Accurately detecting R-peaks in electrocardiogram (ECG) signals is important in various health monitoring applications, such as cuffless blood pressure measurements. In some cases, such as with single-arm ECG measurement, the signal may be buried in large amounts of noise. Moreover, many of the current existing ECG monitors pause the reading until the noise conditions are better, which may lead to a large loss in data. In this work, an adaptive approach is introduced, based on Empirical Mode Decomposition (EMD), to accurately detect the R-peaks in an extremely noisy ECG signal obtained using a single-arm measurement. The proposed algorithm starts by examining the Intrinsic Mode Functions (IMFs) extracted from the recorded signal, using the Hausdorff Distance (HD) as a selection tool, before applying the peak detection algorithm. Experimental measurements used to evaluate the algorithm were obtained from 10 healthy subjects for a total of 30 noisy ECG records, containing close to 2000 QRS complexes in total, the majority of which were at an estimated S/N (Signal-to-noise ratio) below −10 dB. The obtained results show a promising technique for detecting R-peaks in extreme noise with a low percentage of detection error ratio and an average percentage error in R-R interval estimation of 6.8%.