PhysioIntent: Multimodal dataset for human intention prediction research

PhysioIntent database was acquired during the master thesis at INESC TEC. The dataset was built to research human movement intention through biosignals (electromyogram (EMG), electroencephalogram (EEG) and electrocardiogram (ECG)) using the Cyton board from openBCI [1]. Inertial data (9-axis) was also recorded with a proprietary device from INESC TEC named iHandU [2]. A camera, logitech C270 HD, was also used to record the participant’s session video, thus better supporting the post-processing of the recorded data and the agreement between the protocol and the participant activity. All data was then synchronized with the aid of a photoresistor, correlating the visual stimuli presented to the user with the signals acquired.

The acquisitions are divided into two phases, where the 2nd phase was performed to improve some setbacks encountered in the 1st phase, such as data loss and synchronization issues. The 1st phase study included 6 healthy volunteers (range of age = 22 to 25; average age = 22.3±0.9; 2 males and 4 females; all right-handed). In the 2nd phase, the study included 3 healthy volunteers (range of age = 20 to 26; average age = 22.6±2.5; 2 males and 1 female; all right-handed).

The protocol consists in the execution and imagination of some upper limb movements, which will be repeated several times throughout the protocol. There are a total of three different movements during the session: hand-grasping, wrist supination and pick and place. Each sequence of movements, imagination and execution, as well as the resting periods is called a trial. A run is a sequence of trials that end on a 60s break.

This dataset has two different phases of acquisition. Phase 1 has a total of four runs with fifteen trials each, while phase 2 has five runs with eighteen trials each. During Phase 1, on every run, each movement is imagined and executed 5 times corresponding to a total of 20 repetitions per movement during each session. On phase two, on every run, each movement was executed and imagined 6 times, resulting in 30 repetitions per movement on each session.

In phase 1, 4 different muscles, bicep brachii, tricep brachii, flexor carpi radialis, and extensor digitorum, were measured. For the EEG, the measured channels were: FP1, FP2, FCZ, C3, CZ, C4, CP3, CP4, P3, and P4. During phase 2, only one muscle, extensor digitorum, was measured. For the EEG, the channels measured were: FP1, FP2, FC3, FCz, FC4, C1, C3, Cz, C2, C4, CP3, CP4, P3, and P4.

Before the experiments, the participants were informed about the experimental protocols, paradigms, and purpose. After ensuring they understood the information, the participants signed a written consent approved by the DPO from INESC TEC.

All files are grouped by subject. You can find all the detailed descriptions of how the files are organized on the README file. Also, there is an extra folder called "PhysioIntent supporting material" where you can find some extra material including a script with functions to help you read the data, a description of the experimental protocol and the setup create for each phase. For each subject the data is organized according to the data model ("Subject_data_storage_model") where it is shown that each type of data is present in a different folder. Regarding biosignals (openBCI/ folder), there is the raw and processed data. There is an additional README file for some subjects that contains some particular details of the acquisition.

[1] Cyton + Daisy Biosensing Boards (16-Channels). (2022). Retrieved 23 August 2022, from https://shop.openbci.com/products

[2] Oliveira, Ana, Duarte Dias, Elodie Múrias Lopes, Maria do Carmo Vilas-Boas, and João Paulo Silva Cunha. "SnapKi—An Inertial Easy-to-Adapt Wearable Textile Device for Movement Quantification of Neurological Patients." Sensors 20, no. 14 (2020): 3875.

Data and Resources

Additional Info

Field Value
Author Ana Filipa Ferreira
Last Updated April 30, 2024, 16:09 (UTC)
Created August 24, 2022, 08:37 (UTC)
Citation Ferreira, A. F., Minhoto, V., Cunha, J. P., & Dias, D. (2022). PhysioIntent: Multimodal dataset for human intention prediction research [Data set]. INESC TEC. https://doi.org/10.25747/6101-JD30
Contributor Vitor Minhoto; João Paulo Cunha; Duarte Dias
DOI https://doi.org/10.25747/6101-JD30
Format .csv; mp4
Instrument iHandU; OpenBCI; EEG cap 64 channels; Webcam
Language English
Points of contact vitor.minhoto@inesctec.pt; joao.p.cunha@inesctec.pt; duarte.f.dias@inesctec.pt
Relation This dataset was collected in the context of a Master Thesis that will be presented in the following months. Link will be available when the thesis is published.
Software Python
Temporal Coverage May-August 2022
Type Biosignals, and Inertial measurements