Facial Emotion Recognition Analysis Based on Age-Biased Data

Round 1

Reviewer 1 Report

This paper analyzed the importance of age-biased data in recognizing the emotions of facial expressions using deep learning approach. The paper is easy to follow. However, it needs major revision in terms of following comments:

1.      Introduction section is weak in this paper. The authors support this paper with a smaller number of existing works. Limitations in the state-of-the-art are not revealed. I suggest authors to expand introduction and existing works in the paper. Support your introduction with paper contributions and end introduction with organization of paper.

2.      I suggest authors to evaluate model on other parameters than accuracy as well. The entire discussion in on accuracy only which is not justified.

3.      Major emphasis of this article is how higher accuracy is achieved with data is trained on age attributes. However, authors don’t draw its comparison with any other works and don’t explain the reason for higher performance on age basis in the discussion section.

4.      In sub-sections 3.2 and 3.3, I will like to see comparisons drawn with other works and how your model outperformed them.

5.      More description related to the dataset construction is required. How your creation differentiated data than that of FER2013 and MMA FACILE EXPRESSION.

Author Response

1. Introduction section is weak in this paper. The authors support this paper with a smaller number of existing works. Limitations in the state-of-the-art are not revealed. I suggest authors to expand introduction and existing works in the paper. Support your introduction with paper contributions and end introduction with organization of paper.
   
   Response1: Related works were added in lines 65-85. Line 89-93 shows compare between existing work and ours, and it shows also a limitation of existing works.
2. I suggest authors to evaluate model on other parameters than accuracy as well. The entire discussion in on accuracy only which is not justified.
   
   Response2: The formula for the new metric was summarized in line 149, and Table 3 was modified including the resulting metrics.
3. Major emphasis of this article is how higher accuracy is achieved with data is trained on age attributes. However, authors don’t draw its comparison with any other works and don’t explain the reason for higher performance on age basis in the discussion section.
   
   Response3: We further write a reason for the high accuracy of age-specific data on 206-210 lines, and show public benchmark results of fer2013 and accuracy comparison with our model in Table 5.
4. In sub-sections 3.2 and 3.3, I will like to see comparisons drawn with other works and how your model outperformed them.
   
   Response 4: In Table 5, we additionally prepared the accuracy, recall, precision, and f1 score comparison data of the existing dataset and our work.
5. More description related to the dataset construction is required. How your creation differentiated data than that of FER2013 and MMA FACILE EXPRESSION.
   
   Response 5: The data used in this study used image data used in FER2013 and MMA FACILE EXPRESSION. Image data were distinguished by forming two groups of kids and adults through age. And the image data of MMA FACILE EXPRESSION was converted into color data, and the data of FER2013 was converted into grayscale.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

I have reviewed the manuscript “Facial Emotion Recognition Analysis Based on Age-Biased", Manuscript number: applsci-1826072 that has been submitted for publication in the Applied Sciences MDPI Journal, and I have identified a series of aspects that in my opinion must be addressed in order to bring a benefit to the manuscript.

In this article, the authors use the Convolutional Neural Networks (CNN) algorithm in computing the emotion recognition accuracy to investigate on the importance of age-biased data in recognizing the emotions of facial expressions.

The article under review will be improved if the authors address the following aspects in the text of the manuscript:

Ø Comments concerning the "Abstract" section of the paper:

·       The authors wrote a well-structured abstract that summarizes in one paragraph of less than 300 words that summarizes the major aspects of the entire paper in a prescribed sequence that includes: 1) the overall purpose of the study and the research problem(s) they investigated; 2) the basic design of the study; 3) major findings or trends found as a result of their analysis; and, 4) a brief summary of their interpretations and conclusions.

·       However, I suggest to the authors to correct the first sentence “In this paper, we analyzed the importance of age-biased data in recognizing the emotions of facial expressions.” Should be corrected as follows: “This paper aims at analyzing the importance of age-biased data in recognizing the emotions 6 of facial expressions.”

·       It seems like there is a mistake in line 8: authors should check if the writing “adults (≥14) and kids (≥13).” is correct. I think the correct writing is “adults (≥14) and kids (≤13). ”

Ø Comments regarding the "Introduction" section

·       In a scientific article, an important role of the “Introduction” section is to offer authors the opportunity to analyze in detail the current state of knowledge related to the manuscript’s topic, namely the most relevant scientific works in the field in order to be able to contextualize their study and familiarize the readers with what has been done in the literature and what advancement their study aims to bring in addition to the existing body of knowledge.

·       The authors of the article under review managed to reference key previous studies related this paper. Then, they highlighted some limitations of the referenced articles and state the need for their work by presenting it as an opposition between what the scientific community currently has and what it needs or wants.  Then, the author indicated what they have done in their conducted study to address the existing need, namely the task they set out to solve. Unfortunately, the authors failed to preview the remainder of the paper to prepare the readers for the subsequent structuring of the manuscript.

·       As an suggestion, the authors can use this sentence at the end of the introduction to preview the remainder of the paper: “The reminder of this article is organized as follows: Section 2 presents the experimental method adopted to carry out the study. The results obtained and their discussion are presented in Section 3. The conclusions ends the paper in Section 4.”

