Pulse - proof of concept

Calculate heart rate from mp4 video of thumb taken from a smartphone

Method

1. Calculate brightness on each frame (mean value over each pixel and RGB channel)
2. Remove baseline fluctuations using the Asymmetric Least Squares Smoothing

FFT method

• calculate FFT of signal
• find frequency corresponding to maximum FFT coefficient

Heart beats per minute is given by maximum FFT frequency * 60

Spike detector

• count positive amplitude spikes (and calculate spike occurence [spikes/minute])
• count negateive amplitude spikes (and calculate spike occurence [spikes/minute])

Heart beats per minute is given by the average of the positive and negative spike occurence

Video

Acquired using a Nexus 4 (Ubuntu Touch OS) with the flash LED on. The current script expects a .mp4 file without a soundtrack (bitrate of the video was reduced so it could be uploaded on the web).

Dependencies

• imageio: .mp4 processing
• scipy
• numpy
• matplotlib