In a time of virtual doctors’ appointments and commonplace smartwatches that track your heartbeat, a team of UW researchers have taken steps to further the integration of medicine and technology with a smart speaker algorithm designed to detect irregular heartbeats.
Led by Anran Wang, the team includes Dr. Dan Nguyen, clinical instructor at the UW School of Medicine (UWSOM); Dr. Arun Sridhar, assistant professor of cardiology at the UWSOM; and Shyam Gollakota, Ph.D., associate professor at the Paul G. Allen School of Computer Science and Engineering.
The algorithm uses the components of a smart speaker to detect differences in the intervals between heartbeats, thus indicating an irregular heart rhythm, all without any contact. The algorithm itself has two components: pinpointing only the heartbeat and its location using a microphone array already built into the smart speaker, and working to assess the presence or absence of disease based on the gathered heartbeat intervals .
But how can the algorithm separate the very small vibrations of the heart from much larger motions, like a car rumbling in the background, or even respiration?
“We leverage the characteristics of the heartbeats,” Wang, a Ph.D. student at the Allen School, said. “The heartbeat is around a frequency range of 40 to 120 beats per minute, which is much faster than other motions, so we designed a self-supervised learning approach which tries to find the location and maximize the energy falling within the given frequency range.”
Currently, the algorithm utilizes a seven-microphone array which enables the speaker to have a two-foot range in reliably detecting heartbeats and, in turn, heart rhythm. However, Wang noted that since the prototype only employs cheap parts, teaming up with companies like Amazon, with their Echo smart speakers and optimized microphones, would increase the range.
After successfully locating the heartbeat, tackling the second component of the process — inferring the existence of disease — becomes a matching game with existing scientific data.
“By segmenting the heartbeat signal to individual beats, then you can get the beat-to-beat interval, which is the key information that can be used to detect other diseases like atrial fibrillation,” Wang said.
When asked about next steps for the project, Wang noted that the human motion detection of the smart speaker needs to be further developed to differentiate a wider range of background noises, like TV and cooking noises. In addition, he hopes to make this technology accessible to commercially available smart speakers by collaborating with existing companies.
While this feature is currently designed as a spot check prompted when users stand in front of the speaker, ideally this feature would be used in smart speakers in the home setting as an ambient device to contactlessly monitor heart rhythm in the background .
Wang was vocal about the importance of smart speakers in daily life.
“My motivation is to try to enable as many applications as possible on the smart speaker platform ... because we are living at home every day, and [smart speakers] can assist us in many different aspects,” Wang said.
Reach reporter Sruthi Ravi at firstname.lastname@example.org. Twitter: @Sruthi_Ravi7
Like what you’re reading? Support high-quality student journalism by donating here.