Anomaly Detection for Identity Authentication Using 1D-CNN-LSTM Autoencoder and Doppler Radar with Classifiers

This study presents an integrated framework combining 1D-CNN-LSTM-Autoencoder-based anomaly detection with identity authentication using machine learning classifiers. The 1D-CNN-LSTM Autoencoder enhances system reliability by identifying and excluding faulty intervals in timeseries data caused by system instability or random movements. The dataset, collected under a controlled protocol, features scenarios such as Resting and Valsalva maneuvers for training and testing. To address measurement data interference, which can affect identity recognition accuracy, anomaly detection filters out corrupted signals, maintaining system stability and reliability. Among the machine learning classifiers, Cubic SVM demonstrated superior performance, achieving an accuracy of 8 8. 7 %. This integrated approach enhances system robustness, offering promising applications in secure access control, smart environments, and healthcare monitoring.