Heart anomalies, or congenital heart defects, are structural irregularities in the heart that develop during fetal growth. Decoding these anomalies is a complex task. The success of ECG signal analysis relies heavily on selecting the right feature extraction methods and classification models, with particular emphasis on the feature extraction process. In response to this challenge, this study introduces innovative feature extraction techniques based on Holo-Hilbert spectral analysis, which can help to extract the envelope and carrier of signals. New features are extracted from the processed ECG signal to represent its characteristics. Various machine learning method are then employed to evaluate the classification performance of the signal as either arrhythmic or normal. Experimental results demonstrate that the proposed approach achieves accuracy on par with state-of-the-art methods in the literature, with the Light Gradient Boosting Machine (LGBM) model attaining an impressive classification accuracy of 98.4%.