Introduction: Activation times in intracardiac electrograms (EGMs) of post-ischemic ventricular tachycardia (VT) patients are paramount for characterizing abnormal pathways sustaining the arrhythmia. In this context, the onset of the near-field component can be exploited to characterize the latency of the EGMs arising from the arrhythmogenic substrate, the so-called abnormal ventricular potentials (AVPs). This work proposes a novel approach for AVP onset identification, leveraging the Hilbert-Huang Transform (HHT) and a knee-point detection technique. Methods: A dataset composed of 940 AVPs with onsets marked by multiple experts was used. In analogy with the SAFE-T method, the Empirical Mode Decomposition was applied on each EGM to extract the first Intrinsic Mode Function, on which the HHT was evaluated. Whenever the instantaneous frequency from the HHT was higher than a given threshold (explored between 40 and 320 Hz, with a step of 10 Hz), the instantaneous frequency and amplitude were multiplied to compute an expedient signal. On this threshold-specific signal, we computed the cumulative area, and the AVP onset was identified as its elbow-point. Results: Based on the mean absolute error, the method demonstrated good performance for threshold ranging between 110 and 160 Hz (24.0±0.2 ms). According to the analysis of the absolute error distributions, the 140-Hz threshold led to slightly better results, with median of 13 ms (IQR: 28 ms). In this latter case, the linear correlation between the estimated and reference onsets was 0.67. However, the Bland-Altman analysis revealed that the method tended to underestimate the early onsets. Conclusions: The proposed strategy offers an automated method for detecting the onset of AVPs. This approach is particularly relevant for electrophysiological studies aimed at VT suppression, as it could enable the identification of near-field components in abnormal conducting pathways, as well as a more accurate assessment of their local activation timing.