Research

• Few-Shot Learning for Video Anomaly Detection [paper]

  • To overcome challenges of
    1) Lengthy surveillance videos
    2) Variability in terms of spatial and temporal information
    3) Unseen and unexpected actions

Automatically Measure Response Time of Pharyngeal Swallowing Reflex

Measuring the response time of the pharyngeal swallowing reflex is labor-intensive. 

Our novel framework detects the response time during the pharyngeal swallowing reflex automatically. 

This framework can be a clinically useful tool for (1) estimating the absence or delayed response time of the swallowing reflex in patients with dysphagia and
(2) improving the poor inter-rater reliability of pharyngeal swallowing reflex response time evaluation between expert and unskilled clinicians.

Classification of Dependence in Ambulation in Stroke Patients using Smartphone Videos

Clinicians or physiotherapists look at the person's mobility and balance to figure out how dependent they are on walking after a stroke. The mobility function is commonly used to assess the required dependence or assistance.

Patients with disabilities should follow a rehabilitation plan and keep an eye on dangerous situations in the community, such as the risk of falling. In this study, we use a deep neural network to classify whether a disabled stroke patient needs help walking by using video data from an inpatient rehabilitation therapy session taken with a smartphone.