Personal Identification Using Feature of Floor Vibration by Walking

Abstract

For early diagnosis and treatment of disease, sensor systems that automatically get the physiological data in the toilet and bathroom have been developed. The data that obtained by these sensors need to be classified individually when two or more people live together. However, most of the existing personal identification methods, it is difficult to protect the user's privacy and reduce the burden. In this paper, we developed a personal identification method using features of floor vibration by walking on the wooden floor. From the waveform measured with the vibration sensor, eight features, peak to peak values and time intervals of each step, were extracted. As a result of personal identification by Mahalanobis' distance using the extracted features, the identification rate was 92%. As the number of registration data was increased, Mahalanobis' distance decreased a small value after the third day.