no code implementations • 15 Mar 2024 • Peter Leer, Jesper Jensen, Laurel Carney, Zheng-Hua Tan, Jan Østergaard, Lars Bramsløw
In this study, we propose a DNN-based approach for hearing-loss compensation, which is trained on the outputs of hearing-impaired and normal-hearing DNN-based auditory models in response to speech signals.
1 code implementation • 15 Mar 2024 • Peter Leer, Jesper Jensen, Zheng-Hua Tan, Jan Østergaard, Lars Bramsløw
Our results show that this new optimization objective significantly improves the emulation performance of deep neural networks across relevant input sound levels and auditory-model frequency channels, without increasing the computational load during inference.
1 code implementation • 8 Mar 2024 • Vikas Tokala, Eric Grinstein, Mike Brookes, Simon Doclo, Jesper Jensen, Patrick A. Naylor
Studies have shown that in noisy acoustic environments, providing binaural signals to the user of an assistive listening device may improve speech intelligibility and spatial awareness.
no code implementations • 17 Jan 2024 • Iván López-Espejo, Aditya Joglekar, Antonio M. Peinado, Jesper Jensen
Pre-emphasis filtering, compensating for the natural energy decay of speech at higher frequencies, has been considered as a common pre-processing step in a number of speech processing tasks over the years.
no code implementations • 27 Dec 2023 • Holger Severin Bovbjerg, Jesper Jensen, Jan Østergaard, Zheng-Hua Tan
Our experiments show that self-supervised pretraining not only improves performance in clean conditions, but also yields models which are more robust to adverse conditions compared to purely supervised learning.
no code implementations • 7 Dec 2023 • Philippe Gonzalez, Zheng-Hua Tan, Jan Østergaard, Jesper Jensen, Tommy Sonne Alstrøm, Tobias May
To address this, we extend this framework to account for the progressive transformation between the clean and noisy speech signals.
no code implementations • 5 Dec 2023 • Philippe Gonzalez, Zheng-Hua Tan, Jan Østergaard, Jesper Jensen, Tommy Sonne Alstrøm, Tobias May
We show that the proposed system substantially benefits from using multiple databases for training, and achieves superior performance compared to state-of-the-art discriminative models in both matched and mismatched conditions.
no code implementations • 20 Sep 2023 • Andreas J. Fuglsig, Jesper Jensen, Zheng-Hua Tan, Lars S. Bertelsen, Jens Christian Lindof, Jan Østergaard
Results show that the joint optimization can further improve performance compared to the concatenated approach.
no code implementations • 1 Jun 2023 • Juan F. Montesinos, Daniel Michelsanti, Gloria Haro, Zheng-Hua Tan, Jesper Jensen
Audio and visual modalities are inherently connected in speech signals: lip movements and facial expressions are correlated with speech sounds.
1 code implementation • 1 Mar 2023 • Matthias Blochberger, Filip Elvander, Randall Ali, Jan Østergaard, Jesper Jensen, Marc Moonen, Toon van Waterschoot
Distributed signal-processing algorithms in (wireless) sensor networks often aim to decentralize processing tasks to reduce communication cost and computational complexity or avoid reliance on a single device (i. e., fusion center) for processing.
no code implementations • 19 Nov 2022 • Iván López-Espejo, Ram C. M. C. Shekar, Zheng-Hua Tan, Jesper Jensen, John H. L. Hansen
In the context of keyword spotting (KWS), the replacement of handcrafted speech features by learnable features has not yielded superior KWS performance.
no code implementations • 31 Oct 2022 • Andreas Jonas Fuglsig, Jesper Jensen, Zheng-Hua Tan, Lars Søndergaard Bertelsen, Jens Christian Lindof, Jan Østergaard
The intelligibility and quality of speech from a mobile phone or public announcement system are often affected by background noise in the listening environment.
no code implementations • 20 Nov 2021 • Iván López-Espejo, Zheng-Hua Tan, John Hansen, Jesper Jensen
Spoken keyword spotting (KWS) deals with the identification of keywords in audio streams and has become a fast-growing technology thanks to the paradigm shift introduced by deep learning a few years ago.
