As you may know from my recent musings, I’m currently jolly interested in all things to do with artificial intelligence (AI) and machine learning (ML) (see Why, Hello FPGA and AI — How Nice to See You Together!).
As you may also recall, back in January 2019, the folks at QuickLogic — developers of ultra-low power multi-core voice-enabled SoCs, embedded FPGA IP, and Endpoint AI solutions — acquired SensiML, which specializes in AI for use with the smallest IoT devices.
Although they aren’t ones to boast, the folks at SensiML say that, “The SensiML Toolkit is a breakthrough AutoML tool for edge AI development, empowering teams of all sizes and capabilities to efficiently build intelligent sensing algorithms with or without data science skills.”
I plan on learning a lot more about all of this in the not-so-distant future. However, the reason for my waffling here is that the guys and gals at QuickLogic just called to tell me that the chaps and chapesses at SensiML have launched an initiative in which their AI will help in the fight against the COVID-19 pandemic.
This is really quite exciting. It starts with us all visiting the SensiML COVID-19 Cough Screening Dataset Data Collection Page (try saying that ten times quickly).
The idea is that you use this page and your computer to record a few seconds of the sound of yourself coughing, and then you submit this audio file to their database while also answering whichever non-identifiable health questions you are comfortable sharing.
What they are looking for is a mix that includes healthy people, individuals with respiratory conditions other than the coronavirus, and those having — or suspecting they have — COVID-19.
The folks at SensiML say that, “This technology, based on recent academic research, analyzes recorded cough noises to predict positive COVID cases with over 90% accuracy. While not a replacement for clinical testing, it can play a key role in improving back-to-work employee wellness checks and public facility screening to help protect the health of everyone.”
Once they’ve gathered sufficient audio cough samples, the next stage will be for folks to be able to receive a remote diagnosis. I’m not sure how that part will work. I presume someone who wants to know if they are infected will be able to upload an audio sample via the internet — maybe they’ll be able to call (or cough) it in over the phone — for the AI to listen to it, and for the person to be told, “There’s a probability of xx% you are infected/not infected” (I’ll ask the folks at QuickLogic or SensiML to post a comment explaining how all this will actually work).
Personally, I think this is a wonderful idea and I cannot wait to see this system trained, deployed, and in action. What about you? What are your thoughts on this?