Computer vision technology has benefited greatly from the leaps in neural network processing capabilities, and more and more, vision tech is mimicking automated human vision. But would you trust your baby’s health to a camera?
Sivakumar Nattamai, CEO and co founder of Cocoon Cam, believes that you should, and he has the research to prove it. Nattamai, along with his co-founder Pavan Kumar, created a computer vision-based baby monitor that not only lets parents see and hear their baby sleeping, but analyzes the baby’s chest movements so they can always be sure that baby is safe and sound.
Nattamai was proud to tell me that Cocoon Cam is the only non-invasive, hospital tested baby monitor on the market, and as a new parent myself, I wanted to hear more. I caught up with Nattamai recently to find out more about how computer vision works with his monitors.
How did Cocoon Cam come about?
It came about when my daughter was just a newborn. Before kids, my wife and I used to come home and relax and have a coffee together to talk about our day. Once Leena, my daughter, was born, our minds just wouldn’t let us relax. We could only think about whether she was okay while she was sleeping. The mind of a parent is only concerned with whether the baby is okay, and the only way to know is to buy a baby monitor. So I went and bought the best baby monitor out there, that has both audio and video. But all you get to see on the monitor is this sleeping baby that is lying very still, that’s it. It doesn’t really tell you if she’s okay or not, just that she’s there.
In my engineering mind, I started thinking, “There must be a better way to do this.” I started prototyping different wearable devices like your watch, starting with something that you can attach to your baby’s ear which gives you the heart rate. But obviously, wires hanging from your baby who is two weeks old is not the best thing to do! So, I started experimenting with different non-contact based measurements. I tried radar, and then I tried sonar; they worked great but do you really want to shine radar or sonar rays at your baby all night long?
After a lot of experimentation, I found that the best solution that worked was computer vision, that is, computer analysis of video looking at your baby. People already have video cameras, so I figured that was the best way since the only thing the camera emits is infrared for night vision. It uses heat to see when there is no light, and then we analyze the video. We were actually able to extract respiration information from the video feed, and from that, we can tell that the baby is okay. And the most important thing is that it doesn’t require any type of wearable device, or anything attached to the baby’s clothing or bedding. That’s how this all started.
How are you able to do that in real time?
Just as with all IP cameras, the feed is near-real-time. The time between video being captured to the analytics being displayed to your phone is minimal, at under five seconds.
Cocoon Cam is an IP camera which streams the video to the cloud and your smart phone. This is just like the security cameras used for home security monitoring. But we also built a cloud platform that is monitoring the video stream in a secure manner, so we can run the computer vision algorithms in the cloud and provide you with analytics and graphs sent directly to your phone.
It’s a cloud-based AI system that monitors your baby and uses a video camera as its interface?
Exactly. The camera is the sensor and the interface to you is your mobile phone or your iPad. We support both iOS and Android devices.
How have you been able to do this and other people haven’t?
That’s a question that everybody asks me! The field of computer vision is still emerging, and it’s not well known, first of all. We also developed the software slowly over a period of years, and I had a great way to test videos, since I had my daughter and could just take videos every day, which was crucial as we were in the early stages of experimenting to find what worked. My co-founder, Pavan Kumar, has experience specifically in computer vision. When we combined both my experience and his experience in computer vision, we were able to make something that uses computer vision to extract the baby’s breathing, and make a hardware that is affordable to the mass market. More recently, we’ve been very fortunate to have built a world-class team of computer vision experts, including several with PhD’s and masters degrees in this highly specialized field.
There are a lot of competitors in baby video market, so why you didn’t you just build some software that runs on Nest, for example?
Actually, that’s a great question. Nothing that we’ve done locks us into the hardware that we are currently using. We chose this hardware to launch with because of affordability, as I mentioned before, and also because it’s the lightest-weight camera system available for baby monitoring — which is a safety feature. But the most unique consideration that really sets our system apart is security. We initially got started testing with regular cameras that were available in Best Buy. The problem with most cameras, though, is that the setup to get the video past the firewalls in your house and your router to our servers require you to open up the ports in your router, which makes the video stream insecure.
Since it’s your baby, and your home, we wanted to make sure that security was an important aspect of the product we ultimately shipped to customers. That’s why we started to work directly with a camera manufacturer on custom software that helps us securely transmit the video from your device to our cloud servers. With that comes ease of setup with security; instead of messing with your router, you just use your phone and set up the camera and everything works.
Is this a stage one product? What is the future of this technology?
The future of Cocoon Cam extends far beyond monitoring breathing. For example, other monitor companies advertise that they offer cry detection, but how do you actually detect the difference between a cry and other noise? With our approach using big data, now we have a system that can sense and process both video and audio and sense real crying — cry detection that really works.
Also, now that we have reliable respiration information, there are a lot of other interesting things we can do. For our breathing graph technology to work properly, the baby has to be extremely still. When the baby is still and breathing regularly, you know that the baby is actually sleeping. Because of this, we can then produce metrics on how long they are sleeping and their sleep quality. We can accurately say your baby slept at this time and woke at this time; for example, yesterday when I left my son in the crib and he was crying for a few minutes, and then all of a sudden I heard nothing. What do you do at that point as a parent? With our algorithm, you know that the baby is okay, because you can sense breathing and you know that he’s actually sleeping now.
The other thing that we are working hard on now is the question, how does a parent consume the data that they receive from the technology? That’s going to be a big factor. When you start actually analyzing all this data, how do you help parents gain insights from it that improve their lives, and make it user friendly? From the beginning, we’ve been an engineering-led company, but this year we are starting to focus more on design in both software and hardware, to make the world’s best baby monitoring system.
With the tracking of the breathing, obviously, parents are relying on that for safety information. What are your concerns about risk?
No device is perfect, and it’s important to note that Cocoon Cam is not FDA-approved and is not a medical device. But to help continually validate and improve our product, we are involved in IRB-approved testing at Stanford’s Neonatal intensive care unit and at UCSD. We’ve done enough testing now to know that we are accurate to within five breaths per minute when compared to the hospital’s own breathing monitors that require contact with the newborn’s sensitive skin. Results like that are how we feel comfortable, and I can sleep at night, shipping this product to people saying, this works. I rely on it every day with my daughter and now with my son. Cocoon Cam has given my family that extra peace of mind, and our mission as a company is to do the same for families everywhere.