Hearing Your Blood Sugar: Non-Invasive Glucose Measurement Through Simple Vocal Signals, Transforming any Speech into a Sensor with Machine Learning
Nihat Ahmadli, Mehmet Ali Sarsil, Onur Ergen
TL;DR
This work investigates non-invasive glucose monitoring by analyzing vocal biomarkers. Using the Disvoice feature set and PCA to eight components, a ridge-regularized logistic regression classifier is trained to distinguish BG above or below 100 mg/dL, with the acoustic predictor $z$ formed from eight principal components: $z = \theta_0 + \sum_{i=1}^8 \theta_i x_i$, and probability $p = \frac{1}{1 + e^{-z}}$. On data from 49 participants (70 voice samples), the model achieves about 87% training and 85–86% test accuracy, with LOOCV supporting robustness (average accuracy ~84–86% and $p<0.01$ for the predictor). This suggests potential for a painless, low-cost alternative to conventional glucose monitoring, though validation in larger, diverse cohorts is required to confirm generalizability and practical viability. The study highlights a pathway for integrating vocal analysis into diabetes care, leveraging CAPE-V–based recordings and standardized feature engineering to extract clinically relevant biomarkers from speech.
Abstract
Effective diabetes management relies heavily on the continuous monitoring of blood glucose levels, traditionally achieved through invasive and uncomfortable methods. While various non-invasive techniques have been explored, such as optical, microwave, and electrochemical approaches, none have effectively supplanted these invasive technologies due to issues related to complexity, accuracy, and cost. In this study, we present a transformative and straightforward method that utilizes voice analysis to predict blood glucose levels. Our research investigates the relationship between fluctuations in blood glucose and vocal characteristics, highlighting the influence of blood vessel dynamics during voice production. By applying advanced machine learning algorithms, we analyzed vocal signal variations and established a significant correlation with blood glucose levels. We developed a predictive model using artificial intelligence, based on voice recordings and corresponding glucose measurements from participants, utilizing logistic regression and Ridge regularization. Our findings indicate that voice analysis may serve as a viable non-invasive alternative for glucose monitoring. This innovative approach not only has the potential to streamline and reduce the costs associated with diabetes management but also aims to enhance the quality of life for individuals living with diabetes by providing a painless and user-friendly method for monitoring blood sugar levels.