APEX: Audio Prototype EXplanations for Classification Tasks

2026-05-11

AI Summary

This work addresses the lack of interpretable methods tailored to the acoustic characteristics of audio classification, a domain where visual attribution techniques are often naively applied. To bridge this gap, the authors propose APEX, a post-hoc explanation framework that requires no fine-tuning of the original model. APEX introduces multi-view prototype reasoning to the audio domain for the first time, leveraging four distinct prototype types—time-domain, frequency-domain, time-frequency joint, and local patch—to disentangle and explain input signals. By generating instance-level explanations grounded in acoustic similarity while preserving the model’s original predictions, APEX significantly enhances semantic clarity and alignment with human auditory perception, outperforming conventional gradient-based attribution methods.

Abstract

Explainable AI (XAI) has achieved remarkable success in image classification, yet the audio domain lacks equally mature solutions. Current methods apply vision-based attribution techniques to spectrograms, overlooking fundamental differences between visual and acoustic signals. While prototype reasoning is promising, acoustic similarity remains multidimensional. We introduce APEX (Audio Prototype EXplanations), a post-hoc framework for interpreting pre-trained audio classifiers. Crucially, APEX requires no fine-tuning of the original backbone and strictly preserves output invariance. APEX disentangles explanations into four perspectives: Square-based prototypes to localize transient events, Time-based for temporal patterns, Frequency-based highlighting spectral bands, and Time-Frequency-based integrating both. This yields intuitive, example-based explanations that respect acoustic properties, providing greater semantic clarity than standard gradient-based methods.

Problem

• Explainable AI
• audio classification
• prototype reasoning
• acoustic similarity
• spectrogram interpretation

Innovation

• Audio Explainable AI
• Prototype-based Explanation
• Post-hoc Interpretability
• Time-Frequency Disentanglement
• Output Invariance