Source code for torcheeg.datasets.module.emotion_recognition.seed
import os
from typing import Any, Callable, Dict, Tuple, Union
import scipy.io as scio
from ...constants.emotion_recognition.seed import (SEED_ADJACENCY_MATRIX,
SEED_CHANNEL_LOCATION_DICT)
from ..base_dataset import BaseDataset
[docs]class SEEDDataset(BaseDataset):
r'''
The SJTU Emotion EEG Dataset (SEED), is a collection of EEG datasets provided by the BCMI laboratory, which is led by Prof. Bao-Liang Lu. This class generates training samples and test samples according to the given parameters, and caches the generated results in a unified input and output format (IO). The relevant information of the dataset is as follows:
- Author: Zheng et al.
- Year: 2015
- Download URL: https://bcmi.sjtu.edu.cn/home/seed/index.html
- Reference: Zheng W L, Lu B L. Investigating critical frequency bands and channels for EEG-based emotion recognition with deep neural networks[J]. IEEE Transactions on Autonomous Mental Development, 2015, 7(3): 162-175.
- Stimulus: 15 four-minute long film clips from six Chinese movies.
- Signals: Electroencephalogram (62 channels at 200Hz) of 15 subjects, and eye movement data of 12 subjects. Each subject conducts the experiment three times, with an interval of about one week, totally 15 people x 3 times = 45
- Rating: positive (1), negative (-1), and neutral (0).
In order to use this dataset, the download folder :obj:`Preprocessed_EEG` is required, containing the following files:
- label.mat
- readme.txt
- 10_20131130.mat
- ...
- 9_20140704.mat
An example dataset for CNN-based methods:
.. code-block:: python
dataset = SEEDDataset(io_path=f'./seed',
root_path='./Preprocessed_EEG',
offline_transform=transforms.Compose([
transforms.BandDifferentialEntropy(),
transforms.ToGrid(SEED_CHANNEL_LOCATION_DICT)
]),
online_transform=transforms.ToTensor(),
label_transform=transforms.Compose([
transforms.Select('emotion'),
transforms.Lambda(lambda x: x + 1)
]))
print(dataset[0])
# EEG signal (torch.Tensor[4, 9, 9]),
# coresponding baseline signal (torch.Tensor[4, 9, 9]),
# label (int)
Another example dataset for CNN-based methods:
.. code-block:: python
dataset = SEEDDataset(io_path=f'./seed',
root_path='./Preprocessed_EEG',
online_transform=transforms.Compose([
transforms.ToTensor(),
transforms.To2d()
]),
label_transform=transforms.Compose([
transforms.Select('emotion'),
transforms.Lambda(x: x + 1)
]))
print(dataset[0])
# EEG signal (torch.Tensor[62, 200]),
# coresponding baseline signal (torch.Tensor[62, 200]),
# label (int)
An example dataset for GNN-based methods:
.. code-block:: python
dataset = SEEDDataset(io_path=f'./seed',
root_path='./Preprocessed_EEG',
online_transform=transforms.Compose([
ToG(SEED_ADJACENCY_MATRIX)
]),
label_transform=transforms.Compose([
transforms.Select('emotion'),
transforms.Lambda(x: x + 1)
]))
print(dataset[0])
# EEG signal (torch_geometric.data.Data),
# coresponding baseline signal (torch_geometric.data.Data),
# label (int)
In particular, TorchEEG utilizes the producer-consumer model to allow multi-process data preprocessing. If your data preprocessing is time consuming, consider increasing :obj:`num_worker` for higher speedup. If running under Windows, please use the proper idiom in the main module:
.. code-block:: python
if __name__ == '__main__':
dataset = SEEDDataset(io_path=f'./seed',
root_path='./Preprocessed_EEG',
online_transform=transforms.Compose([
ToG(SEED_ADJACENCY_MATRIX)
]),
label_transform=transforms.Compose([
transforms.Select('emotion'),
transforms.Lambda(x: x + 1)
]),
num_worker=4)
print(dataset[0])
# EEG signal (torch_geometric.data.Data),
# coresponding baseline signal (torch_geometric.data.Data),
# label (int)
Args:
root_path (str): Downloaded data files in matlab (unzipped Preprocessed_EEG.zip) formats (default: :obj:`'./Preprocessed_EEG'`)
