摘要
论文建立了特征参量结合分类器的识别流程。在特征参量方面,论文采用相空间重构的方法获取和弦信号的非线性特征用于分类器的构造和相关的识别实验。由于不同和弦信号对应不同的最优相空间重构参数,而在识别未知和弦以及分类器构造过程中,需要固定延迟时间和嵌入维的数值,因此,会造成特征信息丢失的问题,影响识别准确率。为了弥补上述不足,论文在信号预处理环节中引入了EMD自适应信号分解的方法,相关实验表明,该方法可以较为准确地识别出具体的和弦种类。
The process on the basis of characteristic parameters and classifier is built in this thesis.Nonlinear characteristics extracted from the chord are used in constructing classifier and experiments.Although different chords have different phase space reconstruction parameters,the value of delay-time and embedding dimension has to be fixed.And part of the nonlinear characteristics will be lost which lower the recognition accuracy.For making up the disadvantage mentioned above,EMD adaptive signal decomposition methods are used in preprocessing.The result of experiments show that different type of chords can be recognized by this method.
出处
《计算机与数字工程》
2016年第3期497-500,共4页
Computer & Digital Engineering
关键词
和弦识别
相空间重构
EMD
chord recognition
phase space reconstruction
EMD