Cry-Based Infant Pathology
Classification Using GMMs
H. F. Alaie, L. Abou-Abbas, C. Tadj
Traditional
studies of infant’s cry signal focus more on non-pathological-based
classification of infants. In this article, we introduce an inexpensive
health care system which carries out acoustic analysis of the unclean
noisy infants’ cry signals to extract and measure certain cry
characteristics quantitatively and classifies healthy
and sick newborn infants only according to their cries. In the task of
this newborn cry-based diagnostic system, the dynamic MFCC feature
along with static Mel-Frequency Cepstal Coefficients (MFCCs) are selected and extracted for both
expiratory and inspiratory cry vocalizations in order to make the most
discriminative and informative feature vector. And then, we create a
unique cry-pattern for each cry vocalization type and pathological
condition by introducing a novel idea using Boosting Mixture Learning (BML) method to derive
either healthy or pathological subclass models separately from the Gaussian Mixture
Model-Universal Background Model (GMM-UBM). Our newborn cry-based
diagnostic system (NCDS) has a hierarchical scheme which is a treelike
combination of individual classifiers. Moreover, a score level fusion
of the proposed expiratory and inspiratory cry-based subsystems is
performed to make more reliable decision. The experimental results indicate that the adapted BML method has
lower error rates than Bayesian or maximum a posteriori probability (MAP) adaptation as a reference method.
Keywords:
Gaussian mixture models,
Universal background model, on Mel Frequency Cepstral Coefficients,
likelihood ratio scores, newborn infant cries, expiratory and
inspiratory sounds