chunk_size (int): Number of data points included in each EEG chunk as training or test samples. If set to -1, the EEG signal of a trial is used as a sample of a chunk. (default: :obj:`200`)
overlap (int): The number of overlapping data points between different chunks when dividing EEG chunks. (default: :obj:`0`)
num_channel (int): Number of channels used, of which the first 62 channels are EEG signals. (default: :obj:`62`)
online_transform (Callable, optional): The transformation of the EEG signals and baseline EEG signals. The input is a :obj:`np.ndarray`, and the ouput is used as the first and second value of each element in the dataset. (default: :obj:`None`)
offline_transform (Callable, optional): The usage is the same as :obj:`online_transform`, but executed before generating IO intermediate results. (default: :obj:`None`)
label_transform (Callable, optional): The transformation of the label. The input is an information dictionary, and the ouput is used as the third value of each element in the dataset. (default: :obj:`None`)
before_trial (Callable, optional): The hook performed on the trial to which the sample belongs. It is performed before the offline transformation and thus typically used to implement context-dependent sample transformations, such as moving averages, etc. The input of this hook function is a 2D EEG signal with shape (number of electrodes, number of data points), whose ideal output shape is also (number of electrodes, number of data points).
after_trial (Callable, optional): The hook performed on the trial to which the sample belongs. It is performed after the offline transformation and thus typically used to implement context-dependent sample transformations, such as moving averages, etc. The input and output of this hook function should be a sequence of dictionaries representing a sequence of EEG samples. Each dictionary contains two key-value pairs, indexed by :obj:`eeg` (the EEG signal matrix) and :obj:`key` (the index in the database) respectively.
io_path (str): The path to generated unified data IO, cached as an intermediate result. (default: :obj:`./io/seed`)
io_size (int): Maximum size database may grow to; used to size the memory mapping. If database grows larger than ``map_size``, an exception will be raised and the user must close and reopen. (default: :obj:`10485760`)
io_mode (str): Storage mode of EEG signal. When io_mode is set to :obj:`lmdb`, TorchEEG provides an efficient database (LMDB) for storing EEG signals. LMDB may not perform well on limited operating systems, where a file system based EEG signal storage is also provided. When io_mode is set to :obj:`pickle`, pickle-based persistence files are used. (default: :obj:`lmdb`)
num_worker (int): Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process. (default: :obj:`0`)
verbose (bool): Whether to display logs during processing, such as progress bars, etc. (default: :obj:`True`)
in_memory (bool): Whether to load the entire dataset into memory. If :obj:`in_memory` is set to True, then the first time an EEG sample is read, the entire dataset is loaded into memory for subsequent retrieval. Otherwise, the dataset is stored on disk to avoid the out-of-memory problem. (default: :obj:`False`)
'''
channel_location_dict = SEED_CHANNEL_LOCATION_DICT
adjacency_matrix = SEED_ADJACENCY_MATRIX
def __init__(self,
root_path: str = './Preprocessed_EEG',
chunk_size: int = 200,
overlap: int = 0,
num_channel: int = 62,
online_transform: Union[None, Callable] = None,
offline_transform: Union[None, Callable] = None,
label_transform: Union[None, Callable] = None,
before_trial: Union[None, Callable] = None,
after_trial: Union[Callable, None] = None,
after_session: Union[Callable, None] = None,
after_subject: Union[Callable, None] = None,
io_path: str = './io/seed',
