Acoustic classification of bird species plays an important role in accurately monitoring bird diversity. Traditional methods of bird species identification, such as visual observation and manual recording, can be time-consuming and require specialized expertise. Acoustic classification, which leverages the unique vocalizations of birds, offers a promising alternative for efficient and automated species identification. Previous studies have explored various pre-trained ImageNet models for acoustic classification of bird species. However, applying pre-trained ImageNet models directly to audio classification tasks presents a challenge due to the domain mismatch between visual and auditory data. In this study, we present a novel classification framework based on a pre-trained AudioNet model. In order to improve the performance of bird sound classification, the attention mechanism is used to generate important features and gated recurrent unit is used to capture temporal information. The experimental outcomes verify the effectiveness of our proposed system in the classification of bird audio segments, achieving impressive metrics: an accuracy of 97.88%, a precision of 97.95%, a recall of 97.29%, and an F1-score of 97.58%. These figures reflect the system’s remarkable performance in distinguishing among 20 distinct bird species.