基于卷积神经网络的电子音乐辨识模型  

Electronic Music Identification Model Based on Convolutional Neural Network

作　　者：胡淑娟[1] HU Shujuan(Education Institute, Xiantao Vocational College, Hubei Xiantao 433000, China)

机构地区：[1]仙桃职业学院教育学院,湖北仙桃433000

出　　处：《微型电脑应用》2021年第11期150-153,共4页Microcomputer Applications

摘　　要：现有音频辨识变模型无法分辨电子音乐类型,辨识精度较低,为此,设计基于卷积神经网络的电子音乐辨识模型。通过重建电子音乐信号频谱内谐波信息,对电子音乐信号进行预处理,去除电子音乐信号频谱内的噪声,并将去除噪声后的电子音乐文件制作成波形图。将电子音乐频谱波形图作为输入,利用多层特征融合的混合和采样方式提取输入图像内的特性,利用反向传播算法训练卷积神经网络,通过Soft Max分类器试点电子音乐分类辨识。实验结果显示,所设计模型能够有效去除电子音乐内的噪声含量,在迭代次数达到100次时即可将模型拟合误差降至下限值,且辨识率均值达到98.5%左右。The existing audio identification variable model cannot distinguish the types of electronic music,and the identification accuracy is low.Therefore,an electronic music identification model based on convolutional neural network is designed.By reconstructing the harmonic information in the electronic music signal spectrum,the electronic music signal is preprocessed,the noise in the electronic music signal spectrum is removed,and the electronic music file after noise removal is made into a waveform diagram.Taking the electronic music spectrum waveform as the input,the features in the input image are extracted by the mixing and sampling method of multi-layer feature fusion,the convolutional neural network is trained by the back propagation algorithm,and the electronic music classification recognition is piloted by the Soft Max classifier.Experimental results show that the designed model can effectively remove the noise content in electronic music,and the fitting error of the model can be reduced to the lower limit when the number of iterations reaches 100,and the average recognition rate reaches about 98.5%.

关 键 词：卷积神经网络 电子音乐 辨识模型 谐波信息 混合采样 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]