Uncovering the Complexities of Speech

A person’s voice can communicate much about the speaker’s age, emotions, culture, and even health status. Rahul Shrivastav, Ph.D., studies how people derive such information from speech, which could lead to new treatments for speech and voice disorders, as well as result in some commercial applications.

In his Voice Acoustics and Perception Lab, Shrivastav identifies the subtle nuances of speech and develops standardized tools to measure changes in voice quality.

“From a health care perspective, voice quality measures could be used to detect disease, track the progression of disease, or monitor treatment results,” said Shrivastav, an associate professor in the department of speech, language and hearing sciences.

The research could be applied to the care of patients with Parkinson’s disease, who frequently experience problems with voice quality and articulation. By charting changes in speech intelligibility, clinicians could measure the disease’s progress and determine if speech therapy treatments are working. Shrivastav also studies how voice changes with age, which could help therapists discriminate between patients’ normal age-related speech changes and voice quality changes that signal the presence of disease, such as Parkinson’s.

Shrivastav’s research on perception of emotion in speech may lead to treatments for people with aprosodia, a condition caused by stroke or other brain injuries. Patients with aprosodia are unable to convey emotion in their voice, which can have a devastating effect on their personal relationships.

With funding from the National Institutes of Health, Shrivastav and co-investigator David Eddins, Ph.D., of the University of Rochester, are creating sophisticated algorithms that mimic the way the brain receives vocal signals. Preliminary studies show that the system’s measurement of voice quality is as much as 25 percent better than conventional measures.

Shrivastav, fellow department faculty member Alice Holmes, Ph.D., and Holmes’ former patient, Lee Krause, have developed a software system for fine-tuning digital hearing devices that is radically different from traditional tuning methods that often take multiple clinic visits and many months to complete. Shrivastav finds this project particularly exciting because it has “challenged the existing view on voice quality.”

The team’s software, known as Clarujust, is the first standard analytical approach to tuning cochlear implants and hearing aids. The program tests the patient’s hearing using actual speech sounds, not the tones used in the traditional tuning process. The software quickly analyzes the patient’s speech identification to determine the best cochlear implant or hearing aid settings for a particular patient.       

“This work is exciting because we have completely changed the paradigm,” said Shrivastav, adding that the technology could lead to many new applications.

“We originally focused on a narrow problem and found that the same approach, with some modifications, can be used in a variety of digital hearing devices, such as hearing aids and mobile phones,” he said. “The scope has become much wider than we initially thought.”