The identification of bird species can yield important information on changes in their populations and habitats, making birds an essential species for tracking environmental changes. Traditional methods for identifying birds mainly rely on visual observations, which can be limited in effectiveness, especially in dense forests. In the past, many ornithologists and researchers encountered difficulties identifying different bird species and understanding their patterns. Automating the identification of bird species has become possible because of the advancement of deep learning techniques, particularly in the area of audio processing. Utilizing a bird song dataset (xeno-canto) that includes audio recordings of five distinct bird species, the suggested system will be able to capture a wide variety of vocalizations. The audio inputs are converted into useable representations appropriate for the CNN model using feature extraction techniques as Mel Spectrogram and Constant Q Transform. The estimation of bird populations contributes to larger efforts in biodiversity conservation and ecological research by educating researchers, ornithologists, and wildlife enthusiasts about the presence of different bird species in a given geographic area and the management of bird populations. The proposed CNN model achieves an accuracy of 99.70% with Mel Spectrogram and 99.42% with constant Q Transform.