Dec. 18, 2025 | Ethan Simmons
Extended high frequencies, above 8 kHz, can make speech more intelligible in noisy environments.
When we speak, the air moving through our vocal cords generates soundwaves vibrating at different frequencies: the higher the frequency, the higher the pitch.
Most of our modern audio technology, including hearing aids, headphones, and phone conversations chop off the “top end” of these soundwaves to compress the information coming in while keeping speech understandable.
But these extended high frequencies—8,000 Hz and above—contain important signals in the human voice, especially for comprehending speech in noisy environments.
Two researchers in the Department of Speech and Hearing Science at the University of Illinois Urbana-Champaign, Associate Professor Brian Monson and doctoral student Rohit Ananthanarayana, have patented an algorithm to identify and extract speech signals from noise by using extended high-frequency information.
The technique is novel, and with some investment, could be mapped onto existing modern hearing aids, earbuds and more.
Where are these high frequencies useful? Picture a restaurant date where you’re trying to pay attention to your partner speaking on the other end of the table, and voices of other customers are bouncing all around you.
“In those noisy settings in particular, that’s when these higher frequencies become valuable,” Monson said. “All the background noise masks out and interferes with those low frequencies, whereas these really high frequencies tend to stay pretty stable and unmasked, undegraded by the background noise.”
Humans can hear frequencies vibrating from 20 Hz all the way up to 20,000 Hz. Most modern hearing aids capture frequencies up to 6,000 Hz, which covers most everyday sounds. Most clinical hearing tests don’t test subjects’ hearing above this range.
Consonant sounds called “voiceless fricatives,” such as “s,” “sh,” “f” and “ph” sounds, contain energy above 8,000 Hz. Through grant-funded experiments, the researchers have shown the usefulness of these high-end frequencies—such as determining whether someone is facing you and speaking, or if they’re facing a different direction.
In those noisy settings in particular, that’s when these higher frequencies become valuable.
Brian Monson
Associate Professor, Department of Speech and Hearing Science at IllinoisOne experiment conducted by Monson and Ananthanarayana asked participants to listen to another person speaking, with and without the high-frequency range, and determine whether the speech was directed at them or not.
“If you test listeners’ ability to perform this task, to determine whether someone’s looking at you or looking away, they do quite a bit better at that task if they have access to those really high frequencies,” Monson said.
While performing high-frequency research, the algorithm became an interesting side project to work on. The algorithm was developed to be retrofitted as well: the listening benefits could be implemented on existing tech with a firmware update.
“We wanted to find some way to utilize the information in those extended high frequencies to enhance the target speech signal,” said Ananthanarayana, who began his Ph.D. at Illinois in 2021. “We identified a way that was novel but also feasible to try out with the resources and time that we had.”
The patent itself was four years in the making; Monson filed the provisional patent in December 2021, and the two researchers used their backgrounds in electrical engineering to test, code and strengthen the algorithm further.
To test its efficacy, they ran the algorithm on simulated speech recordings, where the “target talker” uttered short sentences in a noisy environment. The algorithm was judged for its ability to enhance the target signal’s clarity while suppressing background noise.
Together with the Office of Technology Management, Ananthanarayana and Monson put together their patent application to safeguard their idea, which was awarded in October.
“Ideally, someone would take interest in this—whether that’s us or someone else—and run with it to see if it’s implementable in hearing aid technology or other assistive listening device tech, like over-the-counter hearing aids, earbuds, headphones,” Monson said. “We think there’s potential there.”
Editor’s note:
The patent “Speech Identification and Extraction from Noise Using Extended High Frequency Information” was approved in October 2025.
To reach Brian Monson, email monson@illinois.edu
To reach Rohit Ananthanarayana, email rohitma2@illinois.edu
