Long-term Conversation Analysis: Exploring Utility and Privacy

TL;DR

This paper presents a privacy-preserving feature extraction method for long-term conversation analysis that balances utility and privacy using spectral smoothing and McAdams coefficient-based speaker anonymization.

Contribution

It introduces a novel combination of spectral smoothing and McAdams coefficient for privacy-preserving feature extraction in conversational data.

Findings

The method maintains utility in voice activity detection and speaker diarization.

It improves privacy protection in speech recognition and speaker verification.

The combination of techniques balances utility and privacy effectively.

Abstract

The analysis of conversations recorded in everyday life requires privacy protection. In this contribution, we explore a privacy-preserving feature extraction method based on input feature dimension reduction, spectral smoothing and the low-cost speaker anonymization technique based on McAdams coefficient. We assess the utility of the feature extraction methods with a voice activity detection and a speaker diarization system, while privacy protection is determined with a speech recognition and a speaker verification model. We show that the combination of McAdams coefficient and spectral smoothing maintains the utility while improving privacy.