摘要
针对复杂环境下的声音识别问题,提出一种基于深度学习的声音识别方法。首先,通过自适应滤波降噪和梅尔频率倒谱系数(Mel-Frequency Cepstral Coefficient,MFCC)提取等方法提取声音特征。其次,采用L2正则化的卷积神经网络(Convolutional Neural Network,CNN)识别声音,以提高模型的泛化能力和准确性。最后,使用ESC-50数据集对所提方法进行验证和测试。实验结果表明,该方法的精确率、准确率及召回率均优于对比方法。
A deep learning based sound recognition method is proposed for the problem of sound recognition in complex environments.Firstly,sound features are extracted through methods such as adaptive filtering noise reduction and Mel-Frequency Cepstral Coefficient(MFCC)extraction.Secondly,L2 regularized Convolutional Neural Network(CNN)are used to recognize sounds,in order to improve the model’s generalization ability and accuracy.Finally,validate and test the proposed method using the ESC-50 dataset.The experimental results show that the accuracy,precision,and recall of this method are superior to the comparison methods.
作者
付兆婷
FU Zhaoting(Baiyin College,Baiyin Open University,Baiyin 730900,China)
出处
《电声技术》
2024年第5期40-42,共3页
Audio Engineering
关键词
复杂环境
卷积神经网络(CNN)
声音识别
complex environment
Convolutional Neural Network(CNN)
voice recognition
