摘要
利用相空间重构方法提取音符音频中非线性特征参量,将部分参量作为训练集来构造支持向量机(SVM)分类器,另一部分作为测试集进行识别效果的检验,由于固定相空间重构参数后,将会导致部分音符信号的非线性信息丢失,从而降低识别准确率,因此将自适应信号分解和PCA的方法引入到信号预处理环节中,建立了相应的识别流程。
This paper presents a method for notes recognition by utilizing nonlinear parameter based on Reconstructed Phase Space technique. After extracting parameters from note signals, SVM classifier are developed using part of the nonlinear parameters and the rest are used as test data to check out the recognition efficiency. The result shows that the way of using fixed parameters will cause loss of nonlinear imformation, which lower the recognition accuracy. As the way of improvement, adaptive signal decomposition and PCA method are intro duced in the recognition process.
出处
《计算机与数字工程》
2013年第8期1246-1248,1387,共4页
Computer & Digital Engineering
