Digital Speech Algorithms for Speaker De-Identification

TL;DR

This paper evaluates four voice modification algorithms to determine their effectiveness in speaker de-identification by analyzing pitch changes, quality, and reversibility, contributing to privacy preservation in multimedia content.

Contribution

It introduces an assessment framework for voice modification algorithms focusing on de-identification effectiveness, quality, and reversibility.

Findings

Voice modifications significantly alter pitch to confuse gender recognition

Some algorithms produce reversible changes, raising privacy concerns

The study provides insights into the balance between de-identification and speech quality

Abstract

The present work is based on the COST Action IC1206 for De-identification in multimedia content. It was performed to test four algorithms of voice modifications on a speech gender recognizer to find the degree of modification of pitch when the speech recognizer have the probability of success equal to the probability of failure. The purpose of this analysis is to assess the intensity of the speech tone modification, the quality, the reversibility and not-reversibility of the changes made.