Automatic Speech Recognition Automatic Speech Recognition (ASR) +2
no code implementations • 15 Nov 2021 • Andreas Jonas Fuglsig, Jan Østergaard, Jesper Jensen, Lars Søndergaard Bertelsen, Peter Mariager, Zheng-Hua Tan
However, the existing optimal mutual information based method requires a complicated system model that includes natural speech variations, and relies on approximations and assumptions of the underlying signal distributions.
no code implementations • 9 Oct 2020 • Giovanni Morrone, Daniel Michelsanti, Zheng-Hua Tan, Jesper Jensen
In this paper, we present a deep-learning-based framework for audio-visual speech inpainting, i. e., the task of restoring the missing parts of an acoustic speech signal from reliable audio context and uncorrupted visual information.
1 code implementation • 21 Aug 2020 • Daniel Michelsanti, Zheng-Hua Tan, Shi-Xiong Zhang, Yong Xu, Meng Yu, Dong Yu, Jesper Jensen
Speech enhancement and speech separation are two related tasks, whose purpose is to extract either one or more target speech signals, respectively, from a mixture of sounds generated by several sources.
no code implementations • 30 May 2020 • Iván López-Espejo, Zheng-Hua Tan, Jesper Jensen
Despite their great performance over the years, handcrafted speech features are not necessarily optimal for any particular speech application.
no code implementations • 6 Apr 2020 • Daniel Michelsanti, Olga Slizovskaia, Gloria Haro, Emilia Gómez, Zheng-Hua Tan, Jesper Jensen
Both acoustic and visual information influence human perception of speech.
no code implementations • 3 Sep 2019 • Morten Kolbæk, Zheng-Hua Tan, Søren Holdt Jensen, Jesper Jensen
Finally, we show that a loss function based on scale-invariant signal-to-distortion ratio (SI-SDR) achieves good general performance across a range of popular speech enhancement evaluation metrics, which suggests that SI-SDR is a good candidate as a general-purpose loss function for speech enhancement systems.
no code implementations • 22 Jun 2019 • Iván López-Espejo, Zheng-Hua Tan, Jesper Jensen
Our results show that this multi-task deep residual network is able to achieve a KWS accuracy relative improvement of around 32% with respect to a system that does not deal with external speakers.
no code implementations • 29 May 2019 • Daniel Michelsanti, Zheng-Hua Tan, Sigurdur Sigurdsson, Jesper Jensen
Regarding speech intelligibility, we find a general tendency of the benefit in training the systems with Lombard speech.
no code implementations • 15 Nov 2018 • Daniel Michelsanti, Zheng-Hua Tan, Sigurdur Sigurdsson, Jesper Jensen
Audio-visual speech enhancement (AV-SE) is the task of improving speech quality and intelligibility in a noisy environment using audio and visual information from a talker.
no code implementations • 15 Nov 2018 • Daniel Michelsanti, Zheng-Hua Tan, Sigurdur Sigurdsson, Jesper Jensen
Humans tend to change their way of speaking when they are immersed in a noisy environment, a reflex known as Lombard effect.
no code implementations • 2 Feb 2018 • Morten Kolbæk, Zheng-Hua Tan, Jesper Jensen
Finally, we show that the proposed SE system performs on par with a traditional DNN based Short-Time Spectral Amplitude (STSA) SE system in terms of estimated speech intelligibility.
Sound Audio and Speech Processing
no code implementations • 31 Aug 2017 • Morten Kolbæk, Dong Yu, Zheng-Hua Tan, Jesper Jensen
We show that deep bi-directional LSTM RNNs trained using uPIT in noisy environments can improve the Signal-to-Distortion Ratio (SDR) as well as the Extended Short-Time Objective Intelligibility (ESTOI) measure, on the speaker independent multi-talker speech separation and denoising task, for various noise types and Signal-to-Noise Ratios (SNRs).
Sound
3 code implementations • 18 Mar 2017 • Morten Kolbæk, Dong Yu, Zheng-Hua Tan, Jesper Jensen
We evaluated uPIT on the WSJ0 and Danish two- and three-talker mixed-speech separation tasks and found that uPIT outperforms techniques based on Non-negative Matrix Factorization (NMF) and Computational Auditory Scene Analysis (CASA), and compares favorably with Deep Clustering (DPCL) and the Deep Attractor Network (DANet).
1 code implementation • 1 Jul 2016 • Dong Yu, Morten Kolbæk, Zheng-Hua Tan, Jesper Jensen
We propose a novel deep learning model, which supports permutation invariant training (PIT), for speaker independent multi-talker speech separation, commonly known as the cocktail-party problem.