Study on a Pig Vocalization Classification Method Based on Multi-Feature Fusion

• DOI: 10.3390/s24020313
• Publication Date: 2024-01-05T08:43:00Z
• Authors: Yuting Hou; Qifeng Li; Zuchao Wang; Tonghai Liu; Yuxiang He; Haiyan Li; Zhiyu Ren; Xiaoli Guo; Gan Yang; Yu Liu; Ligen Yu
• Keywords: Mel-frequency cepstrum; Centroid; Feature (linguistics)

To improve the classification of pig vocalization using vocal signals and recognition accuracy, a method based on multi-feature fusion is proposed in this study. With typical pigs large-scale breeding houses as research object, short-time energy, frequency centroid, formant first-order difference, Mel cepstral coefficient difference were extracted features. These features improved principal component analysis. A model with BP neural network optimized genetic algorithm was constructed. The results showed that to recognize grunting, squealing, coughing, average accuracy 93.2%; precisions 87.9%, 98.1%, 92.7%, respectively, an 92.9%; recalls 92.0%, 99.1%, 87.4%, 92.8%, which indicated had good precision recall, could provide reference for information feedback automatic recognition.