Ø Comments concerning the "Results and discussion" section:

·       The results presented into this section should provide a concise and precise description of the experimental results, their interpretation as well as the experimental conclusions that can be drawn.

·        The authors must assume more clearly their own results and their own original contributions to the current state of knowledge.

·       After having analyzed the results, the authors should move forward to the analysis within which the authors must extend the comparison between their developed strategy from the manuscript and other strategies that have been designed and used in the literature for this purpose.

·       There are a lot of valuable studies in the scientific literature related to the subject of the manuscript to which the authors can compare, and this comparison will highlight even more the novel aspects that their paper brought in contrast to the existing studies. Consequently, to validate the usefulness of their research, within the analysis part of this section, the authors should compare their method from the manuscript and other ones that have been developed and used in the literature for the same purpose.

·       I consider that the manuscript will be considerably improved if the authors state clearly within the manuscript who are the potential beneficiaries of the devised approach (proving once again the usefulness of the conducted study) and how will these potential beneficiaries be able to make use of the devised approach in their everyday activities.

Ø Comments concerning the "Conclusions" section:

·       The authors should avoid simply summarizing the aspects that they have already stated in the body of the manuscript. Instead, they should interpret their findings at a higher level of abstraction than in the Discussion section. The authors should highlight whether, or to what extent they have managed to address the necessity identified within the Introduction section. The authors should avoid restating everything they did once again, but instead they should emphasize what their findings actually mean to the readers, therefore making the Conclusions section interesting and memorable to them.

·        The authors should write a short Conclusions section, in fact they should conclude in just a few sentences given the rich discussion section that they will have devised in the body of the paper. I strongly recommend the authors to overcome the temptation to repeat material from the Introduction or from the rest of the paper only to make the Conclusions section longer under the false belief that a longer Conclusions section will seem more impressive.

·       Authors should be aware they should “Conclusions” instead of “Conclusion” as they did.

 

Author Response

Comments and Suggestions for Authors

Dear Authors,

I have reviewed the manuscript “Facial Emotion Recognition Analysis Based on Age-Biased", Manuscript number: applsci-1826072 that has been submitted for publication in the Applied Sciences MDPI Journal, and I have identified a series of aspects that in my opinion must be addressed in order to bring a benefit to the manuscript.

In this article, the authors use the Convolutional Neural Networks (CNN) algorithm in computing the emotion recognition accuracy to investigate on the importance of age-biased data in recognizing the emotions of facial expressions.

The article under review will be improved if the authors address the following aspects in the text of the manuscript:

Ø Comments concerning the "Abstract" section of the paper:

• The authors wrote a well-structured abstract that summarizes in one paragraph of less than 300 words that summarizes the major aspects of the entire paper in a prescribed sequence that includes: 1) the overall purpose of the study and the research problem(s) they investigated; 2) the basic design of the study; 3) major findings or trends found as a result of their analysis; and, 4) a brief summary of their interpretations and conclusions.
• However, I suggest to the authors to correct the first sentence “In this paper, we analyzed the importance of age-biased data in recognizing the emotions of facial expressions.” Should be corrected as follows: “This paper aims at analyzing the importance of age-biased data in recognizing the emotions 6 of facial expressions.”
  
  Response 1: We modified it
• It seems like there is a mistake in line 8: authors should check if the writing “adults (≥14) and kids (≥13).” is correct. I think the correct writing is “adults (≥14) and kids (≤13). ”
  
  Response 2: We modified it

Ø Comments regarding the "Introduction" section

• In a scientific article, an important role of the “Introduction” section is to offer authors the opportunity to analyze in detail the current state of knowledge related to the manuscript’s topic, namely the most relevant scientific works in the field in order to be able to contextualize their study and familiarize the readers with what has been done in the literature and what advancement their study aims to bring in addition to the existing body of knowledge.
• The authors of the article under review managed to reference key previous studies related this paper. Then, they highlighted some limitations of the referenced articles and state the need for their work by presenting it as an opposition between what the scientific community currently has and what it needs or wants.  Then, the author indicated what they have done in their conducted study to address the existing need, namely the task they set out to solve. Unfortunately, the authors failed to preview the remainder of the paper to prepare the readers for the subsequent structuring of the manuscript.
• As an suggestion, the authors can use this sentence at the end of the introduction to preview the remainder of the paper: “The reminder of this article is organized as follows: Section 2 presents the experimental method adopted to carry out the study. The results obtained and their discussion are presented in Section 3. The conclusions ends the paper in Section 4.”
  
