“Cry Baby”: Using Spectrographic Analysis to Assess Neonatal Health Status from an Infant’s Cry

Abstract

Infant cry analysis is a multidisciplinary area of research incorporating­ ­pediatrics, neurology, physiology, engineering, developmental linguistics, and ­psychology. It has been proposed in the pediatric literature that the infant cry is a reflection of complex neurophysiologic functions and that analysis of the cry itself can be used to assess the status of the infant’s health. Given the diagnostic importance of infant cry, this chapter presents application of spectrographic analysis to the vocal sounds of an infant, comparing normal with abnormal infant cry. Drawing from a rich body of research on spectrographic analysis predominantly used for performance of speaker recognition, this chapter presents how such spectral features that are used to identify and verify speakers can be applied to assess the neonate’s health status, by comparing a normal to an abnormal cry. Ten distinct cry modes, viz., hyperphonation, dysphonation, inhalation, double harmonic break, trailing, vibration, weak vibration, flat, rising, and falling have been identified for normal infant cry and their spectrographic patterns were observed. This analysis was then extended to abnormal infant cry. It has been observed that the double harmonic break is more dominant for ­abnormal infant cry in cases of myalgia (muscular pain). The inhalation pattern is distinct for infants suffering from asthma or other respiratory ailments such as a cough or cold. For example, for the infant whose larynx is not well developed, the pitch harmonics are nearly absent. As such, there are no voicing or glottal vibrations in the cry signal. In addition, for infants with Hypoxic Ischemic Encephalopathy (HIE), there is an ­initial tendency of pitch harmonics­ to rise and then to be followed by a blurring of such ­harmonics. Finally, an infant cry classification system is analyzed by observing the nature of the optimal ­warping path in the Dynamic Time Warping (DTW) algorithm.