In this study, we introduce a novel approach for generating explainable time-domain embeddings with practical applications in anti-spoofing logical-access (LA) tasks. These embeddings are derived from the evaluation of probability mass functions (PMFs) of the output waveforms obtained from diverse channels. Each element within these embeddings is discernible, facilitating comprehensive analyses of distinct attack types and their detection patterns based on specific frequency bands and similarity measures. Our investigation reveals that different filter bands and similarity measures exhibit superior discriminatory abilities between genuine speech and different attack modalities. Unlike most countermeasure systems, our proposed embeddings offer enhanced interpretability and explainability. The efficacy of our approach is demonstrated through various visualized examples, underscoring the significance of diverse channels and similarity measures in combating different attack scenarios.