  Response 3: We added it the end of the introduction

Ø Comments concerning the "Results and discussion" section:

• The results presented into this section should provide a concise and precise description of the experimental results, their interpretation as well as the experimental conclusions that can be drawn.
• The authors must assume more clearly their own results and their own original contributions to the current state of knowledge.
• After having analyzed the results, the authors should move forward to the analysis within which the authors must extend the comparison between their developed strategy from the manuscript and other strategies that have been designed and used in the literature for this purpose.
• There are a lot of valuable studies in the scientific literature related to the subject of the manuscript to which the authors can compare, and this comparison will highlight even more the novel aspects that their paper brought in contrast to the existing studies. Consequently, to validate the usefulness of their research, within the analysis part of this section, the authors should compare their method from the manuscript and other ones that have been developed and used in the literature for the same purpose.
• I consider that the manuscript will be considerably improved if the authors state clearly within the manuscript who are the potential beneficiaries of the devised approach (proving once again the usefulness of the conducted study) and how will these potential beneficiaries be able to make use of the devised approach in their everyday activities.
  
  Response 4: In Table 5, we compared our model with the previously public FER2013 benchmark, and in Table 6, we organized the accuracy, recall, precision, and f1 score of the existing FER2013, MMA FACILE EXPRESSION dataset, and our dataset.

Ø Comments concerning the "Conclusions" section:

• The authors should avoid simply summarizing the aspects that they have already stated in the body of the manuscript. Instead, they should interpret their findings at a higher level of abstraction than in the Discussion section. The authors should highlight whether, or to what extent they have managed to address the necessity identified within the Introduction section. The authors should avoid restating everything they did once again, but instead they should emphasize what their findings actually mean to the readers, therefore making the Conclusions section interesting and memorable to them.
• The authors should write a short Conclusions section, in fact they should conclude in just a few sentences given the rich discussion section that they will have devised in the body of the paper. I strongly recommend the authors to overcome the temptation to repeat material from the Introduction or from the rest of the paper only to make the Conclusions section longer under the false belief that a longer Conclusions section will seem more impressive.
• Authors should be aware they should “Conclusions” instead of “Conclusion” as they did.
  
  Response 5: Line 244-248 shows additional conclusion and our performance

Author Response File: Author Response.docx

Reviewer 3 Report

 

1.     The abstract should include more numerical results to clarify the results.

2.     Clarify research contributions in the introduction section in details.

3.     It should be explained why the three methods were chosen: MobileNet-V2, SE-ResNeXt50(32×4d), and Res- 90 and NeXt101(64×4d).

4.     Compare the features used in terms of features, size vector and feature selection time with other methods.

5.     Evaluates measures use more than accuracy Each sensitivity, ........ , ROC, F-measure, AUC.

6.     Comparing the results with another research using the same dataset.

7.     Use more than one datasets to confirm the suggested  framework.

8.     References need updating and can be used

https://doi.org/10.3390/s22031184

https://doi.org/10.1016/j.bspc.2021.103441

 

 

 

Author Response

1. The abstract should include more numerical results to clarify the results.
   Response 1: We added the best accuracy of our model and compare the score in line 14
   
   
2. Clarify research contributions in the introduction section in details.
   Response 2: In Line 89-94, we wrote comparisons and achievements with existing works and our models.
3. It should be explained why the three methods were chosen: MobileNet-V2, SE-ResNeXt50(32×4d), and Res- 90 and NeXt101(64×4d).
   Response 3: Line 122-125 specifies the reason for selecting the architecture based on Figure 3.
4. Compare the features used in terms of features, size vector and feature selection time with other methods.
   Response 4: Table 5 shows the comparison results with other works and our models.
5. Evaluates measures use more than accuracy Each sensitivity, ........ , ROC, F-measure, AUC.
   Response 5: Table 3 shows recall, precision, and F1 score.
6. Comparing the results with another research using the same dataset.
   Response 6: Table 6 shows several metrics compared to the same dataset as our model and the model learned with the same architecture.
7. Use more than one datasets to confirm the suggested framework.
   Response 7: The dataset used in the study combined two datasets into one dataset to make it one new dataset. Therefore, there is no need to use more than one data set to confirm the suggested framework.
8. References need updating and can be used

https://doi.org/10.3390/s22031184

https://doi.org/10.1016/j.bspc.2021.103441

Response 8: We modified the reference

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

All my comments are addressed. My decision is to accept in current form.

Reviewer 2 Report

The comments and corrections made as part of my first review have been addressed. The authors have taken into consideration all my questions and comments. Nevertheless, It's my opinion that it will be beneficial to the paper if the authors quote the following papers alongside the reference the reference [1]

Bamisile, Olusola, et al. "Comparison of machine learning and deep learning algorithms for hourly global/diffuse solar radiation predictions." International Journal of Energy Research 46.8 (2022): 10052-10073.

Mukhtar, Mustapha, et al. "Development and Comparison of Two Novel Hybrid Neural Network Models for Hourly Solar Radiation Prediction." Applied Sciences 12.3 (2022): 1435.

Bamisile, Olusola, et al. "Impact of economic development on CO2 emission in Africa; the role of BEVs and hydrogen production in renewable energy integration." International Journal of Hydrogen Energy 46.2 (2021): 2755-2773.

Abba, S. I., et al. "Emerging Harris Hawks Optimization based load demand forecasting and optimal sizing of stand-alone hybrid renewable energy systems–A case study of Kano and Abuja, Nigeria." Results in Engineering 12 (2021): 100260.

 

Reviewer 3 Report

Accept in present form