The science of

WHO

A platform where facial recognition and bio-metrics analysis are combined with machine learning to transform well-being.

We believe that communication is much more than just verbal. To detect and monitor wellbeing clues we are looking at multiple data points including:

Computer Vision

We use computer vision to not only identify at-risk individuals but also to compare and contextualize information received via other sensors. Computer vision has come a long way the last few years. A machine can now pick-up tiny variations in someone’s face reliably.

Each time a user records a video, we analyze 68 facial landmarks in each frame. In a typical video, this represents over 35700 information points.
These points are used to detect mood but also blinking which is one of the more commonly accepted ways to track fatigue.
To understand more about how blinks can indicate fatigue you can look into many studies around PERCLOS analysis for example.

NLP/Voice

We use NLP and Voice analysis technology to hear what the person has to say and what their current concerns are. We are doing research on other approaches to using voice data in our analysis as we believe that voice patterns and intonations are just in their early stages of development.

Machine Learning

ML is used to contextualize the different datasets that are relevant to our analysis.  We augment our datasets with external indicators that impact wellbeing (air quality is a simple example of this)
This is also where we take into account many scientifically backed studies such as Epsworth, PHQ-2, PHQ-9 and trail making.

All these elements are clues that we piece together to better understand how the person is doing.

Data Bias

Okaya’s strength – and long-term plan- is being able to personalize this analysis for each user. It will help us tackle one of the biggest risks in machine learning: Bias.Datasets are biased by default. Here is a fun video summarizing the problem of biased data.

Rather than keeping a blind eye to the problem, leading companies are taking a proactive approach to identifying, quantifying, and modeling data bias to reduce its impact on algorithms.

 

If you’d like to learn more about Okaya feel free to reach out, we’ll be happy to share more information with you.

 

Let’s Get Started

Ready To Make a Real Change?