The analysis of the new-borns' cry using NEONAT and data mining techniques

The new-borns' cries are relevant for their health status. Their analysis may constitute a non-invasive diagnosis method. The paper presents a new tool (NEONAT) that allows digital processing of the vocal signal representing the cry, visualizing the frequency spectrum and several values of interest as well as managing the data regarding mothers and babies. A protocol for performing recordings is proposed. The application has a facility to export data in ARFF format to be processed by the algorithms supplied by the data mining tool WEKA. Testing the data mining opportunities was made building a sex classifier based on the information extracted from the cry. Its performances have been good, and based on that, a process for the detection of neurologic suffering was suggested. The application runs in the Obstetrics and Gynaecology Clinique in the Emergency County Hospital from Timisoara.