Clustering of Respirations as a Biometric Using ARS and Machine Learning Techniques

This paper proposes a personal identification using respirations measured by a Doppler sensor with machine learning techniques. The Doppler sensor is well-known method widely used for non-contact vital sensing. Our challenge is to achieve the personal identification using the respirations which are measured by the Doppler sensor and preprocessed by the accumulation for real-time serial-to-parallel converter (ARS). Through machine learning techniques including the k-nearest neighbor (k-NN) and the support vector machine (SVM), the personal identification between two persons are successful with more than 0.7 in the accuracy and in the F-score. In addition, it is also indicated that ARS results in the better performance with the machine learning techniques, compared with the preprocessing by the fast Fourier transform (FFT) as a preprocessing of data.