This paper proposes a novel diagnosis support method based on peripheral autonomic nervous activity for patients with Parkinson's disease (PD). The method measures rates of change in fingertip plethysmograms and arterial stiffness in transition between a supine position and a standing position of a subject, and calculates fingertip plethysmogram and arterial stiffness changes associated with angular variation from the pre-standing supine position to the standing position during a head-up tilt test. Based on the measured indices, the classification probability of PD presence is finally obtained as a biomarker using a log-linearized Gaussian mixture network. 25 patients with PD symptoms (15 with autonomic defects and 10 without) took part in the experiment. The results showed non-significant differences between the patient groups with autonomic defects and without autonomic defects in comparison of each single index on fingertip plethysmogram and arterial stiffness, but a significant difference (p<0.001) between the two groups was observed in the output index of the proposed system. Moreover, receiver operation characteristics (ROC) analysis showed that the area under the curve (AUC) of the proposed biomarker was 0.83 and the classification rate of PD presence was 100% for the learning data and 80% for the unlearned test data.