Vous êtes ici : Accueil » Kiosque » Annonce


Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?


23 octobre 2017

Statistical learning applied to the discovery of sensory-motor strategies for postural regulation in equitherapy

Catégorie : Post-doctorant

Post-doc position in signal processing and machine learning

Title: Statistical learning applied to the discovery of sensory-motor strategies for postural regulation in equitherapy

Place: LITIS /CETAPS laboratories in Normandie Université, Rouen France

Schedule: 2 years between January 2018 and January 2020.

Wage: 2200€ per month

Contacts: romain.herault@insa-rouen.fr


  • Signal processing
  • Machine/statistical learning (if possible applied to signal processing)
  • Experience in dictionnary learning or deep learning will be a plus
  • Scientific programming (matlab/scipy/tensorflow …)
  • Fluent English

Overall, this project focuses on the study of postural regulation in equitherapy in order to analyze the perceptual-motor skills involved in training, learning and / or rehabilitation. In order to do so we have recorded 3D movement of non pathological and pathological subjects that must follow a postural regulation task in equitherapy. A mechanical horse is oscillating at different frequencies chosen by the experimenter. The success of the therapy is measured by the difference between the oscillations of the body and the horse.

The objective of the project is to extract from this great mass of signals postural patterns related to the different pathologies. In order to achieve this goal, we aim to establish a typology of the different modes of segmental organization of postural regulation in relation to the pathology and the experimental environment (such as the frequency), and learn to recognize the belonging of signal segment to a given typology. The main scientific lock that we want to address is the ability to segment and classify/cluster signals at the same time. Both supervised task (assign an example to given pathologies or typology) and unsupervised task (clustering) will be undertaken on signal segments in order to discover postural patterns that may be unknown to experts. Among the possible paths that we envisage to solve these problems one can cite segmentation based on local variance, dictionary learning or neural network adapted to sequence processing.

Therefore, the post-doc will be involved into two sub-tasks:

  1. Signal segmentation and classification/clustering.
  2. 3D Movement variance analysis (through methods such as local variance or weighted current)

This project is part of a greater scientific project called DAISI which has been cofunded by the European Union with the European Regional Development Fund (ERDF), by the French Agence Nationale de la Recherche and by the Regional Council of Normandie.


Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2018.