| 報告時間：2025-4-25 |
| 報告地點：Room 407 |
| 指導老師：Dr. Hsin-I Chiang |
| 學生：Hui Tsz Yan |
| 摘要 |
|
With the advancement of technology, the health and productivity of dairy cows have become increasingly influential on farm economics, driving the transition toward intelligent livestock management. Cow vocalizations serve as important indicators of emotional states, physiological needs, and health conditions. This study aims to develop a machine learning-based system capable of recognizing cow calls and analyzing their emotional states. A vocalization database was constructed by recording cow calls under various scenarios, and the data were categorized based on vocalization type (cow sounds vs. non-cow sounds), call frequency (high vs. low), and emotional valence (positive vs. negative). Prior to classification, the audio data underwent preprocessing, including noise reduction and signal segmentation. Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) algorithms were employed for model training and classification. The system successfully identified cow vocalizations, call types, and emotional states with high accuracy. SVM achieved classification accuracies of 91.30%, 97.06%, and 90.60% for the three respective tasks, while KNN achieved 93.48%, 91.18%, and 90.60%. Although KNN showed slightly better performance in call type classification, SVM demonstrated more consistent overall performance and was therefore deemed the more suitable model. To validate the system’s accuracy and practical application, cow vocalization samples collected from external sources were analyzed and compared with human expert assessments. The results confirmed the system’s reliability and effectiveness. These findings highlight the potential of cow vocalization recognition technology in health monitoring and emotional assessment, contributing to improved animal welfare, enhanced management efficiency, and the advancement of smart livestock farming. |
| 參考文獻 |
|