This study proposes a human authentication framework based on electrocardiogram signals that are robust to dynamic cardiac morphological conditions. The proposed method incorporates a stationary wavelet transform, an infinite feature selection, and a linear discriminant analysis. Evaluation experiments were conducted under three modulated situations: temporal variation, postural variation, and heart rate variation when exercising. Compared with three state‐of‐the‐art methods, the performance of the proposed method was shown to be better overall, with an equal error rate (EER) of 1.48% under time‐varying situations, 1.74% under posture changes, and 5.47% after exercise. These results indicate that the proposed method achieves a highly increased performance compared with state‐of‐the‐art techniques. Further evaluation of the identification performance of the proposed method on two public databases shows that it performs better than previously proposed methods.