| 1 |
RETRIEVING SPEAKER INFORMATION FROM PERSONALIZED ACOUSTIC MODELS FOR SPEECH RECOGNITION
|
|
|
|
In: IEEE ICASSP 2022 ; https://hal.archives-ouvertes.fr/hal-03539741 ; IEEE ICASSP 2022, 2022, Singapour, Singapore (2022)
|
|
BASE
|
|
Show details
|
|
|
| 2 |
Evaluation of Speaker Anonymization on Emotional Speech ; Analyse de l'anonymisation du locuteur sur de la parole émotionnelle
|
|
|
|
In: JEP2022 - Journées d'Études sur la Parole ; https://hal.archives-ouvertes.fr/hal-03636737 ; JEP2022 - Journées d'Études sur la Parole, Jun 2022, Île de Noirmoutier, France (2022)
|
|
BASE
|
|
Show details
|
|
|
| 3 |
Utterance partitioning for speaker recognition: an experimental review and analysis with new findings under GMM-SVM framework
|
|
|
|
In: ISSN: 1381-2416 ; EISSN: 1572-8110 ; International Journal of Speech Technology ; https://hal.archives-ouvertes.fr/hal-03232723 ; International Journal of Speech Technology, Springer Verlag, In press, ⟨10.1007/s10772-021-09862-8⟩ (2021)
|
|
BASE
|
|
Show details
|
|
|
| 4 |
Speaker Attentive Speech Emotion Recognition
|
|
|
|
In: Proccedings of interspeech 2021 ; Interspeech 2021 ; https://hal.archives-ouvertes.fr/hal-03554368 ; Interspeech 2021, Aug 2021, Brno, Czech Republic. pp.2866-2870, ⟨10.21437/interspeech.2021-573⟩ (2021)
|
|
BASE
|
|
Show details
|
|
|
| 5 |
Privacy and utility of x-vector based speaker anonymization
|
|
|
|
In: https://hal.inria.fr/hal-03197376 ; 2021 (2021)
|
|
Abstract:
We study the scenario where individuals (speakers) contribute to the publication of an anonymized speech corpus. Data users then leverage this public corpus to perform downstream tasks (such as training automatic speech recognition systems), while attackers may try to de-anonymize itbased on auxiliary knowledge they collect. Motivated by this scenario, speaker anonymization aims to conceal the speaker identity while preserving the quality and usefulness of speech data. In this paper, we study x-vector based speaker anonymization, the leading approach in the recent Voice Privacy Challenge, which converts an input utterance into that of a random pseudo-speaker. We show that the strength of the anonymization varies significantly depending on how the pseudo-speaker is selected. In particular, we investigate four design choices: the distance measure between speakers, the region of x-vector space where the pseudo-speaker is mapped, the gender selection and whether to use speaker or utterance level assignment. We assess the quality of anonymization from the perspective of the three actors involved in our threat model, namely the speaker, the user and the attacker. To measure privacy and utility, we use respectively the linkability score achieved by the attackers and the decoding word error rate incurred by an ASR model trained with the anonymized data. Experiments on LibriSpeech dataset confirm that the optimal combination ofdesign choices yield state-of-the-art performance in terms of privacy protection as well as utility. Experiments on Mozilla Common Voice dataset show that the best design choices with 50 speakers guarantee the same anonymization level against re-identification attack as raw speech with 20,000 speakers.
|
|
Keyword:
[INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]; [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]; [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG]; linkability; privacy; speaker anonymization; speaker identification; speech recognition; utility
|
|
URL: https://hal.inria.fr/hal-03197376
|
|
BASE
|
|
Hide details
|
|
|
| 6 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 7 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 8 |
The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03332224 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 9 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 10 |
The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03332224 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 11 |
Enhancing Speech Privacy with Slicing
|
|
|
|
In: https://hal.inria.fr/hal-03369137 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 12 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 13 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 14 |
The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03332224 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 15 |
Privacy and utility of x-vector based speaker anonymization
|
|
|
|
In: https://hal.inria.fr/hal-03197376 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 16 |
Kurdish spoken dialect recognition using x-vector speaker embeddings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03262435 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
|
| 17 |
Evaluation of Speaker Anonymization on Emotional Speech
|
|
|
|
In: 1st ISCA Symposium on Security and Privacy in Speech Communication ; https://hal.inria.fr/hal-03377797 ; 1st ISCA Symposium on Security and Privacy in Speech Communication, Nov 2021, Virtual, Germany (2021)
|
|
BASE
|
|
Show details
|
|
|
| 18 |
An investigation into variability conditions in the SRE 2004 and 2008 Corpora ...
|
|
|
|
BASE
|
|
Show details
|
|
|
|
|
