About the feature matrix
Opened this issue · 4 comments
i wonder how to get the feature matrix for my own dataset,thansk for you help!
For EEG data, first, you can make your origin signals as shape [s, n, m], where s denotes the number of samples, n denotes the number of channels, m denotes the time point of one sample (For example, a sample with a sampling rate of 256 Hz for 30 seconds, the number of time points is 256*30=7680). This is the format of the X file in our process.
After this, you can extract DE features and pack the data. For more detailed process please refer to here.
hi, thanks for your reply, another question is that i've seen different kinds of adaptive graph convolution ,like
, and why you tried to use the middle frame to get the adaptive adjacent matrix. Thanks a lot!
The reason for using the middle frame to get the matrix is: in order to judge the sleep stage of an epoch, we use its previous 2 epochs and its following 2 epochs as the context of the current time slice, which is consistent with the judgment method of sleep experts. At the same time, to simplify the calculation, we think that the middle frame can best represent the brain connection at the current time.
For EEG data, first, you can make your origin signals as shape [s, n, m], where s denotes the number of samples, n denotes the number of channels, m denotes the time point of one sample (For example, a sample with a sampling rate of 256 Hz for 30 seconds, the number of time points is 256*30=7680). This is the format of the X file in our process. After this, you can extract DE features and pack the data. For more detailed process please refer to here.
前辈您好，您上文提到信号形状为[s,n,m]，其中s表示样本数量，n表示通道数量，m表示一个样本的时间点。我在运行这个项目时有3个疑问，希望您能帮我解答。1.n是否都为26，m是否都为7680？2.s的值应该如何计算？3.您能否提供某一个受试者的signals shape 和labels shape的具体数字示例？非常感谢您的帮助！