12-month Postdoctoral research position in

machine learning for neural speech decoding

Place: GIPSA-lab (CNRS/UGA/Grenoble-INP) in collaboration with BrainTech laboratory (INSERM). Both

laboratories are located on the same campus in Grenoble, France.

Team: CRISSP team@GIPSA-lab (Cognitive Robotics, Interactive System and Speech processing).

Context:

This position is part of the ANR (French National Research Agency) BrainSpeak project aiming at developing a

Brain-Computer Interface (BCI) for speech rehabilitation, based on large-scale neural recordings. This post-doc

position aims at developing new machine learning algorithms to improve the conversion of neural signals into an

intelligible acoustic speech signal.

Mission:

Investigate deep learning approaches to map intracranial recordings (ECoG) to speech features (spectral,

articulatory, or linguistic features). A particular focus will be put on 1) weakly or self-supervised training in order to

deal with unlabeled, limited and sparse datasets, 2) introducing prior linguistic information for regularization (e.g.

thanks to a neural language model) and 3) online adaptation of the conversion model to cope with potential drift in

time of the neural responses.

Requirement and Profile:

• PhD in machine learning, signal/image/speech processing

• Advanced knowledges in deep learning

• Excellent programming skills (mostly Python)

• Fluent in English

Duration: 12 months

Salary (before tax) / Month €: Depending on the experience

Starting date: Early 2020

How to apply:

Send a cover letter, a resume, and references by email to:

• Dr. Thomas Hueber, thomas.hueber@gipsa-lab.fr

• Dr. Laurent Girin, laurent.girin@grenoble-inp.fr

• Dr. Blaise Yvert, blaise.yvert@inserm.fr

Applications will be processed as they arise.