A5082995677- Anomaly Detection Techniques and Applications
- Human Pose and Action Recognition
- Parkinson's Disease Mechanisms and Treatments
- Multimodal Machine Learning Applications
- Sports Analytics and Performance
- Dementia and Cognitive Impairment Research
- Neurological disorders and treatments
- Sports Performance and Training
- Context-Aware Activity Recognition Systems
- Voice and Speech Disorders
- Technology Use by Older Adults
- Advanced Sensor and Energy Harvesting Materials
- Mobile Health and mHealth Applications
- Balance, Gait, and Falls Prevention
- Video Analysis and Summarization
- Neural Networks and Applications
- Cerebral Palsy and Movement Disorders
- Video Surveillance and Tracking Methods
University of Bristol
2018-2022
Introduction The impact of disease-modifying agents on disease progression in Parkinson’s is largely assessed clinical trials using rating scales. These scales have drawbacks terms their ability to capture the fluctuating nature symptoms while living a naturalistic environment. SPHERE (Sensor Platform for HEalthcare Residential Environment) project has designed multi-sensor platform with multimodal devices allow continuous, relatively inexpensive, unobtrusive sensing motor, non-motor and...
Parkinson’s disease (PD) is a chronic neurodegenerative condition that affects patient’s everyday life. Authors have proposed machine learning and sensor-based approach continuously monitors patients in naturalistic settings can provide constant evaluation of PD objectively analyse its progression. In this paper, we make progress toward such by presenting multimodal deep for discriminating between people with without PD. Specifically, our architecture, named MCPD-Net, uses two data...
In interdisciplinary spaces such as digital health, datasets that are complex to collect, require specialist facilities, and/or collected with specific populations have value in a range of different sectors. this study we simulated free-living dataset, smart home, 12 participants (six people Parkinson's, six carers). We explored their initial perceptions the sensors through interviews and then conducted two data exploration workshops, wherein showed discussed views on how data, other...
Monitoring the progression of an action towards completion offers fine grained insight into actor's behaviour. In this work, we target detecting moment actions, that is when action's goal has been successfully accomplished. This potential applications from surveillance to assistive living and human-robot interactions. Previous effort required human annotations for training (i.e. full supervision). present approach detection weak video-level labels. Given both complete incomplete sequences,...
Action completion detection is the problem of modelling action's progression towards localising moment - when goal confidently considered achieved. In this work, we assess ability two temporal models, namely Hidden Markov Models (HMM) and Long-Short Term Memory (LSTM), to localise for six object interactions: switch, plug, open, pull, pick drink. We use a supervised approach, where annotations pre-completion post-completion frames are available per action, fine-tuned CNN features used train...
This paper looks to explore the challenges faced when producing a set of annotations from videos produced by pilot study evaluating 24 participants (12 with Parkinson's disease, each accompanied healthy volunteer control participant) who are free-living in house embedded platform sensors. We discuss outcome measures chosen annotate and controlled vocabularies formulated for this task, tools processes, how we intend achieve standardisation normalisation annotations, improve quality...
We introduce completion moment detection for actions - the problem of locating completion, when action's goal is confidently considered achieved. The paper proposes a joint classification-regression recurrent model that predicts from given frame, and then integrates frame-level contributions to detect sequence-level moment. voting node frame's relative position by either classification or regression. method also capable detecting incompletion. For example, missed ball-catch, as well at which...
Monitoring the progression of an action towards completion offers fine grained insight into actor's behaviour. In this work, we target detecting moment actions, that is when action's goal has been successfully accomplished. This potential applications from surveillance to assistive living and human-robot interactions. Previous effort required human annotations for training (i.e. full supervision). present approach detection weak video-level labels. Given both complete incomplete sequences,...