Near-Infrared Data Pre-Processing for Glucose Level Prediction in Blood

Abstract

Estimated, 347 million people suffered from diabetes in 2004, and around 3.4 million patients died from consequences of high blood sugar. There are various techniques investigated by researchers in measuring the glucose level in human blood non-invasively, including ultrasonic sensor implementation, multisensory systems, absorbance of transmittance, bio-impedance, voltage intensity, and thermography. The implementation of near infrared (NIR) in predicting the glucose level in blood had been investigated in and the process of data pre-processing is presented in this paper. The selection of the wavelength region by previous researchers has been a debate as the suitable wavelength chose vary from one researcher to another. Despite the wavelength selection problem, the other fragment that needed a close attention is the process of enhancing the NIR data obtained from the experiment. The data pre-processing techniques used in this paper are the data filtering, data sampling, interval correction, wavelength selection and data distribution phases. The processed data then fed as an input to the linear and nonlinear prediction system implemented in this study.