Enhancing Motor Imagery Classification in Brain–Computer Interfaces Using Deep Learning and Continuous Wavelet Transform
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe aim of this study was to design a lightweight, hardware-friendly CNN architecture that can be applied to resource-constrained devices like small FPGAs for real-time BCI.
Authors provide sufficient background and include relevant references. The materials and methods section is in general adequately described. Please correct:
Abstract
Line 9: Should it be a capital letter after: BCI Competition IV dataset….?
Same mistake in line line 85: Therefore, It is suitable…
Line 92: missing brackets and blank space: Fast Fourier Transform FFTis a…
Table and figures should be self-explanatory-please put explanation of abbreviations
For example: DT, SVM – Figure 5; CEMD-CNN, CSF-UAEL…
References are not prepared in the same manner: missing doi, some articles have the year of publish in bold some not. Some have pages: pp 365-374, others just 1645-1658. Please decide one manner
Author Response
Dear Reviewer,
Thank you for your thorough review of our manuscript, and for providing insightful comments and suggestions. We appreciate the time and effort you have taken to evaluate our work. Below, we have addressed each of your comments in detail and have made the necessary revisions to the manuscript.
Reviewer Comment 1: Line 9: Should it be a capital letter after: BCI Competition IV dataset….?
Response: We have revised it.
Reviewer Comment 1: Same mistake in line line 85: Therefore, It is suitable…
Response: We have revised it.
Reviewer Comment 1: Line 92: missing brackets and blank space: Fast Fourier Transform FFTis a…
Response: We have revised it.
Reviewer Comment 1: Table and figures should be self-explanatory-please put explanation of abbreviations
For example: DT, SVM – Figure 5; CEMD-CNN, CSF-UAEL…
Response: We have revised it.
Reviewer Comment 1: References are not prepared in the same manner: missing doi, some articles have the year of publish in bold some not. Some have pages: pp 365-374, others just 1645-1658. Please decide one manner。
Response: Regarding this issues, we will consult with the MDPI editors and make the necessary adjustments afterward. Therefore, no changes have been made in response to this comment.
We have carefully considered each of your comments and have made revisions accordingly. We hope that the revised manuscript meets your expectations, and we are happy to provide any further clarifications if needed.
Thank you again for your valuable input.
Reviewer 2 Report
Comments and Suggestions for AuthorsSee attachment file
Comments for author File: 
Comments.pdf
Moderate editing of English language required.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThis paper introduced a parallel structure utilizing CNN with CWT for the classification of EEG signals. The proposed technique achieved higher accuracy compared to convenitonal techniques and was written well for understanding. However, there are parts that need to be explained more clearly in order to clarify the differentiation and originality of the proposed technique.
- If the data dimension used in the signal input and processing process is specified, I think I can understand the process of processing the signal more clearly.
- The proposed CNN structure is a commonly used form, so it doesn't feel new. It is necessary to list the differentiation and characteristics of the proposed CNN structure more clearly.
- It is necessary to organize the structure and complexity comparison analysis of various CNN techniques in a table form.
Comments on the Quality of English LanguageMinor editing of English language required.
Author Response
Dear Reviewer,
Thank you for your thorough review of our manuscript, and for providing insightful comments and suggestions. We appreciate the time and effort you have taken to evaluate our work. Below, we have addressed each of your comments in detail and have made the necessary revisions to the manuscript.
Reviewer Comment 1: If the data dimension used in the signal input and processing process is specified, I think I can understand the process of processing the signal more clearly.
Response: The input data size is (40, 150, 3), please see line 270. In addition, lines 220-240 and Figure 1 describe the parameter selection during the training process in detail.
Reviewer Comment 2: The proposed CNN structure is a commonly used form, so it doesn't feel new. It is necessary to list the differentiation and characteristics of the proposed CNN structure more clearly.
Response: Our goal is to design a system suitable for small or portable devices, so the chosen CNN architecture must be as simple and efficient as possible. While CNNs are not a new technology and have already achieved success in other fields, we acknowledge this in our introduction (lines 50-68), where we explain the differences and innovations compared to existing technologies. Building on previous work, we have achieved the following results: 1) superior accuracy compared to other models; 2) the first integration of a two-branch CNN structure with CWT; 3) exploration of the impact of different preprocessing methods and classifiers on MI-EEG data; 4) minimal hardware implementation complexity.
Reviewer Comment 3: It is necessary to organize the structure and complexity comparison analysis of various CNN techniques in a table form.
Response: We have revised Table 2 to provide a brief overview of the structures of other CNN techniques and their hardware implementation complexities.
We have carefully considered each of your comments and have made revisions accordingly. We hope that the revised manuscript meets your expectations, and we are happy to provide any further clarifications if needed.
Thank you again for your valuable input.
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsAll review comments have been well revised.
