Inverse Dynamics Modeling and Analysis of Healthy Human Data for Lower Limb Rehabilitation Robots
Round 1
Reviewer 1 Report
*Experimental results should have been given in more detail.
Author Response
Thank you very much for your comment. In the experimental section, we added new experiments and showed the results with more pictures and data in the article.
(The detailed modification has been given and highlighted in the revised version).
Reviewer 2 Report
Thank you for your submission. It's an interesting topic.
However:
1. The experimental analysis is particularly exceptionally weak;
2. There is no comparison or co-analysis with state-of-the-art methods;
3. The parameter configuration used for the experiments was superficially introduced;
4. The dataset lacks detailed information;
5. The application prospects of the model are unknown.
The above five points make the manuscript merely a conference paper proposing a method, instead of a journal article with a rigorous, comprehensive analysis of a new model. Furthermore, a single-person, small-scaled data volume cannot support a strict journal article, for the results are person-dependent and academically limited — you cannot tell whether your conclusion of the model is universally correct and applicable.
Some details, for example:
9:1 – randomly? Pseudo-randomly? Orderly? If randomly, How to preserve experimental reproducibility? Why not cross-validation?
If I got it rightly, your pilot data was a one-person one. Line 253: "subjects"?
Such academic flaws are common in the article, which reduces its readability and feasibility, though I like your idea.
Because the dataset is not publicly available and no other publicly available dataset was used (of course, this is also because your study is to some level new), I cannot be positive about the reproducibility of the conclusions.
Moreover, please note the statements involving data protection, privacy, ethics, and other information about the human signal acquisition, which I did not see in the article. These elements are necessary for an article about biosignal processing experiments involving human participants.
Author Response
Many thanks for your valuable advice.
The replies are in the attachment.
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Comments for author File: Comments.pdf
Author Response
Many thanks for your valuable advice.
The replies are in the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Thank you for the revisions. Most of the issues I talked about were clarified or further refined.
However, I still emphasize that you should focus on external datasets to validate your approach. For example, the wearable sensor-based CSL-SHARE healthy human dataset published in 2020 has 17 channels containing four channels of lower limb EMG, a two-dimensional electrogoniometer to capture the knee joint in sagittal and frontal surfaces, and three-dimensional inertial signals located on the thigh and the shank — two accelerometers and two gyroscopes. It collected nearly 400 hours of data from 20 people, 22 daily and sports activities, and is well segmented and calibrated, which will also enrich the scope of your study. There is also a "wearable real-time human activity recognition system using biosensors integrated into a knee bandage" that should be an important reference for your current and future experimental design.
Even though I can reluctantly accept that you do not use external data for validation in this proof-of-concept article, you should at least cite some relevant publicly available databases and include the extension of the peer dataset-based study in future research plans (e.g. mentioned at the end of the paper). This is not only a tribute to similar studies by academic front-runners, but also gives the reading public a reference and inspiration to do their experiments for interests. For example, there are already nearly 20 studies on gait analysis, activity recognition, rehabilitation applications, feature extraction, etc., applying the CSL-SHARE dataset. Just like the feature extraction, feature stacking and feature space reduction study (https://doi.org/10.5220/0008851401350140 and https://doi.org/10.5220/0010840500003123), which can also provide a reference for one of your future research directions because I'm pretty certain that feature study should be one of your essential future research directions on your current work.
In addition, I believe the following two state-of-the art research work on wearable- biosignal-based gait and activity modeling will enrich your article's background introduction and future work: "Statistical Analysis of a Multimodal Wearable Sensor-based Human Activity Dataset." and "Biosignal Processing and Activity Modeling for Multimodal Human Activity Recognition."
Minors:
- You could use British or American English. It's totally up to you. But it is not good to MIX them. For example, in Section 2.3, "realised" is BE, but "modeling" is AE. It's not a fatal problem, but at least it's a matter of respect for language and academic rigor.
- In addition, I recommend using formulas as part of sentence grammar so that you have to pay attention to punctuation before and after the formulas, as well as to word case. You can refer to good publications by native language writers to make changes.
- In a graph containing multiple curves, I do not recommend using the black dashed line to represent one of the signals. Try a lighter curve with a slightly larger width?
Author Response
Thank you very much for your valuable advice.
All responses are inside the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Thanks for addressing most of my comments. I have no other questions regarding the scientific soundness of this paper, though the authors should still proofread the paper and correct some grammar errors.
Author Response
Thank you very much for your comment.
We re-checked the paper again and made grammar corrections.
The revised parts have been indicated in blue font.
