Exploratory Evaluation of Speech Content Masking

TL;DR

This paper explores a new approach called content masking to protect speech content privacy by concealing specific words or phrases, evaluating different masking techniques and their effects on speech recognition and speaker verification.

Contribution

It introduces the concept of content masking in speech privacy, evaluates baseline masking methods using VQ-VAE and WaveRNN, and analyzes their impact on downstream speech tasks.

Findings

Masking affects automatic speech recognition accuracy.

Different masking strategies influence speaker verification performance.

Masking location and type significantly impact privacy and utility balance.

Abstract

Most recent speech privacy efforts have focused on anonymizing acoustic speaker attributes but there has not been as much research into protecting information from speech content. We introduce a toy problem that explores an emerging type of privacy called "content masking" which conceals selected words and phrases in speech. In our efforts to define this problem space, we evaluate an introductory baseline masking technique based on modifying sequences of discrete phone representations (phone codes) produced from a pre-trained vector-quantized variational autoencoder (VQ-VAE) and re-synthesized using WaveRNN. We investigate three different masking locations and three types of masking strategies: noise substitution, word deletion, and phone sequence reversal. Our work attempts to characterize how masking affects two downstream tasks: automatic speech recognition (ASR) and automatic speaker verification (ASV). We observe how the different masks types and locations impact these downstream tasks and discuss how these issues may influence privacy goals.