recognition and behavior concerning risks and effectiveness in the COVID-19 pandemic. The proposed EWS consists of a smartphone application for COVID-19 related activities sensors data collection, features extraction, classifying the activities, and providing alerts for spread presentation. We collect the novel dataset of COVID-19 associated activities such as hand washing, hand sanitizing, nose-eyes touching, and handshaking using the proposed EWS smartphone application. We evaluate several classifiers such as random forests, decision trees, support vector machine, and Long Short-Term Memory for the collected dataset and attain the highest overall classification accuracy of 97.33%. We provide the Contact Tracing of the COVID-19 infected person using GPS sensor data. The EWS activities monitoring, identification, and classification system examine the infection risk of another person from COVID-19 infected person. It determines some everyday activities between COVID-19 infected person and normal person, such as sitting together, standing together, or walking together to minimize the spread of pandemic diseases.

OBJECTIVE: Semantic segmentation and activity classification are key components to create intelligent surgical systems able to understand and assist clinical workflow. In the operating room, semantic segmentation is at the core of creating robots aware of clinical surroundings, whereas activity classification aims at understanding OR workflow at a higher level. State-of-the-art semantic segmentation and activity recognition approaches are fully supervised, which is not scalable. Self-supervision can decrease the amount of annotated data needed.
METHODS: We propose a new 3D self-supervised task for OR scene understanding utilizing OR scene images captured with ToF cameras. Contrary to other self-supervised approaches, where handcrafted pretext tasks are focused on 2D image features, our proposed task consists of predicting relative 3D distance of image patches by exploiting the depth maps. By learning 3D spatial context, it generates discriminative features for our downstream tasks.
RESULTS: Our approach is evaluated on two tasks and datasets containing multiview data captured from clinical scenarios. We demonstrate a noteworthy improvement in performance on both tasks, specifically on low-regime data where utility of self-supervised learning is the highest.
CONCLUSIONS: We propose a novel privacy-preserving self-supervised approach utilizing depth maps. Our proposed method shows performance on par with other self-supervised approaches and could be an interesting way to alleviate the burden of full supervision.

Improper manual material handling (MMH) techniques are shown to lead to low back pain, the most common work-related musculoskeletal disorder. Due to the complex nature and variability of MMH and obtrusiveness and subjectiveness of existing hazard analysis methods, providing systematic, continuous, and automated risk assessment is challenging. We present a machine learning algorithm to detect and classify MMH tasks using minimally-intrusive instrumented insoles and chest-mounted accelerometers. Six participants performed standing, walking, lifting/lowering, carrying, side-to-side load transferring (i.e., 5.7 kg and 12.5 kg), and pushing/pulling. Lifting and carrying loads as well as hazardous behaviors (i.e., stooping, overextending and jerky lifting) were detected with 85.3%/81.5% average accuracies with/without chest accelerometer. The proposed system allows for continuous exposure assessment during MMH and provides objective data for use with analytical risk assessment models that can be used to increase workplace safety through exposure estimation.

UNASSIGNED: Machine learning has been used for classification of physical behavior bouts from hip-worn accelerometers; however, this research has been limited due to the challenges of directly observing and coding human behavior \"in the wild.\" Deep learning algorithms, such as convolutional neural networks (CNNs), may offer better representation of data than other machine learning algorithms without the need for engineered features and may be better suited to dealing with free-living data. The purpose of this study was to develop a modeling pipeline for evaluation of a CNN model on a free-living data set and compare CNN inputs and results with the commonly used machine learning random forest and logistic regression algorithms.
UNASSIGNED: Twenty-eight free-living women wore an ActiGraph GT3X+accelerometer on their right hip for 7 days. A concurrently worn thigh-mounted activPAL device captured ground truth activity labels. The authors evaluated logistic regression, random forest, and CNN models for classifying sitting, standing, and stepping bouts. The authors also assessed the benefit of performing feature engineering for this task.
UNASSIGNED: The CNN classifier performed best (average balanced accuracy for bout classification of sitting, standing, and stepping was 84%) compared with the other methods (56% for logistic regression and 76% for random forest), even without performing any feature engineering.
UNASSIGNED: Using the recent advancements in deep neural networks, the authors showed that a CNN model can outperform other methods even without feature engineering. This has important implications for both the model\'s ability to deal with the complexity of free-living data and its potential transferability to new populations.