Fuzzy Adaptive PSO-ELM Algorithm Applied to Vehicle Sound Quality Prediction
Round 1
Reviewer 1 Report
The article has additional experiments needed. The subjective and objective evaluation of sound quality is not present in the paper.
Comments:
1)Line 28, What is the meaning of clearness? There is no such concept as clearness in sound quality evaluation. This is a good indication that the authors do not have a good understanding of the basic concepts of objective evaluation parameters of sound quality.
2)The references cited by the authors are not standardized and relevant references are not summarized.
3)The font in Figure 1 is not very clear.
4)The paper simply cited the subjective and objective data of sound quality from others' papers, lacking the process of subjective and objective evaluation of sound quality. Subjective and objective evaluation of sound quality is the basis for the sound quality model establishment.
Author Response
Re Comments:
- I deeply apologize for the misunderstanding. Clearness in sound refers to its purity and recognizability. A high level of clearness makes it easier for listeners to pick up specifics and components, which improves perception and communication. Although it isn't a direct indicator of sound quality, it influences comfort in general.To prevent misunderstandings in the future, adjustments have been made in text.
- Sorry, modifications have been made. Thank you for your advice.
- Figure 1 has been replaced, thank you for your advice.
- Thank you sincerely for your valuable advice. We understand that the results of both subjective and objective evaluation of sound quality significantly influence the accuracy of the prediction model. Unfortunately, due to certain constraints, we were unable to complete these tests in a timely manner. As a result, we have opted to reference data from a comprehensive and meticulously conducted experiment. This method ensures the accuracy of our prediction model and provides a basis for comparing to the results of the old model, thereby validating the advantages of the current model.
Reviewer 2 Report
I have read the manuscript "Fuzzy adaptative PSO-ELM algorithm applied to vehicle sound prediction" which is very interesting and suitable for the scope of the journal.
I have some comments regarding the section 2, which is very clear and detailed, however I suggest to develop clear definitions of the fitness function and the population used for eqs. 15,16, 17. Besides, subsection 2.3 seems to be missing.
Could you please include more details on the dataset used. Maybe a table with the description of the variables could be fine. I could not understand what variables were used as inputs. As far as I understand the variables shown in table 1 are the outputs of the model.
Do you think 25 samples are enough for the training and 5 for testing?
In section 4 you presented the table 4 with the matrixes of weights and thresholds. Are those results relevant if the weights are randomly initialized?
Did you use some early stopping technique for avoiding overfitting?
Please take into account these comments to improve your interesting manuscript.
Author Response
Thank you for your positive feedback and valuable suggestions.
We have reviewed the content of section 2 carefully and made necessary improvements.
In addition, in order to better clarify the training process of the PSO-ELM sound quality prediction model, we have enhanced the description of the population setting, dataset usage, and training steps.
For enhanced reliability of the model evaluation, a larger dataset is preferred. Since the sample size is limited, using 25 samples for training and 5 samples for testing can provide some insight.
The weight and threshold value matrices in Table 4 represent the best solutions found through the PSO optimization process, which are the results of iterative optimisation search with random initialisation data. There is a certain correlation.
To address the concern of overfitting, we adjusted the maximum number of iterations during the optimization process. Once the model's performance reaches a plateau or no longer shows improvement, we halt the optimization process early to prevent overfitting.
Finally, thank you again for your valuable comments!
Reviewer 3 Report
Usually, to prove the performance of the stochastic optimization algorithms, the mean value and standard deviation of the fitness function over several benchmarks are presented.
The algorithm needs to be evaluated using a multimodal function to ensure that it converges to global minima.
The authors claim that the "PSO-ELM model provides a reliable method for vehicle sound quality prediction", but the parameters used to evaluate the sound quality are valid for several scenarios. Please detail the characteristics of the data used in the experiments.
In general, the work is well written, but there are a few formatting (reference to multiple equations, font type, the name of some variables does not match in the equation and paragraph) and grammar issues.
Author Response
Thank you sincerely for your valuable comments. We have carefully incorporated your comments into the article, making necessary improvements. In the fourth part of the paper, we compared our fuzzy adaptive PSO-ELM algorithm with three other sound quality prediction models. We evaluated the relative mean error and goodness of fit to confirm the accuracy of our model.
Furthermore, we have explored and compared three different adaptive inertia factor control methods to guarantee the algorithm's convergence to the global minimum. The results reveal that the fuzzy adaptive PSO-ELM model achieves the same prediction accuracy with the fewest iterations, highlighting its efficiency and effectiveness.
