A person’s heart rate can be accurately extracted from only a video of the person’s face, recorded with a regular web camera. While this opens up many opportunities for telehealth, it also raises privacy concerns, for example during an interview over video chat, which may expose sensitive health information about an applicant.
The goal of this project is to develop a novel method to conceal a person’s pulse in video recordings. The first method will superimpose arbitrary but imperceptible oscillations in the video stream to mask the true heart rate, while the second will attempt to completely remove all sensations such that existing algorithms fail. We will use available datasets of face videos and annotated heart rate signals, and we will build on and extend state-of-the-art machine learning algorithms.
_Note: Candidates should be experienced in machine learning and corresponding toolchains (e.g., PyTorch or Tensorflow), ideally being familiar with methods in Computer Vision. Experience in signal processing is beneficial but not required._
A person’s heart rate can be accurately extracted from only a video of the person’s face, recorded with a regular web camera. While this opens up many opportunities for telehealth, it also raises privacy concerns, for example during an interview over video chat, which may expose sensitive health information about an applicant.
The goal of this project is to develop a novel method to conceal a person’s pulse in video recordings. The first method will superimpose arbitrary but imperceptible oscillations in the video stream to mask the true heart rate, while the second will attempt to completely remove all sensations such that existing algorithms fail. We will use available datasets of face videos and annotated heart rate signals, and we will build on and extend state-of-the-art machine learning algorithms.
_Note: Candidates should be experienced in machine learning and corresponding toolchains (e.g., PyTorch or Tensorflow), ideally being familiar with methods in Computer Vision. Experience in signal processing is beneficial but not required._
Not specified
- Björn Braun, bjoern.braun@inf.ethz.ch
- Andreas Fender, andreas.fender@inf.ethz.ch
- Björn Braun, bjoern.braun@inf.ethz.ch - Andreas Fender, andreas.fender@inf.ethz.ch