Guillaume Fuchs

Fraunhofer Institute for Integrated Circuits

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Markus Multrus

Fraunhofer Institute for Integrated Circuits

Nicola Pia

Fraunhofer Institute for Integrated Circuits

Srikanth Korse

International Audio Laboratories Erlangen

Max Neuendorf

Fraunhofer Institute for Integrated Circuits

Kishan Chand Gupta

Jadavpur University

Christian R. Helmrich

Fraunhofer Institute for Telecommunications, Heinrich Hertz Institute

Sascha Disch

Fraunhofer Institute for Integrated Circuits

Jérémie Lecomte

Fraunhofer Institute for Integrated Circuits

Kishan Gupta

Fraunhofer Institute for Integrated Circuits

Ralf Geiger

International Audio Laboratories Erlangen

Cooperative Institutions

Fraunhofer Institute for Integrated Circuits

Friedrich-Alexander-Universität Erlangen-Nürnberg

International Audio Laboratories Erlangen

Fraunhofer Society

Université de Sherbrooke

Leibniz University Hannover

Technical University of Munich

VoiceAge (Canada)

NTT (Japan)

University of Bonn

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

A Streamwise Gan Vocoder for Wideband Speech Coding at Very Low Bit Rate

Ahmed Mustafa Jan Büthe Srikanth Korse Kishan Gupta Guillaume Fuchs

Recently, GAN vocoders have seen rapid progress in speech synthesis, starting to outperform autoregressive models in perceptual quality with much higher generation speed. However, autoregressive vocoders are still the common choice for neural generation of speech signals coded at very low bit rates. In this paper, we present a GAN vocoder which is able to generate wideband speech waveforms from parameters coded at 1.6 kbit/s. The proposed model is a modified version of the StyleMelGAN vocoder that can run in frame-by-frame manner, making it suitable for streaming applications. The experimental results show that the proposed model significantly outperforms prior autoregressive vocoders like LPC-Net for very low bit rate speech coding, with computational complexity of about 5 GMACs, providing a new state of the art in this domain. Moreover, this streamwise adversarial vocoder delivers quality competitive to advanced speech codecs such as EVS at 5.9 kbit/s on clean speech, which motivates further usage of feedforward fully-convolutional models for low bit rate speech coding.

Codec2

10.1109/waspaa52581.2021.9632750

Cite

Citations (12)

A Streamwise GAN Vocoder for Wideband Speech Coding at Very Low Bit Rate

arXiv (Cornell University) (2021)

Ahmed Mustafa Jan Büthe Srikanth Korse Kishan Chand Gupta Guillaume Fuchs

Recently, GAN vocoders have seen rapid progress in speech synthesis, starting to outperform autoregressive models in perceptual quality with much higher generation speed. However, autoregressive vocoders are still the common choice for neural generation of speech signals coded at very low bit rates. In this paper, we present a GAN vocoder which is able to generate wideband speech waveforms from parameters coded at 1.6 kbit/s. The proposed model is a modified version of the StyleMelGAN vocoder that can run in frame-by-frame manner, making it suitable for streaming applications. The experimental results show that the proposed model significantly outperforms prior autoregressive vocoders like LPCNet for very low bit rate speech coding, with computational complexity of about 5 GMACs, providing a new state of the art in this domain. Moreover, this streamwise adversarial vocoder delivers quality competitive to advanced speech codecs such as EVS at 5.9 kbit/s on clean speech, which motivates further usage of feed-forward fully-convolutional models for low bit rate speech coding.

Wideband audio

10.48550/arxiv.2108.04051

Cite

Citations (0)

A new AMS beam line at the Erlangen tandem accelerator facility

Nuclear Instruments and Methods in Physics Research Section B Beam Interactions with Materials and Atoms (1990)

M. Baumgärtner U. Emmerling Werner Ernst E. Finckh Guillaume Fuchs

Faraday cup

Accelerator mass spectrometry

Tandem

Tandem accelerator

10.1016/0168-583x(90)90370-a

Cite

Citations (13)

Fundamental Frequency Model for Postfiltering at Low Bitrates in a Transform-Domain Speech and Audio Codec

Interspeech 2022 (2020)

