Rupal Patel Captures Voice Melodies

Healthy adults with fully developed vocal systems convey more information by producing speech and changing the melody of their voice.

But children and adults with severe speech-motor disorders tend to rely more heavily on melodic cues, such as volume and duration. Patel uses her understanding of speech melody to create computational tools that can dramatically improve a disordered speaker’s ability to interact with the world.

In one project, Patel overlays meolodic fluctuations from a disordered speaker’s voice with a sentence spoken by a healthy donor of the same demographic. By merging the two signals, she creates a novel synthetic voice that conveys the user’s personal identity.

For children learning how to read, Patel also develops digital tools with visual cues—such as a rising and falling line—that signal pitch changes. Research suggests that by understanding the melody of speech earlier, children may achieve greater reading comprehension.