In this paper, a non-invasive method for estimating blood glucose level is proposed. We investigate blood viscosity and harmonic distortion using a model with the assumption that high blood viscosity caused by high blood glucose level generates turbulence in blood flow. This consequently leads to the distortion of fundamental frequency of pulsation. Therefore, the harmonic components appear owing to the pressure on the vascular wall. Using the proposed non-invasive method, we calculate the amplitude of the harmonic components. The blood glucose level is estimated with a single regression model whose explanatory variable is the calculated harmonic component. Through validation experiment with 33 measured data, the third harmonic component achieved the highest estimation accuracy with 60% of the data meeting the ISO criteria.
Keywords: non-invasive, blood glucose level, regression model, diabetes, harmonic distortion, blood pressure monitor, Fast Fourier transform