Skip to main content
. 2019 Oct 4;87(3):576–580. doi: 10.1038/s41390-019-0592-4

Fig. 1.

Fig. 1

Spectrograms from 5-s audio samples of each cry type showing the distribution of frequencies across time for four different infants. Acoustic features were used to train a machine learning algorithm to predict across three primary cry states: hungry, fussy, pain. This algorithm was tested on infant cries from colic to assess whether acoustic features of pain were present in cries from infants with parental-assessed colic