Research

Our research focuses on understanding how the human auditory system processes complex sounds in real-world conditions and how we can diagnose hearing loss that affects these processes. We combine electrophysiology, psychophysics, and computational approaches.

See a list of our publications here.

Fast hearing diagnosis with the parallel ABR

Fast hearing diagnosis with the parallel ABR

An infant being tested for hearing loss has only as long as their nap. The diagnostic auditory brainstem response (ABR)—the standard test of an infant's hearing—is traditionally measured one frequency, one ear, and one stimulus level at a time, an exam that often runs longer than the patient will sleep. Together with the PoloLab at the University of Minnesota, we are developing the parallel ABR (pABR), which measures responses to every frequency in both ears at once and collapses the threshold search from three dimensions to one. The result is the same waveforms an audiologist already knows how to read, collected in a fraction of the time: in adults with widely varying hearing loss, pABR thresholds correlate with the behavioral audiogram at r = 0.90 and are 2.5 times faster to obtain than with a standard clinical system. Even when a session ends early, the pABR provides at least some information about every frequency in both ears, where serial testing would leave most untested. Our next step is to validate the pABR in infants with hearing loss—the patients who stand to benefit most from a faster, more complete diagnostic test.

Related publications

How the brain encodes natural sounds

How the brain encodes natural sounds

Listening in the real world means picking up a single voice in a crowded room, following a melody over a distant hum, or catching a friend's name on a noisy sidewalk. Most of what we know about how the brain processes sound, however, comes from laboratory experiments that have necessarily reduced these scenes to clicks, beeps, and isolated syllables—the price of the experimental control that traditional methods require. We have developed new ways to measure the human auditory brainstem response (ABR) directly from continuous, naturalistic sounds like speech and music, using carefully designed stimuli ('peaky speech') and signal-processing approaches that account for how the auditory periphery actually responds to complex sound. With these tools we have shown that the brainstem tracks individual talkers in a mixture, that responses to music and speech share more than they differ, and that masking by competing voices shrinks and slows subcortical responses in measurable ways. Our current focus is developing a deep neural network model of human auditory encoding. To do this we will use nearly all the data we have collected in the past decade, while also collecting tens of hours of EEG from each of a small group of listeners—data unlike anything currently available. We will use the model to inform the next generation of hearing aid signal processing and audiologic diagnosis.

Related publications

How the visual system shapes hearing

How the visual system shapes hearing

Watching a talker's face helps you understand them in a noisy room. The reason for this is partly obvious—lip-reading conveys phonetic information that the auditory signal alone may not—but the visual system also acts on hearing in subtler ways, biasing what the brain attends to even when the visual cues themselves carry no specific message. We spent several years investigating how vision shapes auditory selective attention, finding that a simple visual signal whose timing matches a target voice can make that voice easier to hear in a mixture, and that this effect operates even when the visual stimulus contains no useful information of its own. In collaboration with Jennifer Bizley's lab at University College London, we showed that the same principle holds in the auditory cortex of the awake animal: visual cues coherent with one of two simultaneous sounds bind to it neurally, helping the brain segregate competing auditory streams. Together these studies established temporal coherence as a genuine mechanism by which vision contributes to listening, complementing the more familiar role of lip-reading.

Related publications