[Paper Review] MindBigData 2022 A Large Dataset of Brain Signals

David Vivancos, Felix Cuesta|arXiv (Cornell University)|Dec 27, 2022

EEG and Brain-Computer Interfaces4 citations

TL;DR

MindBigData 2022 introduces a large-scale, publicly available dataset of brain signals collected using commercial and custom EEG devices across diverse human activities. The dataset enables benchmarking of machine learning models for decoding mental and physical tasks from raw EEG, supporting advancements in brain-computer interfaces with applications in healthcare, industry, and human-computer interaction.

ABSTRACT

Understanding our brain is one of the most daunting tasks, one we cannot expect to complete without the use of technology. MindBigData aims to provide a comprehensive and updated dataset of brain signals related to a diverse set of human activities so it can inspire the use of machine learning algorithms as a benchmark of 'decoding' performance from raw brain activities into its corresponding (labels) mental (or physical) tasks. Using commercial of the self, EEG devices or custom ones built by us to explore the limits of the technology. We describe the data collection procedures for each of the sub datasets and with every headset used to capture them. Also, we report possible applications in the field of Brain Computer Interfaces or BCI that could impact the life of billions, in almost every sector like healthcare game changing use cases, industry or entertainment to name a few, at the end why not directly using our brains to 'disintermediate' senses, as the final HCI (Human-Computer Interaction) device? simply what we call the journey from Type to Touch to Talk to Think.

Motivation & Objective

• To create a comprehensive, up-to-date dataset of brain signals across varied human activities to support research in neural signal decoding.
• To provide a standardized benchmark for evaluating machine learning models on raw EEG data.
• To explore the limits of EEG technology using both commercial and custom-built devices.
• To facilitate innovation in brain-computer interfaces (BCIs) with real-world applications in healthcare, industry, and entertainment.
• To advance the evolution of human-computer interaction from Type to Touch to Talk to Think.

Proposed method

• Data collection was conducted using commercial and custom-built EEG headsets to capture brain signals during a range of cognitive and physical tasks.
• Each sub-dataset was collected with detailed documentation of the headset type, setup, and experimental procedures.
• The dataset includes raw EEG signals synchronized with task labels representing mental or physical activities.
• Standardized preprocessing and labeling protocols were applied to ensure consistency and reproducibility across sub-datasets.
• The dataset is released in a structured format to support integration with machine learning frameworks.
• Applications are demonstrated through potential use cases in BCI, highlighting the transition from input modalities like touch and speech to direct brain-based interaction.

Experimental results

Research questions

• RQ1 How well can machine learning models decode mental and physical tasks from raw EEG signals using a large, diverse dataset?
• RQ2 What are the performance limits of consumer-grade and custom EEG devices in capturing task-specific brain activity?
• RQ3 How does dataset diversity across tasks and subjects affect model generalization in brain signal decoding?
• RQ4 What are the practical implications of using such a dataset for developing real-world brain-computer interfaces?
• RQ5 To what extent can raw EEG data enable the transition from traditional HCI modalities to direct brain-based interaction?

Key findings

• The MindBigData 2022 dataset comprises a large-scale collection of EEG signals from multiple subjects performing a wide range of cognitive and physical tasks.
• The dataset includes data from both commercial and custom EEG devices, enabling comparative analysis of signal quality and usability.
• Standardized data collection procedures ensure reproducibility and facilitate benchmarking across different machine learning models.
• The dataset supports the development of BCIs with applications in healthcare, industry, and entertainment, demonstrating real-world potential.
• The work enables progress toward the vision of 'thinking' as the primary human-computer interaction modality, moving beyond typing and speaking.
• The dataset is publicly available, promoting open science and accelerating research in neural signal decoding and BCI innovation.