The rPPG is perhaps the most important remote contact-free method of extracting heart rhythm information from video observing wide applications in health care and computing. In this work, three rPPG approaches, namely convolutional neural networks (CNNs), Independent Component Analysis (ICA) and CHROM, are explored. Testing has been performed on thirty participants in different environments under varying light and motion conditions. It follows from the videos collected with the RGB camera and the facts provided by the pulse oximeter. For this work, an averaged model evaluation incorporated factors such as MAE, RMSE, Pearson’s coefficient and success rate. CNN method was the best among the others, with an average error of 2.1 beats per minute (bpm) in comparison with 3.7 bpm for CHROM and 5.2 bpm for ICA. CNN also had a high success rate of 93.2% in the presence of motion and changes in illumination. The robustness of rPPG over a wide spectrum of skin tones, on the other hand, addressed one of the inherent vestiges of the majority of the rPPG studies. The accuracy and efficiency of any approach in rPPG highlight the epitome of deep learning in rPPG, though requiring more computational power. The work becomes significant in the non-contact monitoring of vital signs as it offers details from head-to-head in the comparison of the traditional and the contemporary approaches of rPPG while giving hope for surpassing precision and robustness of the existing applications in healthcare as well as telemedicine.