DAEJEON, Feb. 16 (Korea Bizwire) — South Korean researchers have developed an artificial intelligence (AI)-based, ultra-sensitive voice recognition sensor by imitating the structure of the human ear.
A research team from the Korea Advanced Institute of Science and Technology (KAIST) announced Monday that they had developed a voice recognition sensor designed after a human cochlea, and are taking steps for commercialization.
A trapezoid membrane in the cochlea causes resonance phenomenon, amplifying sounds to allow the human ear to recognize sound from a distance.
Similarly, the research team applied a trapezoid-shaped piezoelectric film that vibrates in certain parts depending on frequencies to the sensor.
Once a sound vibrates the film, the resonant phenomenon that ensues generates a highly sensitive voltage signal.
The research team built on a similar concept proposed in 2018 to successfully create a prototype.
The voice recognition sensor was tested in an actual smartphone and an AI speaker, demonstrating outstanding capability in voice recognition due to less noise.
With 90 percent accuracy, the sensor is capable of identifying the speaker with less voice data than conventional systems.
M. H. Lee (firstname.lastname@example.org)