Comparison of Mental State Discrimination Accuracy by Pulse Wave Using Multi-Layer Perceptron and Recurrent Neural Network

Abstract

This research aimed to compare of accuracy for machine learning using pulse wave. The subjects were 32 healthy young adults. They were divided to two groups by psychological tests. The pulse waves were measured during four emotional audiovisual stimuli. The subjects were discriminated into the mental stable or the mental unstable by Multi-Layer Perceptron (MLP) and Recurrent Neural Network (RNN) by using pulse wave, and the accuracy was calculated. The rate of the RNN was higher than that of the MLP for the most of the stimuli. These results suggest that RNNs would suitable for machine learning using pulse wave.