Real-Time Sound Detection of Rose-Ringed Parakeet Using LSTM Network with MFCC and Mel Spectrogram

Nowadays, machine learning, deep learning, and artificial intelligence are essential in identifying bird presence in rice fields and other agricultural applications. Automated identification techniques make it easier for farmers to monitor bird populations and identify potential threats to rice crops without human intervention. Using sound-based analysis, artificial intelligence (AI)--driven systems can distinguish between bird species that benefit the environment and those that might cause harm. This research presents a unique approach to realtime vocalization identification of Rose-Ringed Parakeets. A Long Short-Term Memory (LSTM) neural network is used in the proposed approach, and it is improved by Mel Frequency Cepstral Coefficients (MFCC) and Mel Spectrogram features. The audio data was processed to extract Mel Spectrogram and MFCC features, which were then utilized to capture the unique spectrum attributes shown by parakeet sounds. The LSTM framework for identifying Rose-ringed Parakeets from their auditory signals was developed and implemented using a Raspberry Pi 4 B single-board computer. Because of its adaptability, our method is ideal for real-world use in situations like the prompt detection and control of parakeet populations. Automating AI procedures reduces the possibility of human error and ensures a continuous and accurate monitoring system. Farmers and researchers can make data-driven decisions that enhance crop yields, preserve biodiversity, and promote sustainable agricultural practices with these technologies.