Vocalization Patterns in Laying Hens - An Analysis of Stress-Induced Audio Responses

Abstract

AbstractThis study presents a novel approach to understanding chicken vocalizations using advanced machine learning techniques. Employing a Convolutional Neural Network (CNN) model, the research aimed to classify and analyze the vocalization patterns of laying hens under various conditions. The dataset comprised vocal recordings from chickens exposed to two distinct stressors – umbrella and dog barking – as well as control groups, across different age stages. The primary objective was to assess how external stressors, age, and the timing of stressor application influence the vocal behavior of chickens. The classification results from the CNN model revealed distinct vocal patterns between control and treated groups, indicating that different types of stressors elicit unique vocal responses in chickens. Notably, the model was successful in distinguishing between pre-stress and post-stress vocalizations, suggesting a significant impact of stressor application on chicken vocal behavior. Additionally, the study found that the age of the chickens played a crucial role in their vocal response to stressors. Younger chickens exhibited different vocalization patterns compared to older ones, with these variations becoming more pronounced over time. The analysis utilized Mel Frequency Cepstral Coefficients (MFCC) as the feature set, which effectively captured the spectral characteristics of the bird songs, providing a robust basis for classification. The findings of this research contribute to a deeper understanding of chicken behavior, particularly in relation to environmental stressors and developmental changes. This knowledge holds significant potential for enhancing welfare monitoring practices in poultry farming, offering a non-invasive and technologically advanced method to assess the well-being of chickens in various rearing conditions.