Investigating Effect of Sleep and Meditation on HRV and Classification using ANN

Abstract

Driver sleep detection plays an important role to avoid the fatal accidents caused due to the drowsiness of the driver. EEG measurement is effective in the case of sleep detection but it is comparatively difficult to obtain during driving. Different states of subjects give different heart rate variability measures making it possible to distinguish the states of subjects. During meditation state, the heart rate variability (HRV) is increased and higher HRV makes it easy to switch from the rest position to an alert position without dizziness. The proposed algorithm utilizes HRV to distinguish between normal state, sleeping state and meditation state. The mean distance between two R peaks observed in the normal state is 0.66 seconds which increased to 0.87 seconds in meditation state and increased further in sleeping state to 1.2 seconds. Spectral analysis of the HRV showed that power in the low-frequency band(0.04Hz to 0.15Hz) decreased by 34.32% and the power in the high-frequency band (0.15Hz to 0.4Hz) increased by 60.01% in sleep state compared against the normal state. In meditation state, the fluctuation observed in the R to R interval is more, resulting in an overall increase in HRV. The standard deviation of heart rate in meditation state increased to 6.14 from 5.28 obtained in non-meditation state. In this paper, the artificial neural network is used to distinguish between normal, meditation and sleeping state. Time domain, frequency domain, and nonlinear features are extracted from the dataset collected from the Physionet website and neural network is trained using these features. 70% of the samples are used for training, 15% for validation and the rest 15% are used for testing the neural network. The overall accuracy of the neural network observed is 85.4%. It classified all the samples of the sleeping state correctly giving 100% accuracy for the class. For normal and meditation class, the observed accuracy is 85.5% and 71.4% respectively.