This shows you the differences between two versions of the page.

Link to this comparison view

automatic_classification_of_newborn_cry_melody [2014/03/15 12:59] (current)
bziolko created
Line 1: Line 1:
 +S. Orlandi, A. Bandini, S.D. Barbagallo, C. Manfredi (Italy)
 +**Automatic classification of newborn cry melody**
 +Institiutions: Dept. of Information Engineering, University of Firenze, Firenze, Italy - Dept. Electronic and Information Engineering “Guglielmo Marconi”, University of Bologna, Bologna, Italy
 +Fundamental frequency (F0) of infant’s cry exhibits different shapes in time domain that make up the melody of crying. Recent literature describes 4 main F0 shapes: rising, falling, symmetric and plateau. Prevalence or absence of some of these shapes may be related to neurological diseases. Objectives: The perceptual analysis of cry melody is time consuming and prone to errors due the huge irregularity of its shape. Thus, while being completely non-invasive and highly informative, it is rarely performed. Aim of this study is setting up an automatic system to identify and classify the melody of newborn infant cry in a short time. Methods: The procedure picks out cry-episodes from 2/3 minutes of crying recording and estimates F0 with a newly developed robust method. The four main F0 shapes are found with an iterative fitting procedure that closely resembles perceptive classification. The method is tested on synthesized cry signals. Results: The comparison between automatic and perceptual classification provides 97% of accuracy. The method is also tested on real cry signals coming from healthy newborns. Conclusions: This method may support clinical investigation in neonatal intensive care units as an aid to early diagnosis of neurological disorders being contact-less, cheap and non-invasive.
Copyright © XXII PVC Organizing Committee 2013. All Rights Reserved.