Data Glove with Bending Sensor and Inertial Sensor Based on Weighted DTW Fusion for Sign Language Recognition

Round 1

Reviewer 1 Report

The topic of this text is a system called Sign-Glove that is designed to recognize sign language and improve communication between people with hearing impairments and those who are not. The system uses a bend sensor and an inertial sensor node to collect data on the shape and movement of the hand in order to recognize sign language words. The text also discusses the challenges of communication between hearing-impaired people and those who are not, and the interfaces that were created to display the meaning of sign language words.

It has the potential to set an important example in terms of the subject of study. Corrections and improvements are essential based on the comments below:

1- The use of gloves and existing studies have not been adequately examined in the literature. There is very little work. Authors need improvement. For this reason, it is useful to scan the following resources:

* Seçkin, M., Seçkin, A.Ç. & Gençer, Ç. Biomedical Sensors and Applications of Wearable Technologies on Arm and Hand. Biomedical Materials & Devices (2022).

* Seçkin, A. Ç. (2021). Multi-Sensor Glove Design and Bio-Signal Data Collection . Natural and Applied Sciences Journal , Full Papers of 2nd International Congress of Updates in Biomedical Engineering , 87-93 

2- The physical, textile, electronic component and communication structure of the proposed glove has not been adequately explained. It should be shown on the images according to their placement.

3- Electronics are placed on the breadboard. It is obvious that it will cause difficulties during hand movements. Also, as far as I can see, most parts are integrated with tape. What are the handicaps, disadvantages and assumptions that these situations can cause should be emphasized in the discussion and conclusion part.

4- Figure-4 is not clear enough. Instead, sensor signal patterns that occur during movements should be shared.

5- The quality of the publication will increase if comparisons are made with similar or competing studies. I strongly recommend.

6- Obtained data should be submitted to the editor and reviewer. In this way, it can be checked whether the work is plagiarism, fiction or misleading. Otherwise, the article should be considered suspicious and rejected.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a nice article, which describes the development of a Sign-Glove system to recognize sign language and a proposed DTW data fusion algorithm. Considering that this is a proof-of-concept experimental work, the article could be accepted with some (if possible) improvements:

- More state-of-the-art references must be included. All references are from 2016 and below.

- The breadboard Sign-Glove system is not, at this stage, a wearable proof system. One wire detached is enough to make all the system to stop working. Hence, the hardware could be implemented in a Printed Circuit Board (PCB),

- The Arduino Uno board is a rather old and high-sized board. Authors should consider using a smaller and faster board (ex: Seeed XIAO or Seeed XIAO BLE Sense boards).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Authors use accelerometers and bending sensos to detect the motions (static and dynamic status) of fingers and hands for recognizing the sign language of 20 words. The texts and results of this manuscript are not well. My suggestion is that it needs the native English speaker to revise its grammar and texts. The major problems have to be modified.

1. In the texts, the number of references should be increased. In “Introduction”, authors describe the many concepts, but these concepts lack the references to support them.

2. My suggestion is to combine “introduction” and “Related works”.

3. Figure 1 shows the sign-glove device used in this study. So, does the contribution of this study be the DTW? If yes, I think the materials of this study is weaker.

4. In line 134, “we can guess….”. I do not agree the description of this sentence. My suggestion is authors should shows the results of these studies, or uses these results as the benchmarks to describe why do these studies not approach to criteria.

5. In Table 1, the performance of some studies are better than this study, and recognize more words. Thus, authors should explain what’s the benefits in this study.

6. In Fig. 2, do “Recognize hand-gesture” and “Recognize hand-shape” be executed in device or PC?

7. authors should combine two system, bending sensors and WondeSense, as one one system, only use a wireless device to transmit the acquired signal from bending sensors and wondesense.

8. Authors should describe the specifics of used system, like as bending sensor, wondesense, wondebox, and sampling rate et al.

