5G AI-IoT system for bird species monitoring and bird song classification

Abstract

Identification of animal species is a crucial aspect of biology and ecology, particularly in ornithology, where collaboration with other disciplines aims to develop effective methods for bird protection and environmental quality assessment. Leveraging artificial intelligence (AI) and Internet of Things (IoT) technologies, advancements in birdsong identification have been achieved. Machine learning and deep learning techniques, including Imagenet-based Convolutional Neural Networks (CNNs) like EfficientNet and MobileNet, have been employed for image feature comparison. Spectrograms of birdsongs have been analyzed, with Deep CNNs (DCNNs) proving effective in reducing model size for birdsong classification. A 5G IoT-based system for raw audio collection has been implemented, testing various CNNs for bird identification from audio recordings. While Imagenet-based CNNs exhibit high accuracy (up to 75%), their training requires a substantial number of parameters, hindering efficiency during inference. In response, two Deep CNNs were designed to reduce parameters while maintaining accuracy, enabling integration into Small Board Computers (SBCs) or Microcontroller Units (MCUs). These DCNNs demonstrated a 6.7% improvement in overall accuracy compared to other networks, offering a lighter solution for deployment in SBCs and MCUs.