Feature Analysis Focused on Temporal Alteration of the Eyeblink Waveform Using Image Analysis

Abstract

In this paper, we propose an eyeblink feature parameter for automatically classifying conscious and unconscious eyeblinks. For the feature parameter, we focus on eyeblink waveform integral values, which are defined as a measurement record of the progression of eyeblinks. Previous studies have used duration time and waveform amplitude as the feature parameters. The integral values, on the other hand, have characteristics of both of those feature parameters. We obtain these parameters using a National Television System Committee format video camera by splitting a single interlaced image into two fields. We use frame-splitting methods to obtain and analyze the integral value of the eyeblink waveform. We experimentally compared the feature parameters to automatically classify conscious and unconscious eyeblinks. Duration time and amplitude did not significantly differ in some subject cases; however, we confirmed a significant difference when using the integral value. Our results suggest that eyeblink waveform integral values are effective for discriminating conscious eyeblinks. We believe that the integral value of the eyeblink waveform is applicable to an eyeblink input interface.