A dataset for the investigation of upper limb torque prediction from EMG signals

Creators

• Quesada, Lucas (Contact person)^{1, 2}
• Verdel, Dorian (Researcher)³
• Bruneau, Olivier (Supervisor)⁴
• Berret, Bastien (Supervisor)^{1, 2}
• Amorim, Michel-Ange (Supervisor)^{1, 2}
• Vignais, Nicolas (Supervisor)^{5, 6}

Description

Motivation

EMG-driven exoskeleton assistance requires the use of intention detection models to associate electromyographic signals with some feature of human movement, such as angular position, velocity, or joint torque. The goal of this dataset is to provide data that allows the benchmarking of such models for a variety of movements.

Short description

This dataset includes kinematic, dynamic and electromyographic data from 17 participants (11 males, age 28.2 ± 7 years, height 175.4 ± 7 cm, weight 70 ± 11 kg). These data were collected during the performance of a sagittal plane upper limb tracking task for single joint (elbow flexion/extension) and multiple joint (elbow and shoulder flexion/extension).

Methodology

A detailed description of the data collection methodology can be found here: https://www.biorxiv.org/content/10.1101/2024.01.11.575155v1

Data Description

The dataset consists of 17 folders, one for each participant. Inside each folder you will find:

• A metadata file (SXX.json) containing information about the subject: age, sex, weight, height, and upper limb masses and lengths.
• An OpenSim model file (scaledModel.osim) containing a scaled upper limb model of the given subject.
• A MVC folder containing EMG data from maximal voluntary contraction trials.
• A SJ folder containing trial folders for the single joint condition (elbow flexion/extension).
• A MJ folder containing test folders for the multi-joint condition (elbow and shoulder flexion/extension).

EMG Data

The files containing EMG data have the following header:

TIME, DELTAnt, DELTMed, DELTPost, PECT, LATI, TRILong, TRILat, TRIMed, BICLong, BICShort, BRA, BRD

This corresponds to a time stamp and EMG signals from the anterior, median and posterior deltoids, pectoralis major, latissimus dorsi, long, lateral and median triceps, long and short biceps, brachioradialis and brachialis.

MVC Data

The MVC folder contains two files: emgMVCElbow.csv and emgMVCShoulder.csv. They were collected during the realization of maximum voluntary contraction tasks and contain raw EMG data sampled at 2 kHz.

Trial Data

Single-joint condition

Each single-joint trial contains 5 files:

• emgFilt.csv: Filtered EMG signals (20–450 Hz bandpass filter, rectification, 3Hz lowpass, normalized with MVC), sampled at 100 Hz.
• emgRaw.csv: Raw EMG signals, sampled at 2 kHz.
• humanPositions.csv: Angular position of the elbow in radians, sampled at 100 Hz.
• humanVelocities.csv: Angular velocity of the elbow in rad/s, sampled at 100 Hz.
• muscleTorque.csv: Joint torque of the elbow in N·m, sampled at 100 Hz.

Multi-joint condition

Each multi-joint trial contains 5 files:

• emgFilt.csv: Filtered EMG signals, same processing as single-joint, sampled at 100 Hz.
• emgRaw.csv: Raw EMG signals, sampled at 2 kHz.
• humanPositions.csv: Angular positions of the upper limb in radians, sampled at 100 Hz.
• humanVelocities.csv: Angular velocities of the upper limb in rad/s, sampled at 100 Hz.
• muscleTorque.csv: Joint torques of the upper limb in N·m, sampled at 100 Hz.

For multi-joint trials, kinematic and torque files use the following headers:

TIME, elv_angle, shoulder_elv, shoulder_rot, elbow_flexion, pro_sup, deviation, flexion

Columns of interest for sagittal plane movements: shoulder_elv (shoulder flexion/extension), elbow_flexion (elbow flexion/extension).