Range-Based Equal Error Rate for Spoof Localization

TL;DR

This paper introduces a range-based equal error rate (EER) metric for spoof localization in audio, providing a more accurate evaluation method than traditional point-based EER by measuring misclassified ranges.

Contribution

The paper proposes a novel range-based EER metric for spoof localization, adapting the binary search algorithm for its calculation, improving performance evaluation accuracy.

Findings

Range-based EER offers a more accurate performance measure.

Range-based EER aligns better with actual misclassified ranges.

Proper temporal resolution is crucial for fair evaluation.

Abstract

Spoof localization, also called segment-level detection, is a crucial task that aims to locate spoofs in partially spoofed audio. The equal error rate (EER) is widely used to measure performance for such biometric scenarios. Although EER is the only threshold-free metric, it is usually calculated in a point-based way that uses scores and references with a pre-defined temporal resolution and counts the number of misclassified segments. Such point-based measurement overly relies on this resolution and may not accurately measure misclassified ranges. To properly measure misclassified ranges and better evaluate spoof localization performance, we upgrade point-based EER to range-based EER. Then, we adapt the binary search algorithm for calculating range-based EER and compare it with the classical point-based EER. Our analyses suggest utilizing either range-based EER, or point-based EER with a proper temporal resolution can fairly and properly evaluate the performance of spoof localization.