AI for Extraction of Biomedical Signals from Headphones

In an age of so many new wearable devices, e.g., smartwatches, glasses, rings, clothing, and so on, headphones can be recognized as the first widely adopted wearable device. They have been around for more than a century and have been used mostly as an output device for listening to music, or, in the
recent decades, talking on the phone. Even more recently, Ohmic has developed a technology formed by a suite of hardware and software solutions that enables headphones to go beyond their initial purpose. Ohmic has created a dongle to which any wired headphone can be connected to and that transforms any simple headphone into a smart version of itself, enabling gesture recognition (tapping and sliding), user identification, and biometric monitoring (heartbeat). It does that by using a circuit to cancel the incoming audio signal while amplifying and processing the signals of interest.

While the prototype can already read biometric signals, user activities like walking and running interfere with the desired signal causing a reduction in the signal-to-noise ratio (SNR). These so-called motion artifacts are a very well-known problem1 for wearable devices in general, however it can be mitigated by use of artificial intelligence to extract desired features, e.g., heart rate and heart rate variability. The latter being used to detect human emotions.

In the current context of making any headphone smarter, the intern student will be researching and developing the most appropriate machine learning strategies that can be used to extract biometric features, more specifically to ensure signal integrity, to understand the needs for future development,
and to lay the basis for the path from R&D prototypes to mass market products.

More precisely, the main tasks will be :
? Working on a literature review and comparison table of machine learning techniques for
biometric signals, highlighting their respective advantages and drawbacks.
? Developing AI algorithms for feature extraction of biometric signals, such as heart rate, heart
rate variability, respiration rate, and in-ear canal characterization.
This project is stand-alone and no other projects are planned at this time.

Faculty Supervisor:

Giovanni Beltrame

Student:

Partner:

TandemLaunch Inc

Discipline:

Computer science

Sector:

Finance and Insurance; Professional, scientific and technical services

University:

Polytechnique Montréal

Program: