摘要
针对多音轨MIDI文件,提出一种多音轨MIDI音乐主旋律识别方法,通过对表征音乐旋律特征的音高、音长、音色、速度和力度5个特征向量的提取,构建基于BP神经网络的情感模型,并且用200首不同情感特征的歌曲对其进行训练和验证。实验结果显示取得了较好的效果。
The audio track of music melody includes a lot of useful information of music melody, which is the basic of music character recognition and also the premise work in the design of the performance plan of music foundation .Five eigenvectors：pitch, length, tone tempo and strength are extracted for the expression of music melody, by which, the basic music character recognition system can be set up. A emotion model is formed by using BP neural network.200 songs with different emotional characteristic songs will be used as the sample data for the training and validation of the neural network. The results of validation shows the effectiveness of the emotion model.
出处
《电子设计工程》
2015年第8期71-74,共4页
Electronic Design Engineering
