Automated analysis of newborn cry: relationships between melodic shapes and native language

Recent research studies have shown that since the last trimester of pregnancy human fetuses are able to listen to and possibly memorize auditory stimuli from the external world, both as music and language are concerned. In particular, they exhibit a specific sensitivity to prosodic features such as melody, intensity, and rhythm that are essential for an infant to learn and develop the native language. This paper presents first results concerning the automated mother language identification of a set of about 7500 cry units coming from French, Arabic and Italian mother-tongue healthy full term newborns. Acoustical parameters and 12 different melodic shapes are computed with the BioVoice software tool and their classification is performed with Random Forest and 4 neuro-fuzzy classifiers. Results show up to 95% differences among the three languages.