Revisiting the Privacy of Low-Frequency Speech Signals: Exploring Resampling Methods, Evaluation Scenarios, and Speaker Characteristics

TL;DR

This paper investigates how low-frequency audio resampling and anti-aliasing filtering affect speech privacy and utility, revealing that anti-aliasing filters are crucial for privacy protection without severely impacting speech recognition performance.

Contribution

It provides a comparative analysis of resampling methods and evaluates privacy and utility impacts across different speaker characteristics and filtering techniques.

Findings

Anti-aliasing filtering significantly enhances speech privacy.

Sampling at 800 Hz retains most speech intelligibility.

Speaker sex and pitch influence privacy and utility outcomes.

Abstract

While audio recordings in real life provide insights into social dynamics and conversational behavior, they also raise concerns about the privacy of personal, sensitive data. This article explores the effectiveness of restricting recordings to low-frequency audio to protect spoken content. For resampling the audio signals to different sampling rates, we compare the effect of employing anti-aliasing filtering. Privacy enhancement is measured by an increased word error rate of automatic speech recognition models. The impact on utility performance is measured with voice activity detection models. Our experimental results show that for clean recordings, models trained with a sampling rate of up to 800 Hz transcribe the majority of words correctly. For both models, we analyzed the impact of the speaker's sex and pitch, and we demonstrated that missing anti-aliasing filters more strongly compromise speech privacy.