ziyujia / hetemotionnet Goto Github PK

I have read your paper “HetEmotionNet: Two-Stream Heterogeneous Graph Recurrent Neural Network for Multi-modal Emotion Recognition”, what an excellent and interesting paper.
According to my understanding in section 5.2 Baseline Models and Settings, a 1s non-overlapping window is employed to divide each trial into several samples, the dimension of input signals should be 32x40x60x32x4 (subjects x trials x segments x channels x bands) for DEAP dataset, so we got 32x40x60=76800 1s-length samples with dimension of 32x4. There seems to be no dimension of time_step, how to extract the temporal feature?

您好 抱歉打扰您了 不知道您能否提供一下从数据集到时空图序列构造的代码的 我将十分感谢您的帮助

您好，很抱歉打扰您，上传的代码中似乎缺少了从数据集构成时空图序列和频空图序列的部分？还是我没有找到呢？