Shane Grayson and Wilson Zhu, USA
New parents are frequently awakened by the cries of their newborn babies. Attempts to stop these cries sometimes result in increasingly louder cries. By first transforming these cries into waveforms, and then into sound spectrograms, the efficiencies and accuracies of different computer learning modules were tested: a support vector machine, a 2-layer neural network, and a long short-term memory model. Finally, an automatic sorter that categorizes each cry was developed. Using this method, it is possible to eliminate error and time wastage when trying to calm a baby. The results of testing the programs demonstrate a high accuracy rate for determining the source of a baby’s cries. This program will enable parents to calm their crying babies in a shorter amount of time, giving them more peace of mind, and perhaps allowing them to get more sleep.
Infant Cry, Deep Learning, Convolutional Neural Network, Audio Classification.