基于节拍谱的语音音乐分类模型

Speech Music Classification Model based on Beat Spectrum

语音音乐分类是语音信号处理领域的重要研究方向。针对以往方法通过提取短时能量、短时幅度等特征参数来区分语音和音乐而忽视音乐具有节拍的特性,为此提出了基于节拍谱的话音分类模型。此模型在针对语音和音乐两类信号中,先对语音信号进行预处理,并对待分类的信号计算得到梅尔频率倒谱系数,再计算梅尔频率倒谱系数的相似矩阵和相似矩阵的自相关,得到待分类信号的节拍谱,最后通过阈值判断待信号类别。经试验结果证明,此模型对比传统分类模型,分类准确率提高到98%。

Speech and music classification is an important research direction in the field of speech signal processing.Aiming at the previous method to distinguish between speech and music by extracting feature parameters such as short-term energy and short-term amplitude,while ignoring the characteristics of music with beats,a speech classification model based on beat spectrum is proposed.In this model,for speech and music signals,the speech signal is preprocessed first,and the Mel frequency cepstral coefficient is calculated for the signal to be classified,and then the autocorrelation of the similarity matrix and the similarity matrix of the Mel frequency cepstrum coefficients are calculated to obtain the beat spectrum of the signal to be classified.Finally,the threshold value is used to determine the signal category.The experimental results indicate that compared with that of traditional classification models,the classification accuracy of this model is 98%.