Libro

This Book examines how background noise affects bird call recognition by employing a Convolutional Neural Network (CNN) combined with a Bi-directional Long Short-Term Memory (BiLSTM) layer. Ambient noise complicates the extraction of bird calls, resulting in misclassifications, inter-species confusion, and longer training durations. Although Noise-Aware CNN–BiLSTM hybrid models are proficient in capturing both spatial and temporal features from spectrograms, noise distorts these representations, forcing the model to focus on irrelevant signals. To address this issue, the proposed architecture leverages localized features for call pattern recognition and temporal modeling for improved noise robustness.