Perpetually Under Construction!!

Selected Projects

• Socio-behavioral features differentiating depression and psychosis patients from healthy controls
  
  In this project, we use continuous audio recordings from free-living across multiple days to automatically infer a subject's in-person social network and derive socio-behavioral features relevant to diagnosing and monitoring depression and psychosis.
  
  

• Speaker Diarization for free-living audio recordings
  
  In this project, we are developing an accurate and contextualized diarization processing pipeline for processing audio from free living. This is being developed in the context of ongoing mental health application projects and could likely be useful for several other healthcare and consumer applications.
  
  

• Mobile sensing for schizophrenia relapse prediction
  
  In this project, we are working on mobile sensing solutions to automatically identify behavioral patterns preceding a schizophrenia relapse with the help of machine learning/deep learning models.

• Unobtrusive blood pressure monitoring theme [Expertise: Machine learning and signal processing]

  • Improved oscillometry

  • Surrogate for blood pressure measurement

  • Pre-ejection period

  • Signal quality indicator

  • Blood pressure and cardiovascular physiology

• Patient monitoring theme [Expertise: Machine learning and computer vision]

  • Machine learning model for actigraphy

  • Deep learning model for scene understanding

  • Patient status monitoring and deterioration detection using physiological signals

• Cloud Data Platform theme

  • Generic, scalable framework for (real-time) data acquisition from connected devices, real-time processing, storage, distributed computing, and visualization. Support for different user models in the platform: data producer, study administrator, researcher, and observer.

  • Architected and implemented the data platform with the above framework on the AWS cloud. Stack: s3,Kinesis,Lambda,EC2,Cognito,DynamoDB,EMR,RDS,Glacier

  • Distributed data processing framework abstracting over Hadoop and spark for unified and easy usage by other researchers

• Machine learning for wearable healthcare applications theme

  • Stress detection with physiological sensors

  • Activity recognition, energy expenditure, and cardiorespiratory fitness monitoring

  • Athlete's training and recovery monitoring

  • Privacy-aware speech feature detection

• Deep learning for wearables and critical care applications

• Computer vision-based delirium detection in ICU patients

• Unobtrusive core brain temperature monitoring of neonates

• Epileptic seizure monitoring using physiological sensors

• Software-defined radio stack implementation on top of OMAP processors

• Collaborative spectrum sensing for cognitive radio

• Line follower robot using an 8051 microcontroller and sensor peripherals