Doctoral Research Theme

My research primarily deals with creating real-time health systems in a mobile environment. Let me break this down, one key point at a time:
  • Real-time : All the systems that I create are modeled to capture relevant information as it takes place.
  • Health : My systems are geared towards finding technological solutions, or providing better toolkits for understanding underlying phenomena, of health related problems.
  • Mobile : I like my systems to be implemented in a way such that people can take them along with themselves.

Doctoral Advisor

I am extremely fortunate to have Octav Chipara as my advisor. Octav heads the Mobile Systems Laboratory (MSL) at the University of Iowa's Computer Science Department.

Ongoing Projects

Exploring relationships between objective and subjective HA evaluation metrics

Using the data collected with AudioSense from over 50 individuals I am trying to (i) establish new, and (ii) validate existing relationships between reported outcome measures 
like listening effort, speech perception etc. and objective measures like SNR.

Completed Projects

Predicting Success of Treatment For Hearing Impaired via Ecological Momentary Assessment
Novel Patient

Using the data collected with AudioSense from 34 individuals with mild-to-moderate hearing loss, we explored two primary questions on how 
well can we predict the success  of a prescribed treatment for i) individuals who have never used a hearing aid, and ii) individuals who want to change their hearing aids. We found that:
  • For new patients, we can predict with accuracies above 90%. This is dependent, however, on the data containing 
    relevant contextual information, and some information about the patient as well
  • For change of hearing aids, we can predict the success of the treatment around 89% accuracy. Idiosyncrasy nullifies all 
    positive effects, if any exist, of adding information from other users.

Associated publication: ICHI 2015

Auditory Context Evaluation using Ecological Momentary Assessment

Context Evaluation
Hearing aid users experience several different contexts throughout their hearing aid assisted life. Among these contexts, there are some which are more prevalent than the others. Understanding these contexts is key to understanding the performance of the hearing aids. In this project, we:
  • Show the prevalence of different contexts, highlight the importance associated with them.
  • Propose a method of combining outcome measures.
  • Show that it is possible to discriminate between good and bad outcome scores with an accuracy of 78%.

Associated publication: Pervasive Health 2014

Stream Processing for High Rate Mobile Sensing Applications

With the development of new smartphones there has been a giant leap in multimodal sensing capabilities of these devices. With this increase in sensing capabilities,

there has been an increase in mobile phone based health applications. The challenge lies in a system that allows these applications to run while managing the already limited resources in an efficient way. In this project, we:
  • Present CSense, a high rate robust stream processing toolkit.
  • Conduct static analysis on the stream flow graphs of sensing applications to optimize frame exchange across components.
  • Show that our flow analysis reduces CPU utilization by as much as 45% compared to the baseline.

Associated publication: IPSN 2014

Tool for real-time and in-situ evaluation of hearing aids

Traditionally, hearing aid performance measurements are conducted in controlled lab settings.
The state-of-the-art evaluation techniques in the field of Audiology involve pen and paper. These methodologies are full of problems such as memory bias and being unable to generate real-world contexts and hence are unable to predict real-world outcomes of hearing aids. The challenge lies in trying to capture these contexts in real-time. In this project, we:
  • Present AudioSense, a system to capture contexts in real-time and in-situ.
  • Show that AudioSense is robust against network problems and performs with a reliability of 100% in test conditions.

Associated publication : CBMS 2013 (Best Student Paper)