Machine Learning based Bird’s Emotions Recognition using Joint Dataset of Audio and Visual Images

Accurate recognition of bird emotions is essential for advancing bird behavior studies and conservation efforts. This paper introduces a learning approach for bird emotions recognition, combining Convolution Neural Networks (CNN) with Bidirectional Long Short-Term Memory (BiLSTM) networks. This work provides a robust framework for bird emotions classification, offering implications for behavioral research, conservation strategies, and understanding bird communication patterns. Additionally, the model's framework can be adapted for other fields of research, creating opportunities for cross-domain applications. As the model evolves, integrating additional data types, like environmental factors, could further refine predictions and enhance research capabilities in bird behavior. Furthermore, the scalability of the approach ensures its applicability to diverse datasets, expanding its potential across various domains. Four different experiments have been carried out for this investigation for four different bird emotions classes like happy, sad, fear and matting. The best performance of 97.50 % accuracy is achieved with clean data, while noisy data reduced the accuracy to 78.75 %. Mixed data offered balanced outcome with 96.88 % accuracy and attained faster training time. This proposed hybrid CNN-BiLSTM multimodal model for bird emotions recognition holds promise for tackling complex problems that require a holistic understanding in case of multiple data sources.