Attention-Based Hybrid Deep Learning Network for Human Activity Recognition Using WiFi Channel State Information

Round 1

Reviewer 1 Report

The paper proposes development of a deep-learning framework that enables human activity recognition using WiFi channel state information measurements, eliminating the need for manual feature extraction. The paper provides good introduction. Results are well-presented. Overall a well-structured paper.

1. What is the main question addressed by the research? AF: The paper looks into WiFi based sensing methods to recognise Human Activity using deep learning techniques.    2. Do you consider the topic original or relevant in the field? Does it address a specific gap in the field? AF: The chosen area, indeed, is worth exploring. Recently, there has been a good deal of research in HAR using smart phones or wearable devices. This paper utilises WiFi based sensing which is unique and makes this article interesting.    3. What does it add to the subject area compared with other published material? AF: As stated above (Q2), it uses different source of data as compared to other researchers which makes it unique.   4. What specific improvements should the authors consider regarding the methodology? What further controls should be considered? AF: A more critical analysis of the results can be added.   5. Are the conclusions consistent with the evidence and arguments presented and do they address the main question posed? AF: Yes. The conclusion summarises the outcomes. The authors also discuss the limitations.    6. Are the references appropriate? AF: Yes   7. Please include any additional comments on the tables and figures.

 

Look fine to me.

Author Response

Dear reviewers, 

We would like to express our gratitude for your comments. Our response and the corresponding revision have been addressed in this letter. In the revised paper, the modifications are written in yellow. With this revision, we hope the manuscript gives a better insight into the proposed idea and is in a better shape of presentation.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors introduces a versatile framework for recognizing human activities by utilizing CSI data and evaluates its effectiveness on two public dataset. Results demonstrated that the proposed method surpasses previous state-of-the-art techniques. However, some major revisions have to be made before the publication of this paper.

1.     Gave a detailed information of the three participants of CSI-HAR dataset, e.g. the age.

2.     In StanWiFi dataset, the original dataset had seven categories, but only six were used in this study to facilitate comparison with previous works. Please explain why the activity of pick up are not included in previous works.

3.     Is there exist individual difference of different subjects? it was suggested to add the results of each subject.

4.     It is not clear for the part of method, especially, how the method of PCA can be used for noise de-noising? It was suggested to plot and evaluate the SNR of CSI before and after de-noise.

5.     And why the subsequent five components of PCA are utilized for feature extraction. What is the cumulative contribution rate of the components?

6.     The title of Figure.10 is wrong. “Figure 10. Recognition performance of the proposed network: (a) without attention mechanism (b)with attention mechanism”

Author Response

Dear reviewers, 

We would like to express our gratitude for your comments. Our response and the corresponding revision have been addressed in this letter. In the revised paper, the modifications are written in yellow. With this revision, we hope the manuscript gives a better insight into the proposed idea and is in a better shape of presentation.

Author Response File: Author Response.pdf

Reviewer 3 Report

1.Is there an impact on the overall detection accuracy before and after using PCA to de-noise?

2. The various evaluation indicators in Tables 5 and 6 are the mean values. What do the values after parentheses represent and how can they be calculated? Please explain

3. Is there an error in the presentation of the CSI-HAR dataset, with only three participants?

Minor editing of English language required, please further check the spellings and grammars 

Author Response

Dear reviewers, 

We would like to express our gratitude for your comments. Our response and the corresponding revision have been addressed in this letter. In the revised paper, the modifications are written in yellow. With this revision, we hope the manuscript gives a better insight into the proposed idea and is in a better shape of presentation.

Author Response File: Author Response.pdf