3237 ニューラルネットワークによるストレス状態識別に関する一考察

概要

論文の詳細を見る
This paper investigates stress emotion recognition from the pulse signal as a part of emotion recognition from bio potential signals. The pulse data is measured by using an electric pulse sensor which detects variety of blood flow quantity from his/her ear. Feature pattern vector for stress emotion classification is composed with time series data of pulse and multi-layer neural network that involves self-organization map in hidden layer is proposed to achieve a stress emotion classifier. As an alternative classifier, LVQ and SVM are tested. Using the collected pulse data from Japanese male subjects, the classifier is trained and the classification test is carried out with the leave-one-out method. The obtained classification rate of two emotional staes (relax/stress) is 66.7% using proposed neural network.
社団法人日本機械学会の論文
2003-08-05