Scientists at the University of California San Francisco (UCSF) have recently published a paper describing their research toward the ultimate speech generating device that uses thought to create speech. Their current approach is the first step toward computer generated speech for people who are totally unable to communicate as a result of neurological impairments. Their experiments required the use of electrodes placed on the surface of the brain of volunteers. The volunteers were epilepsy patients whose neural activity was being measured with electrodes on the surface of their brain. The researchers recorded signals from brain areas that control the tongue, lips and throat muscles. Neural net technology was used to map the signals from the electrodes into known speech patterns. While this technology is at its infant stage, it may someday assist more natural speech for people unable to communicate.
For speech disadvantaged patients who lack a laryngeal voice source, research at UltraVoice is leading to the creation of products with more natural and understandable sound. This research is aimed toward mimicking the original voice patterns of the speaker before their vocal impairment.
Many people have records or videos of their speaking which are used to determine the parameters of their speech which are unique to them. In layman’s terms these parameters include the basic frequency of the voice, the rate and extent to which the frequency changes in normal speech, and the unique “vocal fingerprint” of the speaker.
Every larynx vibrates with a unique sound which makes voices immediately recognizable. When you hear a commercial which employs the voice of a famous actor, it is recognized immediately and listened to more acutely because of the vocal familiarity of the speaker’s “vocal fingerprint.” Imagine how laryngectomees will feel when reunited with a close copy of their original voices instead of the mechanical or gravely voice that is their current only option. Stay tuned for more updates on this exciting research.