Automatic Generation of a Training Data Set for Posture Estimation Aimed at Unconstrained Electrocardiogram Measurement during Sleep

Abstract

Monitoring the health condition of the heart on a daily basis can lead to the prevention of worsening or relapse of heart diseases, early detection of heart failure etc. Common electrocardiogram measurement devices are not suited for daily health care monitoring, due to the difficulties of wearing electrodes and the troublesome attachment of the electrodes. Therefore, the performance of electrocardiogram measurement by means of capacitively coupled electrodes in which electrodes are not affixed to the human body have been proposed. However, as a downside effect of this solution, there are difficult issues to be overcome such as how to maintain the relative positions of the human body and the electrodes, so the authors have developed a measurement system which can handle various sleeping posture changes using an electrode array. This system will estimate subjects’ posture and position in advance of measuring an electrocardiogram to achieve robust electrocardiogram measurement all in a single electrode array formfactor. In this paper, the method of measuring capacitance using the electrode array and of automatically generating the training set for posture estimation will be discussed.