Gumpy: a Python toolbox suitable for hybrid brain-computer interfaces.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Hybrid brain-computer interfaces (BCIs) integrate multiple biosignals for improved performance.
• Existing toolboxes may lack comprehensive features for modern BCI research, including advanced deep learning models.

Purpose of the Study:

• To introduce gumpy, a novel, free, and open-source Python toolbox for developing hybrid BCIs.
• To provide researchers with a versatile platform for processing and decoding electroencephalography (EEG) and electromyography (EMG) signals.

Main Methods:

• Gumpy integrates a wide array of signal processing techniques utilized in the BCI field over the past two decades.
• The toolbox incorporates diverse classification algorithms, ranging from traditional machine learning to deep neural networks.
• It supports EEG and EMG biosignal analysis, visualization, real-time streaming, and decoding.

Main Results:

• Demonstrated efficacy through offline studies in predicting movement from EEG motor imagery and decoding grasp movements from sEMG.
• Successfully applied in real-time applications, including robot arm control via SSVEP and prosthetic hand control using sEMG.
• Gumpy-based deep learning models achieved accuracy comparable to or exceeding state-of-the-art results on benchmark datasets.

Conclusions:

• Gumpy empowers researchers to implement online hybrid BCIs with advanced EEG and EMG signal processing and decoding techniques.
• The toolbox's deep learning capabilities demonstrate potential to match or surpass current state-of-the-art accuracy.
• Gumpy facilitates the development of more robust BCI decoding algorithms, paving the way for future BCI advancements.