基于云模型的音乐情感表示与识别研究

项目名称： 基于云模型的音乐情感表示与识别研究

项目编号： No.61303137

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 张克俊

作者单位： 浙江大学

项目金额： 25万元

中文摘要： 近年来，利用人工智能方法来研究音乐情感，得到了学术界和工业界的广泛关注。然而，目前相关的研究还刚刚起步，亟需有效的音乐情感表示方法和识别模型。 针对上述问题，本项目利用云模型在知识表达时具有不确定中带有确定性、稳定之中又有变化的特点，首次将云模型引入音乐情感研究领域。首先，利用云模型表示和解释音乐情感的模糊性和随机性，研究基于深度机器学习选择音乐情感特征参数，构建音乐情感数据库，有效地解决音乐情感表示问题；其次，利用模式云表示基因表达式编程中的精英群体，并基于泛函分析构造适应值函数，提出了云基因表达式编程，并进行收敛性、时间复杂性分析；最后，利用差分进化算法优化云基因表达式编程的集成学习系数，基于音乐情感数据库，构建音乐情感识别模型，有效地解决音乐情感识别问题。 本项目综合利用云模型、机器学习和基因表达式编程等，部分解决了音乐情感表示与情感识别等科学问题，具有重要的理论意义与应用价值。

中文关键词： 云模型；音乐情感识别；基因表达式编程；；

英文摘要： With the development of Music Information Retrieval and Computation Intelligence，Artificial Intelligence has been widely used to solve the problem of music emotion recognition in academia and industry. However, the research is still in its infancy, and there are many research problems, like lack of effective methods to represent music emotion and to detect the emotion in music. To solve those problems, Cloud Model is used to represent the emotion in music, which fully expressed the fuzziness and randomness of music emotion, and Deep Learning is utilized to extract the music emotional features to build a database of music emotion, which effectively solve the problem of music expression; Then, the model for music emotion recognition based on cloud Gene Expression Programming,unique fitness function and ensemble method, was presented to solve the problem of music emotion recognition. It's worth to mentioning that it is the first time that Cloud Model has been used in the field of music emotion research. Comprehensive utilization of the Deep Learning, Ensemble Method and Gene Expression Programming is an integrated innovation, the research results can be applied to music emotion exploring and indexing, etc., will have important theoretical value and application prospect.

英文关键词： Cloud Model；Music Emotion Recognition；Gene Expression Programming；；