This study investigates the feasibility of using phonocardiogram (PCG) signals to support the early identification of pre-hypertension through supervised machine learning. Hypertension induces subtle mechanical changes in the cardiac cycle, affecting the acoustic properties of heart sounds (S1 and S2) and the timing of systole and diastole. Using a dataset of 78 participants, we extracted temporal, spectral, and statistical features from raw PCG recordings using the TSFRESH library. Correlation-based Feature Selection (CFS) was applied to identify relevant attributes. Several classification models were evaluated to distinguish between normotensive and pre-hypertensive/hypertensive subjects. The Logistic Regression model achieved the most consistent baseline performance, with an accuracy of 62.54% and an AUC of 0.62. While preliminary, these results suggest that PCG signals contain latent information correlated with blood pressure levels. This reinforces PCG’s potential as a low-cost, passive, and non-invasive screening tool for cardiovascular monitoring, warranting further research into advanced signal processing to enhance predictive power.