Caractérisation des cris des nourrissons en vue du diagnostic précoce de différentes pathologies

Abstract

The use of newborn cry signals in diagnosis is based on many theories proposed lately. The main objective in these researches is the cry signals modeling and spectrographic analysis. It has been shown that the newborn cry acoustics are linked to particular medical conditions. This thesis is motivated by improvement of the accuracy of pathological cries recognition. This can be performed by the combination of several acoustic parameters from spectrographic analysis and parameters that describe the configuration of vocal tract and vocal folds. Acoustic characteristics representing the vocal tract were widely applied to the classification of the cries. However the usefulness of vocal folds characteristics in the automatic recognition, as well as their effective techniques extraction have not been exploited deeply. In this context, we have performed a qualitative characterization of healthy and pathologic newborns cries using characteristics that have been defined in the literature and which describe vocal tract and vocal folds behavior during the cry. This step allowed us to identify the most relevant features in the differentiation of the studied pathological cries. For the extraction of selected characteristics, we have implemented effective measure methods that avoid the overestimation and underestimation of characteristics. The proposed and used approach for characteristics quantification facilitates the automatic analysis of cries and allows efficient use of these features in the diagnostic system. We also conducted experimental tests for the validation of all proposed approaches in this thesis. The results are suitable and show an improvement of the cry-based pathology recognition. The work presented in this thesis is a collection of three articles published/submitted in various journals. Two other papers published in conferences are presented in the annexes.

Date	21 May 2014
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Chakib Tadj (Supervisor)