Convolutional Neural Network Based Speaker De-Identification

Abstract

Concealing speaker identity in speech signals refers to the task of speaker de-identification, which helps protect the privacy of a speaker. Although, both linguistic and paralinguistic features reveal personal information of a speaker and they both need to be addressed, in this study we only focus on speaker voice characteristics. In other words, our goal is to move away from the source speaker identity while preserving naturalness and quality. The proposed speaker de-identification system maps voice of a given speaker to an average (or gender-dependent average) voice; the mapping is modeled by a new convolutional neural network (CNN) encoder-decoder architecture. Here, the transformation of both spectral and excitation features are studied. The voice conversion challenge 2016 (VCC-2016) database is used to train the system and examine performance of the proposed method. We use two different approaches for evaluations:(1) objective evaluation: equal error rates (EERs) calculated by an i-vector/PLDA speaker recognition system range between 1.265 -3.46 % on average for all developed systems, and (2) subjective evaluation: achieved 2.8 naturalness mean opinion score (MOS). Both objective and subjective experiments confirm the effectiveness of our proposed de-identification method.

Evaluating X-Vector-Based Speaker Anonymization Under White-Box Assessment

“…The anonymization of speaker's identity can be performed with various methods [1,3,7,12,15]. In this article, our contributions are based on the Baseline-1 (referred to as the baseline in this article) of the VoicePrivacy challenge that anonymizes speech using x-vectors and neural waveform models [3].…”

Preventing Fake Information Generation Against Media Clone Attacks

“…For speech, several classic approaches for speaker anonymization are explored in [31]- [34], while a neural-network-based encoder-decoder approach has been proposed in [35].…”

Speaker Anonymization Using X-vector and Neural Waveform Models

“…For logical anonymization, Jin et al [11] presented a voice transformation (VT) system to change the speaker identity into another special speaker. Similarly, Bahmaninezhad et al [12] utilized a convolutional neural network (CNN) as a VT function and averaged different transformation results as a means to anonymize speech. Magarienos et al [13] and Pobar and Ipšić [14] improved the convenience of the VT-based method to enable the user to select an approximate transformation from a pool of pre-trained VT models for speaker anonymization.…”