Patients with diabetes mellitus (DM) may experience problems such as peripheral tissue necrosis and gradual hardening of blood vessels; these may prevent normal metabolic behavior and may even require amputation under severe conditions. Therefore, evaluating the peripheral blood circulation state of patients with DM is crucial to enabling physicians to perform timely interventional therapy to prevent symptoms from worsening. To improve the general problems of treatment, such as high cost, the infection risk or misjudgment of specific groups, a smart blood perfusion monitoring system is proposed to noninvasively evaluate patients’ peripheral blood circulation state. This system uses near-infrared spectroscopy to noninvasively monitor real-time changes in peripheral blood perfusion with force is applied on the arm. On the basis of changes in peripheral blood perfusion with pressure, several indexes related to blood circulation state are proposed. Finally, a neural network technique was successfully applied to classify patients’ blood circulation state. From the experimental results, F-measure, sensitivity, positive predictive value and accuracy are 82.75%, 80.00%, 85.71% and 83.33%, respectively. The experimental results show that the proposed indexes (Indexes I-IV) are significantly related to blood circulation state and can be used to effectively evaluate peripheral blood circulation.