Datasets:

MigoXV
/

eeg-bme-emotion-parquet

split string	user_id string	source_file string	source_variable string	cohort string	trial_id int16	emotion_trial_id int16	emotion_label int8	emotion_name string	channels int16	samples int32	sampling_rate_hz int16	eeg_dtype string	eeg_shape list	source_start_sample int32	source_end_sample int32	eeg_sha256 string	eeg list
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_neu	healthy_control	1	1	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	0	12,500	bb7cb1b8167cd6915bbb86e340c4689672b54cb45f49737d432e8f11ed459744	[-4.46328067779541,-4.74399995803833,-2.825688123703003,-2.495619058609009,-4.521764755249023,-5.720(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_neu	healthy_control	2	2	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	12,500	25,000	233212e1f013cbb48121fc6e145558140cc5e46a24afff0681f4cf3b2babe658	[4.178714275360107,2.2890827655792236,2.762737512588501,3.4736602306365967,3.294534206390381,1.96019(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_neu	healthy_control	3	3	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	25,000	37,500	5c11c62a4e2af43edfea5193cf6a05be9ca57c024499d12a212207916d071d94	[0.01885785162448883,-0.4286016523838043,4.61472749710083,6.046268939971924,7.093820571899414,5.7564(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_neu	healthy_control	4	4	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	37,500	50,000	c6d475f30dc5ab0a58999437c2c223a3d38c2f6190bf8448733f868150b067e3	[3.6595442295074463,3.541503667831421,1.8403899669647217,1.8467046022415161,3.2790884971618652,4.103(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_pos	healthy_control	5	1	1	positive	30	12,500	250	float32	[ 30, 12500 ]	0	12,500	80da45fd12a651a91558902ea57b4e1cc3e8b24add156433b26b89a4a5381552	[2.3997116088867188,2.869703531265259,0.8408568501472473,-1.4648009538650513,-1.2115223407745361,0.6(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_pos	healthy_control	6	2	1	positive	30	12,500	250	float32	[ 30, 12500 ]	12,500	25,000	4066b9a26c2edb50376532b113fc8167d98d76b7a00aec8de978037c53123d7c	[-4.634337425231934,1.2513718605041504,1.7959047555923462,1.881667137145996,2.49954891204834,3.55274(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_pos	healthy_control	7	3	1	positive	30	12,500	250	float32	[ 30, 12500 ]	25,000	37,500	19ad38303f10419013dcd82233b952e4950a65adfc0ce452a32bb9dd3873a2d0	[3.1811187267303467,2.1354429721832275,-1.2016334533691406,-2.6607582569122314,-2.1701464653015137,-(...TRUNCATED)
train	HC1003	训练集/正常人/HC1003timedata.mat	EEG_data_pos	healthy_control	8	4	1	positive	30	12,500	250	float32	[ 30, 12500 ]	37,500	50,000	72d9b610017faccb464eea97a828ff4ed894bcc6e3602a9f57ac9c6077a64052	[2.8864598274230957,2.018190860748291,3.1917672157287598,-5.607523441314697,-3.5993828773498535,-3.3(...TRUNCATED)
train	HC1005	训练集/正常人/HC1005timedata.mat	EEG_data_neu	healthy_control	1	1	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	0	12,500	35980afea498c022cede385f158d749f5415c81b0eef07abbd7d5ba67967b07c	[-1.9149547815322876,1.8708583116531372,3.71028470993042,3.3174643516540527,2.3761487007141113,2.254(...TRUNCATED)
train	HC1005	训练集/正常人/HC1005timedata.mat	EEG_data_neu	healthy_control	2	2	0	neutral	30	12,500	250	float32	[ 30, 12500 ]	12,500	25,000	21ec54d0df1988d178f218ac3841b5e5bfa810addc0ba7d94996f0716b2c4833	[-1.9406745433807373,3.5466153621673584,4.362205982208252,3.3784735202789307,1.3078402280807495,1.46(...TRUNCATED)

End of preview. Expand in Data Studio

BME EEG Emotion Parquet

赛题四脑电情绪识别数据的 Parquet 转换版本。数据按 HuggingFace Datasets 原生 split 组织：

train: 384 条样本
validation: 96 条样本
test: 80 条样本

每条样本对应一段视频的 EEG 数据。eeg 字段为扁平数组，可通过 eeg_shape 和 eeg_dtype 还原为二维数组。

from datasets import load_dataset
import numpy as np

ds = load_dataset("MigoXV/eeg-bme-emotion-parquet")
row = ds["train"][0]
eeg = np.asarray(row["eeg"], dtype=row["eeg_dtype"]).reshape(row["eeg_shape"])

训练集和验证集来自原训练数据，按被试级别拆分，避免同一被试同时出现在训练集和验证集。测试集来自公开测试集，真实情绪标签未公开。

Downloads last month: 24