Utilizing sound analysis to automatically identify patterns of vocalization in healthy broiler chickens.

Analyzing broiler chicken vocalizations offers insights into their behavior and affective state. The objective of this trial was to identify the normal vocalization patterns of broiler chickens throughout the on-farm stage, considering different times of the day (morning, midday, afternoon, and night). Initially, a systematic review was conducted to identify broiler chicken vocalizations detected by sound analysis described in the literature. For data collection, ten one-day-old male Ross 308 slow-feathering chicks were housed in a 2m² pen, isolated from other broilers and external influences, and monitored continuously. Four well-described vocalizations in broilers—distress calls (DC), short peeps (SP), warble notes (WN), and pleasure notes (PN)—were manually labeled to develop and validate an automated vocalization detector based on a fully convolutional neural network (FCNN). The model achieved a balanced accuracy of 87.9% with minimal confusion. Any vocalization outside the four selected categories was classified as "other" (O). A subsequent trial included three rounds, each containing four pens with 540 male Ross 308 chickens per pen. The FCNN algorithm was used to monitor vocalizations 24/7 from 1 to 42 days of age, analyzing their temporal evolution. Time of day significantly affected SP, DC, W, and PN. Vocalization patterns remained consistent across treatments but varied by age and daily periods. SP peaked in the evening and was lowest at night. DC was highest at midday, W at night, and PN at night but lower at midday and evening. SP and DC were the most common vocalizations. PN was most frequent in 1-week-old broilers. W remained consistently low. DC, SP, and OV increased with age, whereas PN declined, indicating evolving vocal communication. These findings highlight significant age- and time-related vocalization shifts during the on-farm cycle.