EEG Imitation tasks

The dataset used for classification consists of EEG data recorded during imitation learning tasks. The primary features and structure of the dataset are as follows:

Dataset Description
Participants:

Number of participants: 83.
Gender distribution: [e.g., 17 male, 66 female].
Age range: [e.g., 22–40 years].
All participants were right-handed and had normal or corrected-to-normal vision.

Tasks:

The dataset includes EEG recordings from three tasks:
Observation Task: Participants observed gestures performed by a demonstrator (live or via video).
Execution Task: Participants performed gestures without observing the demonstrator.
Simultaneous Observation and Execution Task: Participants simultaneously observed and executed gestures.

Data Acquisition:

EEG was recorded using [31] electrodes placed according to the [standard, e.g., 10–20 system].
Sampling rate: [e.g., 500 Hz].
Bandpass filtering: 1–50 Hz.

Features:

Raw EEG Signals: Data from all electrodes during task performance.
Extracted for each electrode using time-frequency decomposition techniques (fast Fourier transform):

Absolute Power in Alpha Band (8-13 Hz): list 1
Absolute Power in Beta Band (13–30 Hz): list 2
Relative Power in Alpha Band (8-13 Hz): list 3
Relative Power in Beta Band (13–30 Hz): list 4

Task Labels:
Observation task (live or video demonstration) - obser
Execution task (no observation) - execut
Simultaneous observation and execution task. - simult

9Demonstration Modality Labels:
Live or video format of gesture demonstration - scenario: _live or _video
Demonstrator Gender Labels - scenario: male or female

Preprocessing:

Artifacts (e.g., eye blinks, muscle movements) were removed using Independent Component Analysis (ICA).
EEG signals were segmented into epochs corresponding to task events (e.g., gesture onset).
Fast Fourier transform to get absolute and relative alpha and beta power in each trial.

These labels were used as target variables for supervised machine learning models.
This dataset forms the basis for training and testing machine learning algorithms to classify social interaction formats and identify the neural correlates associated with each classification task.

用于分类的数据集由模仿学习任务期间记录的脑电图（EEG）数据组成。数据集的主要特征和结构如下：

数据集描述
参与者：
参与人数：83人。
性别分布：[例如，17名男性，66名女性]。
年龄范围：[例如，22至40岁]。
所有参与者均为右利手，并拥有正常或矫正至正常的视力。

任务：
数据集包括来自三个任务的脑电图记录：
观察任务：参与者观察示范者（现场或通过视频）执行的手势。
执行任务：参与者在未观察示范者的情况下执行手势。
同时观察与执行任务：参与者同时观察和执行手势。

数据采集：
使用[31]个电极按照[标准，例如10-20系统]记录脑电图。
采样率：[例如，500 Hz]。
带通滤波：1-50 Hz。

特征：
原始脑电图信号：任务执行期间所有电极的数据。
使用时间-频率分解技术（快速傅里叶变换）从每个电极提取数据。
α频段（8-13 Hz）的绝对功率：列表1
β频段（13-30 Hz）的绝对功率：列表2
α频段的相对功率（8-13 Hz）：列表3
β频段的相对功率（13-30 Hz）：列表4

任务标签：
观察任务（现场或视频演示）- obser
执行任务（无观察）- execut
同时观察与执行任务 - simult

演示方式标签：
手势演示的现场或视频格式 - scenario：_live 或 _video
示范者性别标签 - scenario：male 或 female

预处理：
使用独立成分分析（ICA）去除伪迹（例如，眼睑跳动、肌肉运动）。
将脑电图信号分割成与任务事件对应的epochs（例如，手势开始）。
快速傅里叶变换以获取每个试验中的绝对和相对α和β功率。
这些标签被用作监督机器学习模型的目标变量。
该数据集构成了训练和测试机器学习算法以分类社交互动格式并识别与每个分类任务相关的神经相关特征的基础。