Inherent emotional feature extraction of neonatal cry

摘要

Identifying the meaning of the cries of a group that cannot speak is a world-concerned research topic. The current effective method is pattern recognition based on training and learning, but it often requires a large sample set and is limited to it. Therefore, a method of directly mining the inherent emotional feature which may not require too many samples and doesn’t have sample limitations is explored to identify the intention of neonatal cry in this study. Based on a database of 150 labeled neonatal audio, the minimum embedding dimension of neonatal cry is extracted as a nonlinear feature representing the activity of the nervous system and the emotion. It is found that the minimum embedding dimension of neonatal cry is higher in the state of pain and lower in the sad state. This result is consistent with related research of brain nervous activities under different emotions and crying characteristics in the state of pain. The extracted emotion-related parameter, which reflects the inherent physiological feature of the human body, can be used to identify and classify emotions by sounds. This technique helps to distinguish the emotions of people who can't speak and take care of them better. Furthermore, it provides a research approach of inherent feature extraction that is small-sample-set-based, non-learning, and can extract the inherent explainable stable feature.