基于深度学习的机器人舞蹈自动生成研究

Dance generation model based on deep learning

在舞蹈机器人构建过程中,针对传统舞蹈生成算法无法精确匹配音乐的问题,结合深度学习的特征提取特点,构建一个基于深度学习的舞蹈生成模型。首先对模型进行整体设计,分别对舞蹈和音频进行特征提取;然后对生成模型的各个功能模块进行设计和实现,利用机器人对舞蹈动作进行展现。实验结果表明,设计的模型生成舞蹈与音频较为匹配,与真实舞蹈动作十分接近,本模型的质量评估值为51.19,该模型对舞蹈姿态序列的分类识别效果准确,音乐匹配率更精准,舞蹈生成效果最佳。

In the process of building a dance robot,a dance generation model based on deep learning is constructed for the problem that the traditional dance generation algorithm cannot accurately match the music and combined with the feature extraction characteristics of deep learning.First,the model is designed to extract the dance and audio;then the functional modules are designed and implemented to show the dance movements.The experimental results show that the designed model generation dance matches the audio,and is very close to the real dance action.The quality evaluation value of this model is 51.19.The model has an accurate classification and recognition effect of the dance posture sequence,the music matching rate is more accurate,and the dance generation effect is the best.