Long-Term Conversation Analysis: Privacy-Utility Trade-off under Noise and Reverberation

TL;DR

This paper investigates how noise and reverberation affect the privacy-utility balance in low-cost speech processing methods suitable for edge devices, highlighting noise's greater impact on privacy and utility.

Contribution

It provides a comprehensive analysis of privacy-utility trade-offs under realistic acoustic conditions using various low-cost privacy-preserving techniques.

Findings

Noise degrades speech recognition and speaker identification more than reverberation.

Some privacy methods maintain utility better under reverberation than noise.

Enhanced privacy correlates with increased speech recognition degradation.

Abstract

Recordings in everyday life require privacy preservation of the speech content and speaker identity. This contribution explores the influence of noise and reverberation on the trade-off between privacy and utility for low-cost privacy-preserving methods feasible for edge computing. These methods compromise spectral and temporal smoothing, speaker anonymization using the McAdams coefficient, sampling with a very low sampling rate, and combinations. Privacy is assessed by automatic speech and speaker recognition, while our utility considers voice activity detection and speaker diarization. Overall, our evaluation shows that additional noise degrades the performance of all models more than reverberation. This degradation corresponds to enhanced speech privacy, while utility is less deteriorated for some methods.