Perceptual Implications of Automatic Anonymization in Pathological Speech

2025-05-01

AI Summary

Automatic speech anonymization lacks empirical validation of perceptual fidelity for pathological speech, hindering its ethical use in clinical data sharing. Method: We conducted a human–machine Turing-style auditory experiment involving five pathological groups (e.g., cleft lip/palate, articulation disorders) and healthy controls, employing state-of-the-art anonymization models (EER 30–40%). We assessed perception via zero-/few-shot identification, repeated-measures ANOVA, and cross-group comparisons using diverse listener backgrounds. Contribution/Results: While anonymized speech retained high intelligibility (91–93%), perceptual quality dropped significantly (83% → 59%, *p* < 0.001). Degradation varied across disorder types (*p* = 0.005), and the native listener advantage vanished post-anonymization. Crucially, automated privacy metrics showed no correlation with human perception, revealing pathology-specific degradation patterns. These findings provide the first empirical basis for establishing perceptually grounded standards for pathological speech anonymization.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Privacy
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Large-scale security measurements

Abstract

Automatic anonymization techniques are essential for ethical sharing of pathological speech data, yet their perceptual consequences remain understudied. This study presents the first comprehensive human-centered analysis of anonymized pathological speech, using a structured perceptual protocol involving ten native and non-native German listeners with diverse linguistic, clinical, and technical backgrounds. Listeners evaluated anonymized-original utterance pairs from 180 speakers spanning Cleft Lip and Palate, Dysarthria, Dysglossia, Dysphonia, and age-matched healthy controls. Speech was anonymized using state-of-the-art automatic methods (equal error rates in the range of 30-40%). Listeners completed Turing-style discrimination and quality rating tasks under zero-shot (single-exposure) and few-shot (repeated-exposure) conditions. Discrimination accuracy was high overall (91% zero-shot; 93% few-shot), but varied by disorder (repeated-measures ANOVA: p=0.007), ranging from 96% (Dysarthria) to 86% (Dysphonia). Anonymization consistently reduced perceived quality (from 83% to 59%, p<0.001), with pathology-specific degradation patterns (one-way ANOVA: p=0.005). Native listeners rated original speech slightly higher than non-native listeners (Delta=4%, p=0.199), but this difference nearly disappeared after anonymization (Delta=1%, p=0.724). No significant gender-based bias was observed. Critically, human perceptual outcomes did not correlate with automatic privacy or clinical utility metrics. These results underscore the need for listener-informed, disorder- and context-specific anonymization strategies that preserve privacy while maintaining interpretability, communicative functions, and diagnostic utility, especially for vulnerable populations such as children.

Problem

• Evaluates perceptual effects of anonymizing pathological speech data
• Assesses listener discrimination accuracy across various speech disorders
• Examines quality degradation patterns in anonymized pathological speech

Innovation

• Human-centered analysis of anonymized pathological speech
• Turing-style discrimination and quality rating tasks
• Listener-informed disorder-specific anonymization strategies