Identity Authentication System using a Support Vector Machine (SVM) on Radar Respiration Measurements

Abstract

Non-contact continuous authentication system can play an important role in maintaining security for a system throughout a login session. Unobtrusive respiration measurement using microwave Doppler radar has shown great promises in healthcare applications and can be a potential replacement for a traditional authentication system. Reported results for identification of human subjects based on radar-captured breathing patterns have focused on using dynamic segmentation which can produce false classifications for similar inhale to exhale area ratios. In this work, a continuous and highly accurate user authentication approach based on non-contact Doppler radar respiration measurement is examined by integrating a Support Vector Machine (SVM) with radial basis function kernel classification. The Fast Fourier Transform (FFT) spectral contents for radar captured respiratory motion traces allowed detection of various unique features and patterns. The proposed system was tested and validated for six test participants with 100% success rate. The proposed authentication system has potential applications including security, health monitoring, IOT applications and virtual reality.