Our fuzzy adaptive PSO-ELM model establishes a connection between the objective evaluation parameters and subjective evaluation parameters of sound quality. This enables us to accurately and efficiently predict the subjective evaluation of sound based on the objective parameters, while also saving valuable testing time and costs. We believe that this approach holds significant practical value in improving the evaluation of automotive sound quality.
Thanks again for your valuable comments.
Reviewer 4 Report
In this paper, the authors proposes a fuzzy adaptive particle swarm optimization (PSO) method based on the fuzzy adaptive PSO-ELM algorithm, to solve the sound quality prediction problem.
The problem and the proposed solutions are interesting. However, several improvements need to be applied to reach a complete research study.
In particular:
1) A section "Related works" must be inserted, in which to describe the relevant works existing in the literature.
2) A more complete discussion about the decisions and motivations by the authors, need to be provided. For example, the decision to use ELM must be motivated by a study which compare other solutions, such as traditiona and deep learning models.
3) The proposed solution must compared with other existing system which solve the same problem.
Minor revision of th english should be made by authors
Author Response
We sincerely appreciate your valuable comments. Your feedback has been taken into consideration, and we have made the necessary revisions in Part 1 and Part 4 of the article to provide a more elaborate and comprehensive description of the related studies. Moreover, to further validate the accuracy and efficiency of our fuzzy adaptive algorithm, we have conducted additional comparisons with other existing researches. This endeavor has allowed us to strengthen the practicality and reliability of our model. Once again, thank you for your valuable input.
Round 2
Reviewer 1 Report
This paper only uses data from other papers to predict and model sound quality. It lacks objective evaluation and analysis of sound quality and subjective evaluation tests, so it cannot be a complete academic paper. This paper has additional experiments needed. The subjective and objective evaluation of sound quality is not present in the paper.
Author Response
Thank you sincerely for your valuable advice. We have enhanced the article by incorporating the calculation of objective parameters and the subjective evaluation of sound quality, as mentioned in the third paragraph of the article. Once again, thank you for your insightful comments.
Reviewer 2 Report
After reading the revised version, I think the improvements better explain the experiment and methodology. However, to improve I think I could be better to state the input variables. As far as I understood, quality variables were used as the outputs. but misunderstand the inputs
Author Response
Thank you sincerely for your valuable comments. To avoid future misunderstanding, Table 2 shows that the objective parameters x for sound quality serves as the input, while the subjective parameter y acts as the corresponding output. The Fuzzy Adaptive PSO-ELM Algorithm allows for the prediction of sound quality's subjective evaluation based on its objective parameters. Thank you for your valuable input again.
Reviewer 3 Report
The authors have improved the work. I recommend this work for publication.
Author Response
Thank you for your positive feedback.
Reviewer 4 Report
I really appreciated that a part of comparisons with other systems was added. however, I don't see significant improvements on the Introduction section. It must be reorganized (possibly into sub-paragraphs) in order to highlight the scenario, the motivations, the overall solution and main contribution. Furthermore, all the part relating to the discussion of the relevant works existing in the literature must be concentrated and described more deeply in a special section "Related works".
Indeed, in the latter, the existing works must be compared with the proposed one in order to highlight the differences from a technical point of view. A table should be inserted at the end of the section highlighting these differences.
The overall quelity of the English is acceptable. However, I suggest a simplification of the sentences in some places in order to make the English more readable.
Author Response
Thank you for your helpful comments. We have made more improvements to Parts I and IV of the article based on your suggestions.
The article begins by listing the advantages of using the ELM algorithm over other existing neural network methods. Then, we explain how the Fuzzy Adaptive PSO-ELM Algorithm overcomes the limitations of the ELM algorithm to present the thesis of this paper.
In Part IV, we compare the fuzzy adaptive PSO-ELM algorithm with the other three algorithms by tabulating and plotting. This comparison highlights the higher effectiveness and accuracy of the fuzzy adaptive PSO-ELM algorithm.
Once again, thank you for your valuable contributions.
Round 3
Reviewer 1 Report
The subjective evaluation process needs to be improved, how the subjective evaluation results are analyzed, and whether cluster analysis and correlation tests are needed are not described in the paper.
Changes in the objective covariates of sound quality for each sample were not analyzed, and only a brief list of each sample was made.
Author Response
Thanks for your feedback. We have added more subjective evaluation of the data process in the third paragraph of the article. To ensure the accuracy of the test data, we evaluated the testers using identical sample pairs. Furthermore, we improved the process of calculating objective test data.
Thanks again for your help! Your feedback really made a difference.