EMG Data Set for Machine Learning and Deep Learning Problems

The Data consists of 3 parts. The dataset includes raw EMG data for 2 users. The users performed 5 hand gestures. i.e. Point, Middle Finger Extension, Closed Fist, Pinch, and Rest. The data is acquired using a Thalmic Labs Myo Armband, which has a sampling frequency of 200Hz. The participants were made to hold a gesture for 5s and relax for 3s. 6 gesture samples were acquired in one session. 12 such sessions were conducted for each gesture for each user. This gives us 72 samples for each gesture. Another data set is prepared that represents the Time Domain features extracted for each processing window The gesture windows were identified and Time domain features were extracted for each window. The feature vector for each window is arranged row-wise. Each row represents a processing window and each column in that row represents an extracted Time Domain feature. The Time Domain Features extracted are Integrated EMG (IEMG), Mean Absolute Value (MAV, MAV1 and MAV2), Simple Squared Integral (SSI), Variance (VAR), Root Mean Square (RMS), Waveform Length (WL), Average Amplitude Change (AAC), Absolute Standard Deviation Value Difference (DASDV), Myo Pulse Percentage (MYOP), Log Detector (LOG), Willison Amplitude (WAmp), Slope Sign Change (SSC) and Number of Zero Crossings (ZC) and Amplitude of First Burst (AFB). The last column represents the Gesture Class point(0), middle finger extension (1), closed-grip (2), pinch (3), rest (4). This data set can be used for gesture recognition problems.

本数据集共包含三个部分。第一部分为两名受试者的原始肌电（Electromyography, EMG）数据，受试者完成5种手部手势动作，分别为点指、伸中指、握拳、捏合及放松。数据采集设备为Thalmic Labs公司的Myo臂环，采样频率为200Hz。实验流程设置为受试者保持指定手势5秒，随后放松3秒；每一轮会话可采集6组手势样本，每名受试者的每种手势均完成12轮会话，最终每种手势可获得72组有效样本。第二部分数据集为各处理窗口提取的时域特征数据。首先对手势动作窗口进行识别，并为每个窗口提取时域特征，特征向量按行排布：每一行对应一个处理窗口，该行的每一列则对应一个已提取的时域特征。本次提取的时域特征包括：积分肌电（Integrated EMG, IEMG）、平均绝对值（Mean Absolute Value, MAV、MAV1及MAV2）、简单平方积分（Simple Squared Integral, SSI）、方差（Variance, VAR）、均方根（Root Mean Square, RMS）、波形长度（Waveform Length, WL）、平均幅度变化（Average Amplitude Change, AAC）、绝对标准差差值（Absolute Standard Deviation Value Difference, DASDV）、肌电脉冲百分比（Myo Pulse Percentage, MYOP）、对数检测器（Log Detector, LOG）、威尔逊振幅（Willison Amplitude, WAmp）、斜率符号变化数（Slope Sign Change, SSC）、过零点数量（Number of Zero Crossings, ZC）以及首次爆发振幅（Amplitude of First Burst, AFB）。数据集的最后一列对应手势类别标签：点指（0）、伸中指（1）、握拳（2）、捏合（3）、放松（4）。本数据集可应用于手势识别相关研究任务。