Smart speaker is a kind of wireless speaker and voice command device with an integrated virtual assistant that offers interactive actions and hands-free activation with the help of voice command. We use smart speaker as a smart time saving device whenever we want to pick a music we want to hear, controlling lights and other connected accessories in our smart home, or to search for a quick piece of information without having to type anything. An innovative research project from the University of Washington suggests that it is possible to transform them from time-saver to life-saver in the very near future!
Smart Speaker; Image Source: digitaltrends.com
According to recent research, the most common place for an out-of-hospital cardiac arrest is the patient’s bedroom. At UW, a team of researchers have found a way to carry out the cardiac arrest detection without the need for any additional hardware or even without touching the user while they are asleep. They plan to do it by taking advantage of smart speakers’ ability to always listen.
Associate professor Shyam Gollakota, UW’s Allen School of Computer Science and Engineering; Image Source: https://www.thebetterindia.com
The work was carried out by the lab of Shyam Gollakota, an associate professor in UW’s Allen School of Computer Science and Engineering. “A lot of people have smart speakers in their homes, and these devices have amazing capabilities that we can take advantage of”, Gollakota said in a statement. This is not the first time that Gollakota’s team has carried out some fascinating work in bringing out something new from a traditional ordinary device. In the the past, they have developed a smartphone app that can hear ear infections. They did it by granting ordinary smartphones diagnostic abilities to do that.
Smartphone app that can detect ear infections; Image Source: npr.org
“Smart speakers are increasingly becoming popular. You don’t need them to be recharged, and so they are plugged in all the time. As a result, you don’t need to be worried about power.” Gollakota told Digital Trends. During a cardiac arrest, it is very common that the victim won’t breathe normally. In a many cases it has been observed that they exhibit a symptom known as ‘agonal breathing’, a type of disordered gasping sound. The tool the team has developed is an app that listens for the signature sounds of cardiac arrest, mainly the ‘agonal breathe’. In the event that these are identified, it can call for help. It will either alert a caregiver or even call emergency medical services.
Picture this: You’re enjoying a pleasant evening in your own room watching movies. Suddenly, your heart stops pumping and blood flow ceases. You start fighting for breath, making guttural gasping noises, and involuntarily twitching but unfortunately there is none to perform first aid or take you to the hospital. The smart speaker on the bookshelf, however, recognizes signs of agonal breathing and calls 911 and you get to live another day!
It is not that easy to identify “agonal” sounds in the midst of other environmental sounds. To recognize these “agonal” sounds and differentiate them from the other sounds heard in a bedroom environment, the team developed a Support Vector Machine(SVM). This is a type of discriminative classifier; mainly a sorting machine in which we can label examples of what it is looking for and then categorize any new examples it experiences so that it becomes capable of skipping those.
Using a smart speaker to detect agonal breathing: a) Agonal breathing detection pipeline, b) Audio waveform and c) spectrogram of agonal breathing, hypopnea, central apnea, and obstructive apnea; Image Source: nature.com
“Our technology was tested on agonal breathing sounds obtained from 911 calls to Seattle King County’s EMS services during cardiac arrests from 2009 to 2017. We evaluated our technology on 164 hours of sleep sounds, collected across 35 different bedroom environments, as well as 82 hours of sleep lab sounds where patients had apneas, hypopneas and snoring events. [These] can sound similar to agonal breathing. We showed that we can identify agonal breathing sounds accurately in all these scenarios”, Gollakota said.
It is so important to reduce false-positive alert rate as it will be very unprofesional to alert either emergency services or loved ones unnecessarily. “To make the algorithm more robust, we will have to augment more data across different geographical areas beyond Seattle King County,” Gollakota said. But according to a recent paper published in the journal npj Digital Medicine, it has been showed that the UW team’s tool can detect instances of agonal breathing with 97% accuracy from a distance of up to 20 feet away. They are currently looking to commercialize the technology through a spinoff company called Sound Life Sciences.
Featured Image Source: technologyreview.com