Feasibility of using Air-conducted and Bone-conducted Sounds Transmitted through Eyeglasses Frames for User Authentication

Recently, the demand for eyeglasses as a local industry has declined. Therefore, there is a need to improve the design and functionalities of eyeglasses manufactured in Japan. This study investigated user authentication using air- and bone-conduction microphones built in the frame of eyeglasses. Therefore, we propose a new user authentication system that uses air- and bone-conducted speech signals for speaker identification using the spectral ratio of air- to bone-conducted speech as a feature quantity. Using machine learning methods, the data were analyzed for a speaker identification system with the logarithmic power spectrum of air- and bone-conducted speech sounds of the five Japanese vowels. The study revealed resistance to voice fluctuations due to speech disorders using the voice data of two persons with a cold.