Poultry Vocalization Signal Dataset for Early Disease Detection
Description
The aim of generating this dataset is to deliver open, accessible and quality machine learning dataset for Poultry farm management. The dataset contains a total of 346 audio signal files which are divided into three folders: healthy, noise and unhealthy. The healthy folder contains 139 audio files, noise folder has 86, while unhealthy folder has 121 files. The audio files are saved in .wav file format.
Files
Steps to reproduce
Bowen University poultry research farm was used as the experimental environment where total number of day-old chicks were acquired and divided into two groups. The first group was treated for respiratory diseases and the other group was not treated for respiratory diseases. The chicks were then placed in a controlled environment, separately from each other, and microphones were positioned at a suitable distance from the birds to avoid background noise and other sounds that may interfere with the analysis. Data was collected at 96 kHz with 24-bit samples. The audio data collection process occurred continuously, three times daily (morning, afternoon and night) for a period of 65days. During this period, food and water are readily available to the birds at all times. After 30 days, the untreated group developed respiratory disease which was evident in their sound. These data were also captured as unhealthy. Audio signals of poultry chickens were collected and stored in MA4 and later converted to WAV format.