Automatic identification of qualitatives characteristics in infant cry

In infant cry analysis it is of great importance to identify the qualitative characteristics present in the cry wave, this is because they provide additional information that allows recognizing variations and similarities between normal and pathological cries. Nowadays the qualitative characteristics analysis is manually done by using visual perception (inspecting spectrograms) and by auditory perception (listening cry recordings). This perceptive analysis is the one expert physicians apply to help them make a diagnosis. In this work we present a method based in the definition of a threshold applied to the energy of the signal, this threshold allows to identify automatically cry units in a sample recording and another threshold that allows eliminating inspiratory cry segments. We also present a method that allows automatically identifying the melodic shape, shifts, glides and noise concentrations in cry units. The whole process implementation as well as some experiments and results are here presented.