Non-Invasive Blood Glucose Estimation Using Photo Plethysmography and Extreme Learning Machine Algorithms

The current research describes a non-invasive blood sugar level estimation device that uses Photo Plethysmography (PPG). Prior research has demonstrated that using a light-sensitive sensor improves blood sugar level prediction. The PPG signal of the individuals is captured using a constructed light sensor-based data-gathering device. The primary advancement of the results paper is the examination of many PPG signal characteristics through the efficient estimate of BGL values using the Only One Pulsed Assessment approach. Regarding the ADC conversion, the sampling rate of the frequency was set at 1 kHz with an accuracy of twelve bits during the dataset collection process. This PPG dataset segment contains 219 sample points. Features such as time and frequency domain characteristics are used in BGL value estimation. Using the previously suggested feature sets, a machine learning network is developed, and a BGL value estimate is carried out. The PPG medical data are also obtained, and both of them are trained to predict patient diabetes or nondiabetes using ELM algorithms. The method that produces the best results is taken into consideration. The top classifier was the ELM, which had a training time of 9 seconds and an efficiency of 99.6%. Based on the research, the Extreme Learning Machine (ELM) achieved more accurately in the machine learning methods.