<![CDATA[Automatic bird species classification based on chirping sounds has become an important solution to support conservation efforts for Indonesia's biodiversity, which comprises 1.835 bird species. This study proposes a classification system that combines Mel-Frequency Cepstral Coefficients (MFCC) feature extraction with Bidirectional Long Short-Term Memory (BiLSTM) architecture to identify 10 commonly found Indonesian bird species. The research dataset utilized 750 bird sound recordings from the xeno-canto.org platform, segmented into 4-second duration clips and augmented to 3,750 samples through pitch shift and time stretch techniques. MFCC feature extraction with 40 coefficients was employed to represent the spectral characteristics of bird sounds, while the BiLSTM model was selected to capture complex bi-directional temporal dependencies in bird vocal signals. In the testing process, an 80:20 data split was performed for training and testing. Confusion matrix analysis confirms the model's capability to distinguish unique characteristics of each species with minimal error rates. Research results demonstrate that the system achieved a classification accuracy of 98%. The combination of MFCC and BiLSTM proves effective for automated and sustainable biodiversity monitoring and bird conservation applications in Indonesia.]]>
Copyrights © 2025
