A Robust Tool for Newborn Infant Cry Analysis

In this paper, a new robust adaptive tool for new- born infant cry analysis is proposed, characterised by high tracking capability, well suited for the signals under study. It performs F0, noise and resonance frequencies tracking, on signal frames of varying length (even few ms), adaptively tailored to varying signal characteristics. Moreover, voiced/unvoiced separation is implemented, allowing disregarding unvoiced parts of the signal where misleading results could be obtained. Plots of F0 and its harmonics, noise tracking, spectrogram with resonance frequencies superimposed, are presented in a coloured-scale. Some added statistics allow further understanding and comparison of results. The new software tool is completely automatic, working with any sampling frequency and F0, and also with strongly corrupted signals, and does not need any manual setting of whatever option to be made by the user, thus being easily usable also by non-experts. Some examples are reported, concerning both healthy and pathological new-born infant cries.