Noninvasive Blood Glucose Level Detection Based on Matrix Pencil Method and Artificial Neural Network

Abstract

A method of improving the resolution of the detected blood glucose level by using the microwave detection technique is proposed in this paper. In this proposed method, the matrix pencil method and the artificial neural network are combined to help improve the resolution of the detected blood glucose level. The matrix pencil method is applied to extract the poles of the received microwave signals. And the artificial neural network which is very popular in the artificial intelligence field in recent years is also utilized to help distinguish the blood glucose level by training the poles extracted from the received signals. The reliability of the method is checked by establishing an earlobe model which is more realistic than it is in the former research. The mean error between the real blood glucose level and the detected blood glucose can be 0.09957% which is minor than 0.1%. The correctness of the method is testified by successfully detecting the blood glucose level with the precision of 1\xa0mg/dl. The UWB microwave detection system can satisfy the detection of the normal range of the plasma glucose level 70–240\xa0mg/dl.