Master 2014 2015
Stages de la spécialité SAR
Speech enhancement using deep neural networks

Site :Speech enhancement using deep neural networks
Lieu :Technicolor is an industry leader in the production of video content for movies, TV, advertising, games and more. The company provides production, postproduction, and distribution services to content creators, network service providers and broadcasters. Technicolor Research Rennes is the largest Technicolor Research Center conducting research in various domains applying to the creation, management and delivery of digital content. For more information on Technicolor R&I Rennes: The internship will be hosted in Imaging Science Lab (ISL) within Technicolor R&I Rennes. More specifically, he/she will join the “Image & Vision” team composing of more than twenty researchers and engineers coming from many different countries. One major goal of the team is to conduct the research towards describing and searching for contents from large data collections as well as enhancing and adapting these contents to any devices.
Encadrant : Ngoc Q. K. Duong, Alexey Ozerov, and Patrick Perez
Dates :du 01/03/2015 au 31/08/2015
Rémunération :1200 euros / month (brut)
Mots-clés : Parcours ATIAM : Acoustique, Parcours ATIAM : Informatique musicale, Parcours ATIAM : Traitement du signal


Speech enhancement aims at improving the intelligibility and/or overall perceptual quality of noisy speech signals. It is a key element in a wide range of applications in daily communication systems and robotics. Numerous approaches have been proposed in the literature exploiting different audio signal processing techniques. This challenging internship aims at exploiting an emerging framework based on deep neural networks (DNN), which has recently gained great success in different tasks in audio, image and language processing. Some recent studies [1,2] have shown the potential of this DNN-based approach for speech enhancement.

The intern will conduct both research and algorithm implemention while investigating the use of deep neural network for speech enhancement. Depending on obtained results, this work may lead to a patent submission and a research publication in an international conference.


[1] Y. Xu, J. Du, L. Dai, and C. Lee, “An Experimental Study on Speech Enhancement Based on Deep Neural Networks,” IEEE Signal Processing Letter, Vol. 21, No. 1, 2014.

[2] D. Liu, P. Smaragdis, and M. Kim, “Experiments on Deep Learning for Speech Denoising,” in Proceedings of the annual conference of the International Speech Communication Association (INTERSPEECH), 2014.