Simple Siamese Model with Long Short-Term Memory for User Authentication with Field-Programmable Gate Arrays

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Attached is the comments.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

English must be improved.

Author Response

Dear Reviewers, 

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper proposes the Siamese model for user authentication using electromyogram (EMG) signals with long short-term memory (LSTM) to achieve a high accuracy of over 99% with limited resources suitable for wearable devices. The paper can be enhanced by addressing the following:

1- The author should add the paper's contribution as 3-4 bullet points in the second to last paragraph of the introduction.

2- A table of symbols will enhance the understandability of the paper.

3- A table of acronyms will increase the readability of the paper.

 

 

Comments on the Quality of English Language

Minor edits and proofreading.

Author Response

Dear Reviewers, 

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors, 

in the introduction section create a table that summarizes the results from the other research papers. Based on that table discuss on disadvantages of the used methods in other literature. After that you need to explicitly write the novelty and the idea in the separate paragraph and based on that create the hypotheses in question format of your research. 

Figures are missing grids, please enlarge the font-size in the matplotlib.pyplot. Based on the graphs I'm pretty sure you have used matplotlib. The following code will allow you to adjust the font-size. 

import matplotlib.pyplot as plt SMALL_SIZE = 8 MEDIUM_SIZE = 10 BIGGER_SIZE = 12 plt.rc('font', size=SMALL_SIZE) # controls default text sizes plt.rc('axes', titlesize=SMALL_SIZE) # fontsize of the axes title plt.rc('axes', labelsize=MEDIUM_SIZE) # fontsize of the x and y labels plt.rc('xtick', labelsize=SMALL_SIZE) # fontsize of the tick labels plt.rc('ytick', labelsize=SMALL_SIZE) # fontsize of the tick labels plt.rc('legend', fontsize=SMALL_SIZE) # legend fontsize plt.rc('figure', titlesize=BIGGER_SIZE) # fontsize of the figure title

Also use the same/similar fonts in the matplotlib pyplots as in the text. Here is the code line how you can change this. 

plt.rcParams["font.family"] = "Times New Roman"

Figure 3 should be the same font as other text in the paper. 

Figure 5 should be either transformed to table or should be enlarged. The font is so small I have to enlarge the document to 305%. It would be better if you create the table instead of figure. 

Figure 7 is confusing. The ROC curve can not be just a straight line. 

Figures 9, and 10 should be either tables or enlarged. 

Conclusions and Discussions are never written together in one paragraph. Disucssion should be before the conclusion in which you explicitly discuss the conducted research, the obtained results, used methods... 

The conclusions section should be rewritten: 

1. first paragraph describes what was done in the paper

2. the second paragraph are the answers to the hypotheses defined in the introduction section that are based on the detailed discussion section 

3. the third paragraph should describe the pros and cons (advantages or disadvantages of conducted research) 

4. the fourth paragraph should contain directions for the future work that are based on the cons (disadvantages) of your proposed research methodology. 

 

 

 

 

 

 

 

 

 

 

 

 

Author Response

Dear Reviewers,

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Same as the last review comments.

Additional comparative experiments are needed. The description of the hardware that has not been tested and integrated should not be included in the manuscript. The complete network block diagram needs to be added, Table II cannot reflect the Siamese structure of the network. The table cannot be given in the form of a screenshot.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Dear Reviewer,

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

the manuscript has been corrected according to my comments and suggestions. The manuscript can be accepted. 

 

Author Response

Dear Reviewer,

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

In response to the previous comments, the author has converted two figures into tables and provided explanations for the rest. However, I disagree with some of the explanations provided. I believe the paper needs content adjustments and additional experiments. The EMG sensor part not involved in the dataset production should not be included in the manuscript, even if the test results are very similar to those of commercial equipment.  The paper needs additional comparative experiments, conducted on the same dataset but using different networks, to support the novelty of the proposed network in addressing the small-sample problem and noise measurement environment issue in EMG user authentication, as described in Section 3.1.

Author Response

Comments 1: In response to the previous comments, the author has converted two figures into tables and provided explanations for the rest. However, I disagree with some of the explanations provided. I believe the paper needs content adjustments and additional experiments. The EMG sensor part not involved in the dataset production should not be included in the manuscript, even if the test results are very similar to those of commercial equipment. The paper needs additional comparative experiments, conducted on the same dataset but using different networks, to support the novelty of the proposed network in addressing the small-sample problem and noise measurement environment issue in EMG user authentication, as described in Section 3.1.

 

Response 1: I genuinely understand your concerns. Logically speaking, it makes sense to exclude the section about the EMG sensor if it wasn't used to generate the dataset in the actual paper. Based on this, you're suggesting to remove the EMG sensor part, but I believe it is necessary for the overall structure.

In my previous papers, the reviewers frequently questioned how to create a comprehensive wearable system without any reference to the small EMG sensor. This is why I included this part. From my perspective, the most important concept is the ability to make the entire system wearable. Just making the neural network wearable is meaningless. While there are some issues to address, I hope you can consider it from the overall system perspective.

Honestly, removing the EMG sensor part is easy. However, it is crucial for the overall content, so please take this into account. If the “Editor” and the other reviewers also agree that it should be removed, I will do so without further argument in the future.

When implementing neural networks using the same or similar datasets, accuracy levels of about 92.5%, 94%, and 94.35% were achieved by our group. Given the low accuracy for practical applications, I have continuously proposed new neural networks. Currently, the performance of the proposed network is the highest and most suitable for application in wearable systems. Based on this, the following content has been added to page 11.

 “Additionally, using the same or similar datasets with different neural networks resulted in accuracies of 92.5% [39], 94% [25], and 94.35% [40], respectively, demonstrating that the proposed network has relatively higher accuracy.”

Author Response File: Author Response.pdf

Round 4

Reviewer 1 Report

Comments and Suggestions for Authors

The title of the paper is "FPGA-based Siamese-LSTM Model". The abstract describes the system model and its hardware implementation, with further details on system integration planned for future work.

If the author insists on the importance of EMC sensors to the entire paper, please complete the PCB,  test data and analysis results after completing the system integration. Currently, a single schematic cannot adequately support the significance of this content to the overall structure of the paper.

Otherwise, consider including this content in a future publication. It can be mentioned in the current paper but does not need to exist as a separate section, and a schematic is unnecessary.

Author Response

Following your comment, the section related to EMG sensors has been excluded from the entire paper.

Author Response File: Author Response.pdf