Using Wearable Sensors to Estimate Vertical Ground Reaction Force Based on a Transformer
Round 1
Reviewer 1 Report
1-The abstract is not clear, In what aspect your method is 32% more effective 11 than the RNN architecture? similarly, What kind of parameters did you compare to say it is 25% more effective than the LSTM architecture? Vague.
2- The title of the paper does not reflect the actual task you have performed.
3- Literatur review is not sufficient on IMU, you need to make a comparative analysis on that similarly for RNN, and TCN to analyse your work related to these.
4- The transformer method, that is actually used is not clearly defined (L57-L60).
5- The paper is not well organised.
6- Results are good but not analysed, there should be a discussion section on MSE for transformer and LSTM.
Author Response
1.The abstract has been modified ，Using Mean Absolute Percentage Error as a parameter to evaluate the effect of the model in this paper.
2.We have changed the title to ''Based Transformer Estimating Vertical Ground Reaction Force Using Wearable Sensors''.
3.In order to reduce the length of the article, we have not covered the related work as a separate section; we have placed related work in the first part of the introduction.
4.We have added a description of the Transformer, which can be found in more detail in the original article, as there are no complex modifications to the model in this article.
5.Added a discussion section.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this manuscript, the authors proposed a new method to estimate ground reaction forces from wearable sensors for a variety of real-world situations. I present my comments and suggestions for changes in relation to the following parts of the article.
- This manuscript could be significantly shortened to improve readability.
- The order of reference numbers is not entirely sequential. Please check again.
- There is no content corresponding to the Figure in the text (Figure 1, Figure 2).
- Please explain the variables (or functions) used in the formulas described in this paper.
- Please include the full terms before using an abbreviation (RNN, TCN, ADC, LSTM, CNN, PCA, MAPE).
- These sentences are difficult to understand and seems to have grammatical problems. I think you need to correct the part where there is a problem with the grammar (Line 67-69, 73-75, 77-80, 121-124, 166-168, 221-223, 233-236, 239-243).
- The limitations of the study have not been mentioned. Could you reflect on this? Moreover, could you identify the focus on future investigations of this topic?
- Please add a discussion section.
- Title
(Line 2) Please remove the period from the title.
- Introduction
(Line 27) In the manuscript, the reference numbers should start with [1], not [4][5].
(Line 37-41) Please separate and write concisely to help the reader's understanding.
(Line 41, 44) These sentences are difficult to understand and seems to have grammatical problems. The location of the reference numbers are different. It should be written according to the "MDPI Style Guide".
- Dataset
(Line 76) Write the full term of the abbreviation only once. The full term for FSRs is already on line 67.
(Line 91) It is described that the subjects performed fast running, jogging, slow walking, brisk walking, and walking up and down stairs without shoes (barefoot). How was the pressure insole secured to the foot? Does this method affect data output and sensor value accuracy?
Please explain in detail how the calibration of the IMU was carried out with a formula.
Did you measure the IMU data using a program like Figure 2?
Please improve the readability of this manuscript by attaching photos of the experiment wearing the pressure insole and the IMU.
- Methods
(Line 171) Please capitalize the first letter.
- Results
(Line 190-193) In this manuscript, it was confirmed that the use of Gate_MSE does not improve the overall effect, and its effect is slightly worse than that of the MSE loss function. On the other hand, it was mentioned that Gate_MSE is more effective in predicting the peak ground reaction force, which is often used to assess injury risk, muscle loss, etc. so that effective peak prediction is more important. Could you say that the performance is good just by showing the shape of the graph? A comparison of quantitative data on this should be included.
- Conclusions
It must be more specific.
Author Response
1.The order of the reference numbers has been changed.
2.Added an introduction to Fig. 1, Fig. 2 and added a picture of the sensor being worn.
3.The grammar in the text has been corrected.
4.A quantitative comparison of Gate_MSE and MSE can be viewed in Table 1.
5.Added a discussion section about the limitations of this paper's approach.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The paper titled"Based Transformer Estimating Vertical Ground Reaction Force Using Wearable Sensors" has been reviewed initially. After 1st review of the paper, a major revision was suggested. The authors have done a good job and addressed almost all of the previous issues. However, a response letter is not provided by the authors.
Author Response
1.Revised grammatical issues in the text to improve readability.
Author Response File: Author Response.pdf
