Research: thesis

Efficient sound coding – Perception of music and speech: Contributions to the evaluation of the theory of “efficient coding” and analysis of its contributions to the study of deafness treated by cochlear implants

Health & well-living

The objective of this project is to evaluate hypotheses concerning the perceptual processing of music and speech.


The aim is to compare the results of perceptual evaluations and statistical analyses on natural signals and their “vocoded” correspondents (cochlear implant simulations).


This comparison will provide essential elements for a better understanding of the differences and similarities related to the perceptual representations required for listeners – whether people with normal hearing or deaf patients with cochlear implants – to perform perceptual processing of speech and music signals.


* MuseICs ( Mus_ique /E_fficient Coding / I_mplant _C_ochléaire / S_imulations).



This project is co-funded by the Région Pays de la Loire (West Creative Industries), the University of Nantes and the CNRS.


Objectives and Methodology

This project aims to develop innovative methods for the study of perceptual mechanisms in hearing. To do so, it compares the results of statistical analyses (Principal Component Analysis in particular), carried out on natural signals represented in spectral bands as well as on cochlear implant simulations, with the results of perceptual evaluations (perceptual  judgement tasks, measurement of surprise effects with respect to predictive expectations in oculometry and evoked potentials). These two corpora of results can be completed by psychophysical methods called “classification images”, which can be interesting to identify the internal structure of physical stimuli with respect to perceptual mechanisms.


From a theoretical perspective, the project will contribute to a better understanding of the nature of the perceptual representations required in the perception of similar physical characteristics in speech and music (melody, rhythm), and may provide essential elements for the study of the perceptual mechanisms involved in the analysis of vocoded signals and among patients with cochlear implants. It will also allow for extending this theoretical knowledge to the prosodic dimension of linguistic facts (melodic, rhythmic).




The modelling of the perceptual mechanisms resulting from the experiments carried out will make it possible to envisage suitable solutions for the processing of audio signals in cochlear implants. The project may also contribute to the development of tests for evaluating the perceptual processing of prosodic facts (melody/rhythm) in foreign languages or in contexts of language disorders.


Halfway through the project, it is planned to submit a collaborative ANR project on the problems of adaptation to variation in deaf patients with cochlear implants for the period 2021-2025.




Duration : three years, from 01/10/2019 to 30/09/2022.



Share on