Recurrent Neural Network with Tunicate Swarm Optimization and Shift Delta Cepstrum for Automated Driver Health Monitoring System

Fatal driving accidents are primarily caused by stress and weariness. Almost 80% of drivers experience stress when operating a vehicle. Anxiety, impatience, lack of concentration, and problems paying attention are all brought on by stress. So by closely monitoring the health of the drives we can avoid major accidents. Now a days there are many inventories in the field of health monitoring of the drivers, even though the accuracy and early detection is challenging. In order to overcome the drawbacks of the existing system, a driver health monitoring system based on optimized Recurrent Neural Network (RNN) is proposed. The collected ECG signal is first pre-processed by Butterworth filter and Savitizky-Golay Filter to improve the quality of the signal. After pre-processing, the features are extracted by using Fractional Fourier transform (FrFT) and Shift Delta Cepstrum (SDC). It enhance the pre-processed signal in the extraction of time and frequency domain. To classify whether the driver is fit, under stress and Cardiac Health issues the RNN algorithm is used for classification. The suggested classifier's learning rate is maximized with the application of Tunicate Swarm Optimization (TSO). To avoid the accident with other vehicle a buzzer is used to alert them in the cause of Stress and heart issues. According to the experimental approach, the techniques achieves 96.6% of accuracy, 94.6% of precision, 91% of Recall and 95% of Specificity. Thus, the driver health be accurately monitored using this proposed automated model.