Newborn infant's cry analysis

The very first cry or the birth cry of an infant carries significant information about the health of an infant and hence, it is considered as the vital parameter in deciding the Apgar count. As an infant grows, the cry acoustics changes with the integration of vocal tract system. Infants are found to produce many sounds apart from crying, which reflect the learning mechanism of the infants of the language spoken in his or her surroundings or the environment. Along with this, infants who have distinct cry sounds or who require large amount of stimulation to produce a cry, are found to be at risk of sudden infant death syndrome (SIDS) or possible neurological disorders. In this paper, newborn infant cries are analyzed using features derived from fundamental frequency (F (0)) contour or pitch contour, energy of the cry signal in different frequency sub-bands and unvoicing present in the infant's cry. For the extraction of fundamental frequency, modified autocorrelation method is used and shown to perform better than traditional autocorrelation-based method. To identify the significance of these features in identifying the reason of crying, ANOVA analysis is applied on these features. It is observed that the F (0) features are not of significance in the newborn cry analysis and presence of unvoicing in the infant's cry varies with the maturity of central nervous system (CNS) and is a discriminative feature of prime importance in newborn's cry analysis. In birth cries, the mean percentage of unvoicing is 84.4 % which drops to 67.7 % in normal infants (20 days-3 months). Birth cry analysis shows that there is very less voicing and hence, less vibration of the vocal folds.