Birdsong Classification Based on Hybrid CNN-LSTM Neural Network

Xin Wang,Chengcheng Ren,Tao Yu,Shuping He
DOI: https://doi.org/10.1109/yac59482.2023.10401432
2023-01-01
Abstract:In this paper, wavelet packet denoising and data enhancement are used to deal with the problems of noisy background noise and small data volume in Birdsong Audio. The Mel frequency cepstral coefficients and Gammatone frequency cepstral coefficients extracted from the audio are fused to form a rich feature vector. In this paper, CNN and LSTM are combined to form a CNN-LSTM combined neural network model, which makes the network classification of birdsong more accurate.
What problem does this paper attempt to address?