10:30 AM - 12:10 PM
[3Rin2-25] A study of observation fluctuation reduction method for ear acoustic authentication
Keywords:Ear acoustic authentication, Machine learning, Equal error rate, Time series data, Frequency analysis
Ear acoustic authentication is a type of biometric authentication that uses the acoustic characteristics of the ear canal.
A special earphone which has a driver and a microphone is used in this system.
The measurement data has error due to the attaching and detaching of earphone each time.
In our previous study, we proposed a special earphone which has a driver and two microphones.
And multiple features are concatenated.
However, the accuracy didn’t improve.
It’s conceivable that the concatenation method is not good.
In this study, we conceive that multiple features obtained by multiple microphones in one measurement complement observation fluctuation.
As a result, the accuracy was improved.
By analyzation of variance, we can say that the features were interpolated, and the learning data increased by using multiple microphones.
We conclude that the method using the earphone which has multiple microphones is effective.
A special earphone which has a driver and a microphone is used in this system.
The measurement data has error due to the attaching and detaching of earphone each time.
In our previous study, we proposed a special earphone which has a driver and two microphones.
And multiple features are concatenated.
However, the accuracy didn’t improve.
It’s conceivable that the concatenation method is not good.
In this study, we conceive that multiple features obtained by multiple microphones in one measurement complement observation fluctuation.
As a result, the accuracy was improved.
By analyzation of variance, we can say that the features were interpolated, and the learning data increased by using multiple microphones.
We conclude that the method using the earphone which has multiple microphones is effective.