A former student of mine just sent me this video:
It’s a robotic vocal mechanism created by Professor Hideyuki Sawada at Kagawa University in Japan.
You can see that there’s an air compressor that pumps air through artificial vocal folds that gives the fundamental frequency. Then the air and sound travel through that fleshy portion and out through the lips. That fleshy portion is a resonator (it even includes that top portion to mimic the nasal cavity and an insert in the tube for the tongue), and the formant frequencies are modified with eight motors underneath and a valve that controls airflow to the nasal cavity.
The website has a schematic of the device. I’m curious if pitch is modified primarily through air pressure or the vocal cords or some combination of both.
If you watch the other videos (links at the end), you’ll see that this device can also mimic vowel sounds in real time. That’s very impressive.
A Talking Robot and the Auditory Feedback Learning for Natural Vocalization