Blood Pressure (BP) estimation is an important component of cardiovascular health monitoring, with Photoplethysmography (PPG) emerging as a promising non-invasive technique. Our study introduces a novel approach for BP estimation by introducing a Siamese network architecture tailored for the inference of BP from PPG signals, employing a dual calibration signal approach as opposed to the conventional single calibration method. We evaluated the model’s predictive performance using Mean Error and standard deviation (ME [STD]), alongside Pearson’s Correlation Coefficient ( \(\rho \) ). Our results demonstrated the following: for diastolic BP, an estimation of 0.24 [5.92] mmHg with a \(\rho \) of 0.81, meeting the Association for the Advancement of Medical Instrumentation (AAMI) standard; however, for systolic BP, the estimation of 0.41 [10.65] mmHg with a \(\rho \) of 0.82 are comparable to current state-of-the-art approaches, but did not meet the STD of AAMI standard. Our study underscores the potential of deep learning-based methods to enhance cardiovascular health monitoring and to develop more precise BP estimation techniques. However, it also indicates that further refinements are necessary for improving systolic BP estimation.