Automatic Classification of Infant's Cry Using Data Balancing and Hierarchical Classification Techniques


One of the fundamental characteristics of communication between human beings is the use of spoken language, an intricate tool, perfected over thousands of years. However, in the first years of our lives we still do not possess this tool to express ourselves, instead we use crying as a means of expressing hunger, pain or other sensations and feelings. The objective of this paper is to investigate whether the use of techniques to artificially increase the number of samples, together with a hierarchical reorganization of the database, can improve the automatic classification of infants' cries. To achieve this, data balancing techniques and hierarchical classification will be used.

Keywords Infant Cry Detection; Data Augmentation.

[full paper]