'Saojeong' Attaches Human Ear-Inspired Sensors to Smartphones and AI Speakers

KAIST Professor Geonjae Lee's Team Develops Prototype, Opens Path to Commercialization

Photo courtesy of Korea Advanced Institute of Science and Technology (KAIST).

[Asia Economy Reporter Kim Bong-su] A sensor modeled after the human ear that can detect voice more accurately has been developed, and it is expected to be utilized in smartphones and artificial intelligence (AI) speakers in the future.

On the 15th, according to the Korea Advanced Institute of Science and Technology (KAIST), Professor Geonjae Lee and Dr. Heeseung Wang from the Department of Materials Science and Engineering recently developed a resonant-type flexible piezoelectric voice sensor capable of implementing highly accurate ultra-sensitive AI-based speaker identification and voice security technology, and succeeded in commercializing it by integrating it into smartphones and AI speakers.

The resonant-type piezoelectric voice sensor is designed so that when the sensor membrane vibrates due to voice, resonance occurs, enabling the acquisition of highly sensitive voltage signals. This is similar to how humans recognize sounds from a distance. In humans, the trapezoidal membrane in the cochlea generates numerous resonances within the audible frequency range, amplifying sounds so that even faint sounds can be accurately perceived. To maximize the effect of this principle, the research team used a very thin flexible piezoelectric membrane to mimic the human ear and created a resonant voice sensor capable of ultra-sensitive sound identification by implementing multiple resonance channels.

Professor Lee’s team first proposed the concept of a resonant-type flexible piezoelectric voice sensor in 2018, and this time, they theoretically clarified the resonance, frequency, and role of the piezoelectric membrane according to the sensor structure, while developing a voice sensor that is highly miniaturized and improved in performance.

By linking future Internet of Things (IoT) technology that accurately controls smart devices from a distance with voice encryption security technology, it is expected to greatly contribute to providing consumer-customized services. In particular, due to its excellent signal-to-noise ratio (SNR) and multiple channels, it has the advantage of improving speaker identification accuracy in AI voice services with a small amount of data.

In fact, when compared under the same conditions with existing commercial capacitive microphones, the research team revealed that the recognition rate for voice analysis and speaker identification was significantly higher, and the error rate could be reduced from 60% to 95% depending on the conditions.

This sensor was already unveiled last year at the Consumer Electronics Show (CES) through Pronics Co., Ltd., a company founded by Professor Lee. Currently, the technology demonstrates highly complete AI voice technology and is pursuing collaboration with leading IT companies in Silicon Valley through Pronics’ U.S. branch.

Professor Geonjae Lee said, "The mobile voice sensor commercialized this time has high sensitivity while drastically reducing size, so it can be applied as a core sensor driving future AI technology," adding, "The mass production commercialization process is also in the final stages, so it will soon be applied in daily life."

This research was conducted with the support of the Human Plus AI Sensor Center of the National Research Foundation of Korea and was published in the June 12 issue of the international academic journal Science Advances.

This content was produced with the assistance of AI translation services.