Context-Aware Activity Recognition Systems

Smartphone Placement Recognition during Walking: Performance Determinants and Real-World Generalizability

This paper develops and validates a classifier to recognize smartphone placement during walking, showing improved accuracy when aggregating pocket placements and highlighting performance variability across conditions.

Towards Continuous Home Monitoring for Dementia: A Real-Time mmWave Radar Framework for Activity Classification and Tracking

This paper introduces a real-time mmWave radar framework for activity classification and tracking in dementia home monitoring, achieving high accuracy in detecting stationary, walking, and absent states.

Validation and optimisation of wearable accelerometer data pre-processing for digital measure implementation and development

This paper validates and optimizes a modular pipeline for preprocessing wearable accelerometer data to ensure quality and traceability in digital health measures.

Edge-First Ground Reaction Force Estimation with Consumer Smartwatches

This paper develops a wearable system using smartwatches to estimate ground reaction force, achieving high correlation with minimal error in healthy participants.

Continuous Monitoring of Head Turns: Compliance, Kinematics, and Reliability of Wearable Sensing

This study evaluates the feasibility of continuous head turn monitoring using wearable sensors, finding high reliability and establishing normative data in healthy adults.

Human motion data expansion from arbitrary sparsesensors with shallow recurrent decoders

This paper develops a deep learning method to reconstruct dense human motion data from sparse sensors, enabling accurate tracking of unmonitored body segments and improving digital biomarker estimates.

Accuracy and Reliability of a Suite of Digital Measures of Walking Generated Using a Wrist-worn Sensor: Performance Characterization Study

The paper evaluates the accuracy and reliability of digital walking measures from a wrist-worn sensor using real-world data and reference devices.

Rapid dynamic naturalistic monitoring of bradykinesia in Parkinson's disease using a wrist-worn accelerometer

This study uses wrist-worn accelerometers to classify bradykinesia fluctuations in Parkinson's patients at the single-minute timescale using individual and group models.

Three-level Sleep Stage Classification Based on Wrist-worn Accelerometry Data Alone

This paper develops an LSTM model to classify sleep stages using wrist-worn accelerometry data alone, achieving 65-81% accuracy with small training samples.

Statistical and machine learning models for classification of human wear and delivery days in accelerometry data

This paper develops models to classify accelerometry data as human wear or delivery days, finding that a hybrid convolutional recurrent neural network performs best.

A Neural Network Based Algorithm for Dynamically Adjusting Activity Targets to Sustain Exercise Engagement Among People Using Activity Trackers

This paper develops a machine learning model to dynamically adjust activity targets for users of activity trackers to improve adherence by personalizing goals based on historical data and environmental factors.

CoRhythMo: A Computational Framework for Modeling Biobehavioral Rhythms from Mobile and Wearable Data Streams

CoRhythMo models biobehavioral rhythms from mobile and wearable data to assess and predict health outcomes.

Thermal sensors improve wrist-worn position tracking

This paper demonstrates that adding thermal sensors to wrist-worn devices improves position tracking accuracy for head locations, with potential applications in monitoring body-focused repetitive behaviors.

Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 UK Biobank participants

The paper develops a machine learning model to classify physical activity and sleep behaviors from sensor data in UK Biobank participants, achieving 87% accuracy.

Assessing actimeters for inclusion in the Healthy Brain Network

The paper evaluates actimeters for inclusion in the Healthy Brain Network, finding the ActiGraph wGT3X-BT most suitable despite no clear overall superiority among devices.