Real-Time Dynamic and Multi-View Gait-Based Gender Classification Using Lower-Body Joints

Round 1

Reviewer 1 Report

The paper addresses a significant research area related to biometric/gender identification. The methodology is well organized and adequate. 

An appropriate english review is necessary.

Author Response

Dear sir, the attached file is reply to reviewer 1 comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

The discussed research is on binary gender classification based on lower joint positions estimated by Kinect. Gender is established using a simple linear regression model.

This paper is apparently a minor extension of the authors' ACM MTAP paper [37].

A major drawback is that gender classification is a rather easy problem with solutions of already almost 100% accuracy and this method is too simplistic. Maybe train a neural net?

Line 296: How do you get the intercept alpha and how do you calculate the Hip (with three lines) and LBJ (with three lines)?

This could be a paper on a workshop or maybe next year's FG, but not a journal.

Author Response

Dear Sir, the attached file is reply to reviewer 2 comments.

Author Response File: Author Response.docx

Reviewer 3 Report

For the purpose of classifying gender, this paper proposes a logistic regression-based model for gender classification using lower body joints. Different joints’ features are extracted using the Kinect Sensor. And different statistical strategies are employed to verify that the lower body joints of each gender are appropriate for gender classification. The results show that the lower body joints are statistically significant for gender classification. The authors created their own dataset  that includes 3D gait features for 373 subjects recorded by Microsot Kinect. The proposed method achieves performance improvement in the real-time experiments, which shows good efforts.

The organization of the paper is sound. No important sections are missing. However, the paper could be improved in several aspects:

1.In addition to binary logistic regression, gender classification can use other classification models such as support vector machine (SVM). If other classification models are compared in the experiment, the experiment can be more sufficient.

2. As the proposed method has real-time gait recognition capabilities, it is better to include the actual recognition time in the experiment.

3. The method [37] used in the comparison experiment should be introduced in “2. Literature Review”.

4. ”The study [48]”in line 124 should be in the list of references.

Author Response

Dear Sir, the attached file is reply to reviewer 3 comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The provided rebutall is insufficient.

Authors only added one self citation and one incomplete table with what pretends to be an experiment with some kind of artificial neural nets.

Why this is a reject: heavy typesetting is necessary, citations to used methods are missing, topic is too narrow and rather easy.

Do not resubmit.

Author Response

Dear Reviewer,

We are thankful to the editor and reviewers for their valuable comments which help us improving quality of the manuscript.

As per above comments the manuscript is revised and the suggested citations by reviewer 2 are added in the revised manuscript at Table No. 10.

Author Response File: Author Response.pdf