MadNad's picture
chore: updated README.
d92b686
metadata
license: mit
language:
  - en
configs:
  - config_name: raw
    data_files: '**/raw/*.csv'
  - config_name: fft
    data_files: '**/fft/*.csv'
tags:
  - gestures
  - classification
  - robotics

Real Time Gesture Recognition Dataset

This repository contains the dataset for the Real-Time-Gesture-Control-System-for-Precision-Control project that focuses on the development of a real-time gesture control system using the uMyo EMG sensors.

Description

The dataset is divided into two folders: dynamic_gestures and static_gestures. The first contains loosely-timed sessions of gesture recordings where the gesture is repeated several times between intervals of rest. The second contains a single recording of each gesture, where the gesture is held for a few seconds. The gestures are: baseline, fist, up, down, peace, and lift. New gesture recordings and more gestures will be uploaded soon. Every subject has two types of recordings: the raw EMG data in the raw folder and the FFT data in the fft folder. Each individual recording is a CSV file dedicated to a single gesture performed by a subject.

Structure

dynamic_gestures
β”œβ”€β”€ session-DD_MM_YY
β”‚   β”œβ”€β”€ subject1
β”‚   β”‚   β”œβ”€β”€ fft
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_baseline.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_fist.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_up.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ ...
β”‚   β”‚   β”œβ”€β”€ raw
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_baseline.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_fist.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_up.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ ...
β”‚   β”œβ”€β”€ subject2
β”‚   └── ...
β”œβ”€β”€ session-DD_MM_YY
└── ...
static_gestures
β”œβ”€β”€ session-DD_MM_YY
β”‚   β”œβ”€β”€ subject1
β”‚   β”‚   β”œβ”€β”€ fft
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_baseline.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_fist.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ fft_subject1_up.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ ...
β”‚   β”‚   β”œβ”€β”€ raw
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_baseline.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_fist.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ subject1_up.csv
β”‚   β”‚   β”‚   β”œβ”€β”€ ...
β”‚   β”œβ”€β”€ subject2
β”‚   └── ...
β”œβ”€β”€ session-DD_MM_YY
└── ...

Contents

Raw EMG Data

The raw EMG data is a header-less CSV file with the columns containing 8 successive EMG readings from the corresponding sensor:

      [sensor 1]  [sensor 2]  [sensor 3]  [sensor 4]  [sensor 5]
[0]:  [8 readings][8 readings][8 readings][8 readings][8 readings]
[1]:  [8 readings][8 readings][8 readings][8 readings][8 readings]
[2]:  [8 readings][8 readings][8 readings][8 readings][8 readings]
[3]:  [8 readings][8 readings][8 readings][8 readings][8 readings]
...

Each row can be treated as a single 5-dimensional data unit in time.

FFT EMG Data

The FFT EMG data is a header-less CSV file with the columns containing 4 FFT bins with ~143 Hz resolution from the corresponding sensor:

      [sensor 1] [sensor 2] [sensor 3] [sensor 4] [sensor 5]
[0]:  [4 bins]   [4 bins]   [4 bins]   [4 bins]   [4 bins]
[1]:  [4 bins]   [4 bins]   [4 bins]   [4 bins]   [4 bins]
[2]:  [4 bins]   [4 bins]   [4 bins]   [4 bins]   [4 bins]
[3]:  [4 bins]   [4 bins]   [4 bins]   [4 bins]   [4 bins]
...