Research and Implementation of Bird Species Identification Method Around Overhead Transmission Lines Based on Voiceprint Recognition

This study explores the method of using voiceprint recognition technology to identify bird species. The ECAPA-TDNN algorithm is adopted to analyze and recognize bird song audio, and a bird song collection device is independently developed. In terms of data support, a dedicated bird song audio database is constructed by combining on-site field collection with the integration of public datasets. For the original audio data, preprocessing operations such as frame division, windowing and denoising, and dynamic Mel filtering are carried out in sequence to improve data quality. In the research, the black-necked crane and magpie are taken as typical cases to deeply analyze their audio features, and model training is conducted based on the ECAPA-TDNN algorithm. Eventually, the weight file for bird song audio analysis is obtained. The recognition results show that in a quiet environment, the ECAPA-TDNN algorithm performs excellently in bird song recognition, but when there is noise interference, its recognition effect needs to be further optimized.