基于低秩表示和稀疏自动编码器的情绪识别研究

Research on Emotion Recognition Based on Low Rank Representation and Sparse Automatic Encoder

针对当今在多模态数据处理过程中存在的两个问题:1)多模态数据中存在冗余信息和噪声;2)如何平衡多模态数据间的关系,我们提出了一种低秩表示和稀疏自动编码器相结合的情绪识别方法(Low Rank Representation and Sparse Automatic Encoder Classification,LSC)。采用低秩表示从数据的高维空间中提取潜在特征,并且可以去除数据噪声和冗余信息;采用自动编码器对提取的特征进行融合,解决了多模态数据信息间关系无法平衡的问题,使特征更具鲁棒性,最后进行分类。进行了实验,证明此方法的有效性。

In view of the two problems in multi-modal data processing today;1)redundancy and noise exist in multi-modal data;2)how to balance the relationship between multi-modal data,a Low Rank Representation and Sparse Automatic Encoder Classification method(LSC)is proposed.The low-rank representation is used to extract potential features from the high-dimensional space of the data,and the noise and redundant information in the data can be removed.The extracted features are fused with an automatic encoder,which solves the problem that the relationship between multi-modal data and information cannot be balanced,and makes the features more robust,and finally classifies them.Experiments are carried out to prove the effectiveness of this method.