Child Cry Classification - An Analysis of Features and Models

This paper presents a study on the classification of child cries based on various features extracted through speech and auditory processing. Certain spectral and descriptive features vary significantly in a child's cry intended for a specific purpose. Firstly, the model was trained using individual features. Later, the best features were selected and the model was again trained by combining these features. Logistic regression, SVM, KNN and Random Forest models were used for classification. A total of 457 samples were used for training/testing the models from the dataset Donate-a-cry corpus.

Automated summary

This study focused on classifying child cries based on extracted features through speech and auditory processing. The model was trained using individual features and then retrained with the best selected features. Different classification models were used with a dataset of 457 samples from the Donate-a-cry corpus.