基于深度学习模型的鸟类保护研究

Research on Bird Conservation Based on Deep Learning Models

摘要: 为了保护鸟类，维持森林生态系统的稳定，针对声音识别技术在鸟类保护系统中的重要性，本文旨在运用深度学习模型实现鸟类声音，伐木，枪声，电锯声的识别，首先通过对采集到的声音运用麦克风阵列法进行声源定位等预处理；其次对声音进行识别，运用梅尔倒谱系数(MFCC)进行特征提取是识别声音的关键步骤，模型测试结果表明，鸟声的识别率为90.2%，电锯和枪声的识别率为96.6%。

Abstract: To protect birds and maintain the stability of forest ecosystems, and considering the significance of sound recognition technology in bird protection systems, this paper aims to apply a deep learning model to recognize bird sounds, logging sounds, gunshots, and chainsaw sounds. First, the collected sounds are preprocessed for sound source localization using the microphone array method. Second, for sound recognition, extracting features with Mel-Frequency Cepstral Coefficients (MFCC) is a crucial step. The model test results show that the recognition rate of bird sounds is 90.2%, and the recognition rate of chainsaw sounds and gunshots is 96.6%.