Sneha Das Tom Bäckström Guillaume Fuchs

Speech codecs can use postfilters to improve the quality of the decoded signal.While postfiltering is effective in reducing coding artifacts, such methods often involve processing in both the encoder and the decoder, rely on additional transmitted side information, or are highly dependent on other codec functions for optimal performance.We propose a low-complexity postfiltering method to improve the harmonic structure of the decoded signal, which models the fundamental frequency of the signal.In contrast to past approaches, the postfilter operates at the decoder as a standalone function and does not need the transmission of additional side information.It can thus be used to enhance the output of any codec.We tested the approach on a modified version of the EVS codec in TCX mode only, which is subject to more pronounced coding artefacts when used at its lowest bitrate.Listening test results show an average improvement of 7 MUSHRA points for decoded signals with the proposed harmonic postfilter. 1

Codec

Sub-band coding

10.21437/interspeech.2020-1067

Cite

Citations (2)

MPEG Unified Speech and Audio Coding – Bridging the Gap

Springer eBooks (2011)

Markus Multrus Max Neuendorf Jérémie Lecomte Guillaume Fuchs Stefan Bayer

Codec2

Sub-band coding

Codec

Bandwidth extension

10.1007/978-3-642-23071-4_33

Cite

Citations (2)

NESC: Robust Neural End-2-End Speech Coding with GANs

arXiv (Cornell University) (2022)

Nicola Pia Kishan Chand Gupta Srikanth Korse Markus Multrus Guillaume Fuchs

Neural networks have proven to be a formidable tool to tackle the problem of speech coding at very low bit rates. However, the design of a neural coder that can be operated robustly under real-world conditions remains a major challenge. Therefore, we present Neural End-2-End Speech Codec (NESC) a robust, scalable end-to-end neural speech codec for high-quality wideband speech coding at 3 kbps. The encoder uses a new architecture configuration, which relies on our proposed Dual-PathConvRNN (DPCRNN) layer, while the decoder architecture is based on our previous work Streamwise-StyleMelGAN. Our subjective listening tests on clean and noisy speech show that NESC is particularly robust to unseen conditions and signal perturbations.

Codec

Wideband audio

End-to-end principle

Robustness

10.48550/arxiv.2207.03282

Cite

Citations (0)

A robust speech/music discriminator for switched audio coding

Zenodo (CERN European Organization for Nuclear Research) (2015)

Guillaume Fuchs

Discriminator

10.5281/zenodo.35900

Cite

Citations (0)

A Novel Scheme for Low Bitrate Unified Speech and Audio Coding – MPEG RM0

Journal of the Audio Engineering Society (2009)

Max Neuendorf Philippe Gournay Markus Multrus Jérémie Lecomte B. Bessette

MPEG-4

Constant bitrate

Source

Cite

Citations (18)

StyleMelGAN: An Efficient High-Fidelity Adversarial Vocoder with Temporal Adaptive Normalization

ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2021)

Ahmed Mustafa Nicola Pia Guillaume Fuchs

In recent years, neural vocoders have surpassed classical speech generation approaches in naturalness and perceptual quality of the synthesized speech. Computationally heavy models like WaveNet and WaveGlow achieve best results, while lightweight GAN models, e.g. MelGAN and Parallel WaveGAN, remain inferior in terms of perceptual quality. We therefore propose StyleMelGAN, a lightweight neural vocoder allowing synthesis of high-fidelity speech with low computational complexity. StyleMelGAN employs temporal adaptive normalization to style a low-dimensional noise vector with the acoustic features of the target speech. For efficient training, multiple random-window discriminators adversarially evaluate the speech signal analyzed by a filter bank, with regularization provided by a multi-scale spectral reconstruction loss. The highly parallelizable speech generation is several times faster than real-time on CPUs and GPUs. MUSHRA and P.800 listening tests show that StyleMelGAN outperforms prior neural vocoders in copy-synthesis and Text-to-Speech scenarios.

Normalization

Intelligibility (philosophy)

Regularization

10.1109/icassp39728.2021.9413605

Cite

Citations (45)

A Scalable CELP/Transform Coder for Low Bit Rate Speech and Audio Coding

Journal of the Audio Engineering Society (2006)

Guillaume Fuchs Roch Lefebvre

Codec2

Source

Cite

Citations (1)