Raw Surface Electromyography Dataset from Myo Arm Band

Published: 26-03-2021| Version 1 | DOI: 10.17632/d4y7fm3g79.1
Contributors:
Airani Mohammad Khan

Description

The data set is a part of the ongoing research work to perform gesture classification using Machine Learning and Deep Learning Techniques. The dataset consists of samples acquired from 10 consenting users. Each user performed 5 different hand gestures. The gesture classes include Index Finger Extension, Middle Finger Extension, Cylindrical Grip, Closed Grip, and Rest. The Myo Armband by Thalmic labs was used for data acquisition. The armband consists of 8 Surface EMG sensor units. The 8 sensors read data every 5ms. This data is stored in a CSV file with the timestamp. The sensor names are labeled as well. Each user wore the armband on their forearm and performed the 5 different gestures.100 instances were collected for each user for each of the gestures. one instance included performing the gesture within a window of 2 seconds followed by a rest window of 2 seconds. This pattern of flexion and extension gestures is performed to acquire the data. Each session was restricted to a batch of 25 instances considering the fatigue that sets in after constant flexion and extension gestures. The gesture data corresponding to each user are organized into 10 folders labeled as "User #", Inside each User folder, the file corresponding to each gesture are organized in a folder labeled with the gesture names "Index Finger Extension", "Middle Finger Extension", "Cylindrical Grip", "Closed Grip", and "Rest". Each CSV file is labeled as "u#-gesture-name-set-#.csv".

Files

Steps to reproduce

1. A Python Program must be developed to automate the process of data acquisition. 2. The Python Program must take input as the user's name, gesture name, and batch number. This data will be used to name the generated CSV File. 3. The user wears the Myo armband. Ensure that the Armband is connected to the computer via Bluetooth Dongle. 4. The python program will play a video to prompt the user to flex/extend the gesture or relax. The user does accordingly. 5. Each batch stops after 25 instances. 6. 4 Such batches must be made to collect 100 instances each for a gesture 7. Repeat the same process for all 10 users.