9. In Eq(1), (2), there are the wrongs. Please check. Moreover, the variables should be explained more clear, and show the values of these variables.

10 How to recognize the words should be explained more clear.

11. I do not agree the board of Arduino is a result of this study. My suggestion is to show the specifics of these devices and sensors in “Methods”, not show the figures.

12. My suggestion is to delete Fig. (8) and (9).

13. How to get the weighting values should be explained more clear.

14. My suggestion is to modify the description of performances and qualities of proposed method. The results in Fig. 11 are too weak.

15. If do not use DTW, authors should explain how to recognize sign-language without DTW.   

16. I suggest to modify the text in “Conclusion”, and add the “Discussions”.

  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have revised the article to a good level. After making minor revisions and improvements, it can be accepted. 1. It would be good to enrich the comparison table. For example authors can add studies done with EMG. 2. If possible, references written in Japanese characters should be translated into English. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Authors use accelerometers and bending sensos to detect the motions (static and dynamic status) of fingers and hands for recognizing the sign language of 20 words. The texts and results of this manuscript are not well. My suggestion is that it needs the native English speaker to revise its grammar and texts. The major problems have to be modified.

1. In the texts, the number of references should be increased. In “Introduction”, authors describe the many concepts, but these concepts lack the references to support them.

DONE

2. My suggestion is to combine “introduction” and “Related works”.

 

Too bad! Authors only delete “Related works”. They do not organize and revise the text of two sections. Moreover, they do not modify the numbers of the other sections.

 

3. Figure 1 shows the sign-glove device used in this study. So, does the contribution of this study be the DTW? If yes, I think the materials of this study is weaker.

 

According to Table 2, the contribution of this study is weighting DWT. The technical soundness is low.

 

4. In line 134, “we can guess….”. I do not agree the description of this sentence. My suggestion is authors should shows the results of these studies, or uses these results as the benchmarks to describe why do these studies not approach to criteria.

 

I do not agree the sentence “But the computational cost of multiple IMU data is high than multiple bending sensors.” What’s “computational cost”? it should be defined clearly, like as spending time or spending money (an expensive hardware). My suggestion is authors should describe more clear to let readers understand the benefit of this study.

 

 

5. In Table 1, the performance of some studies are better than this study, and recognize more words. Thus, authors should explain what’s the benefits in this study.

DONE

6. In Fig. 2, do “Recognize hand-gesture” and “Recognize hand-shape” be executed in device or PC?

 

My suggestion is Hand-Gesture Featureè Hand-Gesture signal.

 

7. authors should combine two system, bending sensors and WondeSense, as one one system, only use a wireless device to transmit the acquired signal from bending sensors and wondesense.

 

How to synchronize?

 

8. Authors should describe the specifics of used system, like as bending sensor, wondesense, wondebox, and sampling rate et al.

DONE

9. In Eq(1), (2), there are the wrongs. Please check. Moreover, the variables should be explained more clear, and show the values of these variables.

10 How to recognize the words should be explained more clear.

Too simple. The hand action is a continuous motion, and each action spend the different time. Thus, how to segment data and make the features should be explained more clear. If not, readers do not have any interesting to cite this paper.

 

11. I do not agree the board of Arduino is a result of this study. My suggestion is to show the specifics of these devices and sensors in “Methods”, not show the figures.

DONE

12. My suggestion is to delete Fig. (8) and (9).

DONE

13. How to get the weighting values should be explained more clear.

DONE

14. My suggestion is to modify the description of performances and qualities of proposed method. The results in Fig. 11 are too weak.

 

“Combo” should be “Combination”. What’s caption of x-axis? Accuracy? If yes, add it. Moreover, if x-axis is accuracy, does not have 120% index.

 

15. If do not use DTW, authors should explain how to recognize sign-language without DTW.

DONE  

16. I suggest to modify the text in “Conclusion”, and add the “Discussions”.

DONE

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 3

Reviewer 3 Report

Authors responded my comments.