An improvement of generalization ability of respitation rate classification using plethysmogram data

Abstract

This paper describes a study on discrimination of respiration rate based on plethysmogram (pulse wave) data. It is known that pulse wave is slightly changed by breathing. By utilizing this behavior, we can create an interface which can control a variety of instruments by breathing. However, the stable discrimination of plethysmogram data is difficult since the shape and the amplitude of pulse wave are different in each day. In this paper, we verify the generalization ability of SVMs by using evaluation data which are measured in different days from when training data were measured. Moreover, we consider the improvement of the generalization ability of SVMs which are trained using training data which are measured in plural days.