EEI 2022
All proceedings of this conference

COVID-19 Sound Recognition Algorithm Based on VGGmobNet

Conference: EEI 2022 - 4th International Conference on Electronic Engineering and Informatics
06/24/2022 - 06/26/2022 at Guiyang, China

Proceedings: EEI 2022

Pages: 5Language: englishTyp: PDF

Authors:
Wu, Chenwen; Cao, Xuetong; Peng, Wei (School of Electronics and Information Engineering, Lanzhou Jiaotong University, Lanzhou Gansu, China)

Abstract:
In order to address the problem of the excessive number of parameters and low detection accuracy of the current network model that facilitates audio diagnosis of COVID-19, we proposed a training method based on a three-channel Mel-Spectrogram, which is obtained by multi-scale window size and hop length. Besides, we put forward an improved VGGish network based on inverted residual block, while reducing the number of parameters, speeding up the model training process by using transfer learning. Compared with AlexNet, VGG16 and other neural network, the VGGmobNet network model proposed in this paper has fewer parameters, with an accuracy of 85.2%.