NonInvasive Blood Gloucose Estimation Using PPG Signal

Abstruct: Our project aims to estimate blood glucose levels in a non-invasive way using PPG signals captured from the smartphone front camera.First, the signal is acquired and then filtered to remove noise and obtain clear pulse signals.After that, we extract 125 features from each pulse, transforming the signal into numerical data that can be processed by machine learning models.The data is stored in CSV format and loaded using Pandas, then preprocessing is applied.Next, we use regression models such as SVR, Linear Regression, Ridge, Lasso, and Elastic Net to predict blood glucose levels.The dataset is split into training and testing sets based on subjects to ensure proper generalization.After training, we extract the coefficients of each feature and use them to perform feature selection by identifying the most important features.We also apply cross-validation to improve the reliability of the results.Finally, one of the most important improvements we applied was averaging multiple pulses for the same subject, which significantly enhanced the prediction accuracy. This work was conducted at Arab International University (AIU), Syria. The official website of the university is: https://www.aiu.edu.sy