In this paper, cough could be detected automatically by quantitatively evaluating the change in the waveform shape of the photoplethysmograph(PPG). By separating the fluctuation component due to heartbeat and the fluctuation component due to respiration contained in the PPG, our method can detect coughs using a linear discriminant analysis from the shape change of the respiratory fluctuation component. Since the cough can be detected without being affected by the surrounding sound environment by using the PPG, there is a possibility that the risk index of a serious disease can be automatically measured non-invasively and easily in hospital rooms and at home. In this paper, we evaluated the detection accuracy of voluntary cough by experiments and the effectiveness method was shown.