基于特征词位置因素的音乐情感智能分类算法

Design of music emotional intelligence classification algorithm based on feature word location factors

针对歌词文本中特征词位置对音乐情感分类的影响问题,文中使用层次分析法来进行特征词在不同位置的权重分析,并对歌词所提取的特征向量进行修正。同时,与音频信号所提取到的特征向量进行多模态数据融合,使用深度置信网络已有监督训练的方式,分析混合融合后的特征向量与音乐情感之间的联系,构建出基于特征词位置因素的音乐情感智能分类算法。测试与实验结果表明,基于特征词位置因素的音乐情感智能分类算法在5种音乐情感样本的测试下,最低准确率为80.1%,平均准确率为83.5%,明显优于未采用位置因素修正的算法,具有良好的有效性与可行性。

In order to explore the influence of feature word position on music emotion classification in lyrics text,AHP is used to analyze the weight of feature words at different positions and correct the feature vector extracted by lyrics.At the same time,multi-modal data fusion is performed with the feature vector extracted from the audio signal,and the deep confidence network is used to supervise the connection between the feature vector and the musical emotion in the form of supervised training,and the feature word position factor is constructed.Musical emotion intelligent classification algorithm.According to the test results,the music emotion intelligent classification algorithm based on the feature word position factor has the lowest accuracy rate of 80.1%and the average accuracy rate of 83.5%under the test of five kinds of music emotion samples,which is obviously better than the algorithm without position factor correction.,with good effectiveness and feasibility.