Baby Cry Decoder: A boon for the Hearing Impaired Caregiver

The cry of a baby, the main means of communication between the baby and the caregiver carries wealth of information about the status and the need of the baby. A caregiver who is hearing impaired is placed at an extreme disadvantage because of his inability to perceive this vital channel of communication with the baby. The main aim of this work is to bridge this gap by building a decoder which captures the cry of the baby and classifies it into various categories which will help the caregiver in taking the appropriate action thus empowering the hearing challenged caregiver. This study introduces several pre-processing techniques as well as a variety of feature extraction schemes such as MFCC, spectrogram, and fundamental frequency, among others to characterize the cries of babies. Machine learning classifiers may be trained using these characteristics. In this research activity, recently proposed neural network architectures like EfficientNet and ResNet are used along with established Machine Learning classifiers such as KNN and SVM for building the classifier. This study thoroughly examines prior research in all relevant areas of decoding cries of babies in order to gain insights into the most recent, cutting-edge works in infant cry signal processing and categorization. The main objective of this research project is to compare various models for the purpose of determining the optimal model for classifying cries of infants.