Deep Learning Based Infant Cry Analysis Utilizing Computer Vision

摘要

Effort to understand infants' cry is crucial as crying is the main form of communication for infants and sensitively responding to their cry is closely related to his/her development. This paper proposes a method that classifies a video of a crying infant using deep learning techniques to effectively accomplish this task. Specifically, the method utilizes various audio feature extraction techniques (Mel Frequency Cepstral Coefficient and Short-Time Fourier Transfrom), and classification models such as autoencoder, deep residual network, and concatenate layer. The experiment results show that the proposed model obtains high accuracy in interpreting infants' cry compared to other machine learning based models.