Biometric Authentication for Dementia Patients with Recurrent Neural Network

The usage of technology to monitor the health status of patients with chronic diseases continue to rise. Dementia is one such chronic disease which demands continuous monitoring and observation to keep track of the health status of the patient. The dementia patients use different services to contact healthcare workers or doctors. These services need login credentials. However, due to progressive and frequent memory loss and confusion, they face significant challenges to access the services. Hence, biometric authentication can play a crucial role to provide better support for them. This paper proposes a biometric-based authentication framework based on a recurrent neural network for dementia patients. The PPG and ECG signals from the wearable devices are examined for authentication purpose. Two distinct features of the signals: instantaneous frequency spectral entropy are provisioned to the LSTM network to train the system. From the dataset of ten participants, the accuracy of the PPG and ECG based identifications reached to 100% and 88.9% and F1 scores reached to 1.00 and 0.86 respectively.