Privacy Aware Acoustic Scene Synthesis Using Deep Spectral Feature Inversion

Gontier, Félix; Lagrange, Mathieu; Lavandier, Catherine; Petiot, Jean-François

doi:10.1109/icassp40776.2020.9053172

Cited by 9 publications

(5 citation statements)

References 15 publications

(17 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A 512-dimensional x-vector encoding the speaker is extracted using a TDNN trained on VoxCeleb-1,2 with Kaldi. In Step 2, for every source x-vector, an anonymized x-vector is computed by finding the N farthest xvectors in an external pool (LibriTTS train-other-500) according to the PLDA distance, and by averaging N * randomly selected vectors among them 5 . In Step 3, an SS AM generates Mel-filterbank features given the anonymized x-vector and the F0+BN features, and a neural source-filter (NSF) waveform model [35] outputs a speech signal given the anonymized xvector, the F0, and the generated Mel-filterbank features.…”

Section: Anonymization Baselinesmentioning

confidence: 99%

See 1 more Smart Citation

Introducing the VoicePrivacy Initiative

Tomashenko¹,

Srivastava²,

Wang³

et al. 2020

Interspeech 2020

116

View full text Add to dashboard Cite

The VoicePrivacy initiative aims to promote the development of privacy preservation tools for speech technology by gathering a new community to define the tasks of interest and the evaluation methodology, and benchmarking solutions through a series of challenges. In this paper, we formulate the voice anonymization task selected for the VoicePrivacy 2020 Challenge and describe the datasets used for system development and evaluation. We also present the attack models and the associated objective and subjective evaluation metrics. We introduce two anonymization baselines and report objective evaluation results.

show abstract

Section: Anonymization Baselinesmentioning

confidence: 99%

“…Current methods fall into four categories: deletion, encryption, distributed learning, and anonymization. Deletion methods [4,5] are meant for ambient sound analysis. They delete or obfuscate any overlapping speech to the point where no information about it can be recovered.…”

Section: Introductionmentioning

confidence: 99%

Introducing the VoicePrivacy Initiative

Tomashenko¹,

Srivastava²,

Wang³

et al. 2020

Interspeech 2020

116

View full text Add to dashboard Cite

show abstract

“…Deletion techniques aim at ambient sound analysis. When recording sound in public places, speech is obfuscated such that no information can be recovered [101]. This can be seen as a similar technique to blurring all faces in video surveillance in public places.…”

Section: A Privacy Preservationmentioning

confidence: 99%

Personal Voice Assistant Security and Privacy—A Survey

Cheng

Roedig

2022

Proc. IEEE

View full text Add to dashboard Cite

Personal voice assistants (PVAs) are increasingly used as interfaces to digital environments. Voice commands are used to interact with phones, smart homes, or cars. In the United States alone, the number of smart speakers, such as Amazon's Echo and Google Home, has grown by 78% to 118.5 million, and 21% of the U.S. population own at least one device. Given the increasing dependency of society on PVAs, security and privacy of these have become a major concern of users, manufacturers, and policy makers. Consequently, a steep increase in research efforts addressing security and privacy of PVAs can be observed in recent years. While some security and privacy research applicable to the PVA domain predates their recent increase in popularity, many new research strands have emerged. This article provides a survey of the state of the art in PVA security and privacy. The focus of this work is on the security and privacy challenges arising from the use of the acoustic channel. Work that describes both attacks and countermeasures is discussed. We highlight established areas such as voice authentication (VA) and new areas such as acoustic Denial of Service (DoS) that deserve more attention. This survey describes research areas where the threat is relatively well understood but where countermeasures are lacking, for example, in the area of hidden voice commands. We also discuss work that looks at privacy implications; for example, work on management of recording Manuscript

show abstract

“…In recent years, some methods of environmental sound synthesis using deep learning approaches have been developed [1]- [3]. Environmental sound synthesis has great potential for many applications such as supporting movie and game production [2], [4], and data augmentation for sound event detection and scene classification [5], [6]. As one of the methods of environmental sound synthesis, environmental sound synthesis using sound event labels as input to the system [1] has been proposed.…”

Section: Introductionmentioning

confidence: 99%

Onoma-to-wave: Environmental sound synthesis from onomatopoeic words

Okamoto¹,

Imoto²,

Takamichi³

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper, we propose a new framework for environmental sound synthesis using onomatopoeic words and sound event labels. The conventional method of environmental sound synthesis, in which only sound event labels are used, cannot finely control the time-frequency structural features of synthesized sounds, such as sound duration, timbre, and pitch. There are various ways to express environmental sound other than sound event labels, such as the use of onomatopoeic words. An onomatopoeic word, which is a character sequence for phonetically imitating a sound, has been shown to be effective for describing the phonetic feature of sounds. We believe that environmental sound synthesis using onomatopoeic words will enable us to control the fine time-frequency structural features of synthesized sounds, such as sound duration, timbre, and pitch. In this paper, we thus propose environmental sound synthesis from onomatopoeic words on the basis of a sequenceto-sequence framework. To convert onomatopoeic words to environmental sound, we use a sequence-to-sequence framework. We also propose a method of environmental sound synthesis using onomatopoeic words and sound event labels to control the fine time-frequency structure and frequency property of synthesized sounds. Our subjective experiments show that the proposed method achieves the same level of sound quality as the conventional method using WaveNet. Moreover, our methods are better than the conventional method in terms of the expressiveness of synthesized sounds to onomatopoeic words.

show abstract

Privacy Aware Acoustic Scene Synthesis Using Deep Spectral Feature Inversion

Cited by 9 publications

References 15 publications

Introducing the VoicePrivacy Initiative

Introducing the VoicePrivacy Initiative

Personal Voice Assistant Security and Privacy—A Survey

Onoma-to-wave: Environmental sound synthesis from onomatopoeic words

Contact Info

Product

Resources

About