From ShutEye to SleepScore, several smartphone apps are available if you’re trying to better understand the impact snoring has on your rest, allowing you to leave the microphone on overnight to record your harsh nasal growls and throat reverberations. But while smartphone apps are useful for tracking the presence of snores, their accuracy remains an issue when applied to real-world rooms with extraneous noise and multiple audible people.
Preliminary research from the University of Southampton examines whether your snoring has a sound signature which could be used for identification. “How do you actually track snoring or coughing accurately?” asks Jagmohan Chauhan, an assistant professor at the university who worked on the research. Machine learning models, in particular deep neural networksmight help verify who performs this hum-phonic symphony.
Although the research is still nascent, it is based on peer-reviewed studies who used machine learning to verify the creators of another data-rich sound often heard piercing the bloodthirsty silence of the night: the cough.
Researchers from Google and the University of Washington mixed the audio of human speech and coughing into a dataset, then used a multitasking learning approach to check who produced a particular cough in a recording. In their studyAI performed 10% better than a human rater in determining who coughed in a small group of people.
Matt Whitehill, a graduate student who worked on the cough identification paper, questions some of the methodology behind snoring research and believes more rigorous testing would reduce its effectiveness. Yet he considers the broader concept of audible identification to be valid. “We showed you could do it with a cough. It seems very likely that you can do the same with snoring,” says Whitehill.
This segment of audio-based AI is not as widely covered (and certainly not in such explosive terms) as natural language processors like OpenAI’s ChatGPT. But regardless, a few companies are finding ways to use AI to analyze audio recordings and improve your health.
Resmonics, a Swiss company specializing in detecting symptoms of lung diseases using AI, has released CE-certified medical software available to Swiss people via the myCough app. Although the software is not designed to diagnose the disease, the app can help users track the number of nighttime coughs they experience and the most common type of cough. This provides users with a more complete understanding of their coughing patterns while deciding if a medical consultation is needed.
David Cleres, co-founder and chief technology officer at Resmonics, sees the potential for deep learning techniques to identify a particular person’s cough or snore, but believes big breakthroughs are still needed for this segment of the AI research. “We learned the hard way at Resmonics that robustness to variations in recording devices and locations is as difficult to achieve as robustness to variations in different user populations,” Cleres writes via email. Not only is it hard to find a dataset with a range of natural coughing and snoring recordings, but it’s also hard to predict the microphone quality of a five-year-old iPhone and where someone will choose to put it. leave overnight.
So the sounds you make in bed at night can be trackable by AI and different from nighttime sounds made by other people in your household. Could snoring also be used as a biometric linked to you, like a fingerprint? Further research is needed before drawing premature conclusions. “If you’re looking from a health perspective, it might work,” Chauhan says. “From a biometric point of view, we cannot be sure.” Jagmohan is also interested in exploring how Signal processingwithout the aid of machine learning models, could be used to help spot snorers.
When it comes to AI in healthcare facilities, passionate researchers and intrepid entrepreneurs continue to face the same problem: a dearth of readily available quality data. The lack of diverse data for AI training can pose a tangible danger to patients. For example, an algorithm used in US hospitals deprioritized care black patients. Without robust datasets and thoughtful model building, AI often works differently in real-world circumstances than in sanitized practice settings.
“Everyone is really moving to deep neural networks,” says Whitehill. This data-intensive approach further increases the need for reams of audio recordings to produce quality cough and snore research. A machine learning model that tracks when you snore or gasp a lung isn’t as memorable as a chatbot who creates existential sonnets on Taco Bell’s Crunchwrap Supreme. It is always worth pursuing with vigor. While generative AI remains a priority for many in Silicon Valley, it would be a mistake to hit the snooze button on other AI applications and ignore their dynamic possibilities.
