Autonomous Surveillance of Infants’ Needs using Cnn-Deep Learning Model for Audio Cry Classification: Artificial Parenting

Abstract

Infants portray suggestive unique cries while sick, having belly pain, discomfort, tired, attention and need change of pumpers among other needs. There exists limited knowledge in accessing the infants’ needs as they only relay information through suggestive cries Levetown (2008). Many teenagers, tend to give birth at an early age, thereby exposing them to be the key monitors of their own babies (Heron et al., 2010). They tend not to have had sufficient skills in monitoring the infant’s dire needs, more so during the early stages of infant developments. Artificial intelligence has shown promising efficient predictive analytics from supervised, unsupervised to reinforcement learning models. This study therefore seeks to develop both android app and IOT gadget that could be used to discriminate the infant audio cries by leveraging the strength of convolution neural networks as a classifier model. Audio analytics from many literature is untapped area by researchers as it’s attributed to messy data, huge data generations which is manifested through 5Vs characteristics, such as velocity, variety, volume, veracity and variability (McAdam et al., 2010). This study therefore strongly leverages on convolution neural networks, deep learning model that is capable of handling the big data. To achieve this, the audio data inform of wave will be converted to images through melospectrum frequencies which will be classified using computer vision CNN model. The librosa library will be used to convert the audio to melospectrum upon which the CNN model will be used to extract the pixels using computer vision thus classifying the audio classes such sick, belly pain, tired, change of pumpers and discomfort. The study long term goal will be incorporated as a device that will be utilized at domestic level and hospital facilities for surveillance of the infant’s health and social needs’ status all time round.