ChatGPT-Based Model for Controlling Active Assistive Devices Using Non-Invasive EEG Signals

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article presented by Dr. Tais Mota and colleagues is, in my opinion, interesting, and the topic they address is well suited for this journal. However, I have a few brief questions for the authors and would appreciate it if they could further clarify some aspects.

My first question concerns the advantages that could arise from using Transformer architectures based on self-attention mechanisms in the context of EEG signal decoding. What are these advantages, and how could they overcome the limitations of the current model?

In the future perspective of integrating a Transformer model, which EEG variables would be best suited for encoding via attention mechanisms?

Additionally, given the moderate accuracy and technological challenges, what are the next steps to be addressed to make this non-invasive brain-computer interface system ready for clinical or industrial applications?

I believe the authors could briefly discuss these points, possibly in the final section of the manuscript, as it could be useful for readers.

To conclude my questions, I believe the text is very well written and does not require any language revision. Moreover, the figures and tables are clear and well presented.

Finally, regarding the citations used by the authors, they are highly relevant, and some are extremely recent. This is particularly important as it further highlights the strength of the manuscript within the current research landscape.

Author Response

Please see attachment. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper investigates ChatGPT-Based Model for controlling active assistive devices using non-invasive EEG signals. The manuscript is overall interesting and it deals with relevant theme. Figures and Tables are clear and visible. I do not detect some major weakness of the paper. I consider that the objectives of the study are resolved in the paper. I can recommend several modification which may improve the paper: 

1. Check if citations are allowed in Abstract.
2. Revise all abbreviations. MoCap has been defined twice, for example. 
3. Add paper outline in the Introduction so that readers can navigate more easily throughout the paper. 
4. I found that it is important to elaborate more why did you use specifically ChatGPT. 
5. Revise if all information about EEG signal pre-processing is inside the text, as I consider it important given that EEG signals can have many artifacts. 
6. Elaborate more why is obtained model's accuracy range from 58-85% (this range seems large). Does something affect this range?
7. Did you perform hyperparamter search for machine learning model?
8. Do you think that incorporating more people (subjects) for getting the data will be hard, or will it affect on the results? WIll it potenatially decrease the obtained accuracy?

Author Response

Please see attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Authors have addressed my comments very successfully.

I do not require any further modifications.