io_size: int = 10485760,
io_mode: str = 'lmdb',
num_worker: int = 0,
verbose: bool = True,
in_memory: bool = False):
# pass all arguments to super class
params = {
'root_path': root_path,
'chunk_size': chunk_size,
'overlap': overlap,
'num_channel': num_channel,
'online_transform': online_transform,
'offline_transform': offline_transform,
'label_transform': label_transform,
'before_trial': before_trial,
'after_trial': after_trial,
'after_session': after_session,
'after_subject': after_subject,
'io_path': io_path,
'io_size': io_size,
'io_mode': io_mode,
'num_worker': num_worker,
'verbose': verbose,
'in_memory': in_memory
}
super().__init__(**params)
# save all arguments to __dict__
self.__dict__.update(params)
@staticmethod
def process_record(file: Any = None,
root_path: str = './Preprocessed_EEG',
chunk_size: int = 200,
overlap: int = 0,
num_channel: int = 62,
before_trial: Union[None, Callable] = None,
offline_transform: Union[None, Callable] = None,
after_trial: Union[None, Callable] = None,
**kwargs):
file_name = file
subject = int(os.path.basename(file_name).split('.')[0].split('_')
[0]) # subject (15)
date = int(os.path.basename(file_name).split('.')[0].split('_')
[1]) # period (3)
samples = scio.loadmat(os.path.join(root_path, file_name),
verify_compressed_data_integrity=False
) # trial (15), channel(62), timestep(n*200)
# label file
labels = scio.loadmat(
os.path.join(root_path, 'label.mat'),
verify_compressed_data_integrity=False)['label'][0]
trial_ids = [key for key in samples.keys() if 'eeg' in key]
write_pointer = 0
# loop for each trial
for trial_id in trial_ids:
# extract baseline signals
trial_samples = samples[trial_id] # channel(62), timestep(n*200)
if before_trial:
trial_samples = before_trial(trial_samples)
# record the common meta info
trial_meta_info = {
'subject_id': subject,
'trial_id': trial_id,
'emotion': int(labels[int(trial_id.split('_')[-1][3:]) - 1]),
'date': date
}
# extract experimental signals
start_at = 0
if chunk_size <= 0:
dynamic_chunk_size = trial_samples.shape[1] - start_at
else:
dynamic_chunk_size = chunk_size
# chunk with chunk size
end_at = dynamic_chunk_size
# calculate moving step
step = dynamic_chunk_size - overlap
while end_at <= trial_samples.shape[1]:
clip_sample = trial_samples[:num_channel, start_at:end_at]
t_eeg = clip_sample
if not offline_transform is None:
t_eeg = offline_transform(eeg=clip_sample)['eeg']
clip_id = f'{file_name}_{write_pointer}'
write_pointer += 1
# record meta info for each signal
record_info = {
'start_at': start_at,
'end_at': end_at,
'clip_id': clip_id
}
record_info.update(trial_meta_info)
yield {'eeg': t_eeg, 'key': clip_id, 'info': record_info}
start_at = start_at + step
end_at = start_at + dynamic_chunk_size
def set_records(self, root_path: str = './Preprocessed_EEG', **kwargs):
file_list = os.listdir(root_path)
skip_set = ['label.mat', 'readme.txt']
file_list = [f for f in file_list if f not in skip_set]
return file_list
def __getitem__(self, index: int) -> Tuple[any, any, int, int, int]:
info = self.read_info(index)
eeg_index = str(info['clip_id'])
eeg_record = str(info['_record_id'])
eeg = self.read_eeg(eeg_record, eeg_index)
signal = eeg
label = info
if self.online_transform:
signal = self.online_transform(eeg=eeg)['eeg']
if self.label_transform:
label = self.label_transform(y=info)['y']
return signal, label
@property
def repr_body(self) -> Dict:
return dict(
super().repr_body, **{
'root_path': self.root_path,
'chunk_size': self.chunk_size,
'overlap': self.overlap,
'num_channel': self.num_channel,
'online_transform': self.online_transform,
'offline_transform': self.offline_transform,
'label_transform': self.label_transform,
'before_trial': self.before_trial,
'after_trial': self.after_trial,
'num_worker': self.num_worker,
'verbose': self.verbose,
'io_size': self.io_size
})
