Multi-View Classification Model for Birdsong Recognition

Birds play a crucial role in nature as sensitive environmental indicators. The variety and activity of bird species can reveal the ecological status of specific areas. this study proposes an innovative multi-view feature fusion model. Experimental results demonstrate that this model achieves a classification accuracy of 90.54%, which represents a significant improvement over single-view models, such as the ResNet model (90.14%) and the Shuffle Transformer model (89.28%). Notably, this model has only 389,905 parameters, far fewer than ResNet's 23,535,080 and Shuffle Transformer's 27,774,452. This not only highlights the model's superior efficiency but also reduces computational and storage burdens, making it an ideal choice for real-time inference and classification tasks on embedded devices.