Scientists and engineers recently have been able to measure people’s heartbeats without even making contact with the skin, including using cheap video technology that detects skin tone changes. This has been possible with live and present subjects, but now a student team at Imperial College London has developed software that can analyze previous recordings of people speaking in front of the camera and estimate their heart rates.
The software uses both audio and video as the inputs, analyzing the skin tone and spectrograms of the person’s voice. It’s not clear how accurate the final product is, but it’s certainly intriguing. There are certainly medical uses that can be envisioned, but the authors also envision using the algorithms to help keep politicians in line.
Here’s an amusing video the team made for one of the prototypes of the software:
More info: Hearing Heartbeat in Audio and Video: A Deep Learning Project…