Subject: 2-year positions on distant-mic ASR From: Emmanuel Vincent <emmanuel.vincent@xxxxxxxx> Date: Thu, 13 Nov 2014 10:25:02 +0100 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>Dear list, We are offering two positions on distant-mic ASR in the context of a large-scale R&D project on voice control of home appliances involving 5 companies and 3 academic labs. Successful candidates will perform research on state-of-the-art acoustic models and contribute to transfer their ideas into a commercial ASR product. * Research engineer (ideal start: Jan/Feb 2015 - duration: 2 years) - collaborate with a speech company on the DNN acoustic model setup - assess various multicondition training and speaker adaptation strategies for reverberant and noisy speech - design improved confidence measures * Postdoc (ideal start: Apr 2015 - duration: 2 years) - explore adaptation of DNNs to the reverberation conditions - explore adaptation of DNNs to noise by uncertainty propagation - design improved uncertainty estimators Salary: 2600 to 3300 €/month gross depending on experience, plus free health insurance and additional benefits Ideal profile: - PhD on speech processing, machine learning, audio signal processing, or applied statistics - proficient programming in C++, shell, Matlab/Python - optional experience with Kaldi or another ASR software To apply: send a CV, a motivation letter, a list of publications, and one or more recommendation letters to emmanuel.vincent@xxxxxxxx and irina.illina@xxxxxxxx Applications will be assessed on a continuous basis until Dec 12. Please apply as soon as possible before that date. References: [1] M. Ravanelli, M. Omologo, "On the selection of the impulse responses for distant-speech recognition based on contaminated speech training", in Proc. Interspeech, 2014. [2] B. Dumortier, E. Vincent, "Blind RT60 estimation robust across room sizes and source distances", in Proc. ICASSP, 2014. [3] M.L. Seltzer, D. Yu, Y. Wang, "An investigation of noise robustness of deep neural networks," in Proc. ICASSP, 2013. [4] B. Li, K.C. Sim, "An ideal hidden-activation mask for deep neural networks based noise-robust speech recognition," in Proc. ICASSP, 2014. [5] D.T. Tran, E. Vincent, D. Jouvet, “Fusion of multiple uncertainty estimators and propagators for noise robust ASR”, in Proc. ICASSP, 2014. -- Emmanuel Vincent PAROLE Project-Team Inria Nancy - Grand Est 615 rue du Jardin Botanique, 54600 Villers-lès-Nancy, France Phone: +33 3 8359 3083 - Fax: +33 3 8327 8319 Web: http://www.loria.fr/~evincent/