基于基音频率的数字化音乐情感分类方法

Classification Method of Digital Music Emotion Based on Fundamental Sound Frequency

为提升音乐分类准确率,提出基于基音频率的数字化音乐情感分类方法。其中通过添加滤波器对音乐信号进行预加重,引入帧移实施分帧处理,使用截断函数实施截断处理。基于多项式拟合设计AMDF基音频率检测算法,对预处理后的音乐信号进行数字化分析。提取Mel频率倒谱系数、线性预测倒谱系数、共振峰参数等特征参数。基于CNN-LSTM构建音乐情感分类模型,完成数字化音乐情感分类。实验结果表明,该方法在不同情感类别与不同帧移大小下,分类准确率均较高,误报率均较低,性能较好。

In order to improve the accuracy of music classification,a digital music emotion classification method based on basic sound frequency is proposed.It adds a filter to pre-increase the music signal,introduces frameshift to implement frame processing,and implements truncation function.AMDF base frequency detection algorithm is designed based on polynomial fitting to analyze the preprocessed music signals.Characteristic parameters such as Mel frequency inversion coefficient,linear prediction and resont parameters are extracted.Based on CNN-LSTM,the music emotion classification model is constructed to complete the digital music emotion classification.Experimental results show that the proposed method in different emotion categories vs High classification accuracy,and low false alarm rate showed good performance.