Underwater Noise Modeling and Its Application in Noise Classification with Small-Sized Samples

Round 1

Reviewer 1 Report

‘machine learning has garnered intense attention from researchers worldwide. New techniques integrated with ML or deep neural networks (NN)’ – acronyms must be added next to the related phrase: machine learning (ML). Research questions and hypotheses must be constructed based on more specific supporting sources, preferably as recent as possible. The methodology is unclear. More development and depth of the methodology and analysis are needed. The manuscript will benefit from further discussion of key concepts and methodological criteria in order to offer a better articulation between theory and data. Some cited sources are not developed. Some portions of the text are poorly edited. E.g., ‘z. On the other hand，Since sample’. Figures should be improved, unified as style, and thoroughly explained. The discussions require more structure and there is a need of offering a clear assessment of reviewed literature. Several statements made in the paper are not supported by adequate empirical evidence or by making reference to relevant literature. You should compare your results with others in terms of concrete data for better research integrative value. The Conclusions section is way too short. There is some discussion of the limitations of the study however these are not considered in terms of the implications on the study findings. The reference list includes several not-peer reviewed sources.
The relationship between deep learning-based object detection technologies and geospatial big data management algorithms as regards machine learning-based underwater noise classification has not been covered, and thus such sources can be cited:
Lăzăroiu, G., Andronie, M., Iatagan, M., Geamănu, M., Ștefănescu, R., and Dijmărescu, I. (2022). “Deep Learning-Assisted Smart Process Planning, Robotic Wireless Sensor Networks, and Geospatial Big Data Management Algorithms in the Internet of Manufacturing Things,” ISPRS International Journal of Geo-Information 11(5): 277. doi: 10.3390/ijgi11050277.
Blake, R., and Frajtova Michalikova, K. (2021). “Deep Learning-based Sensing Technologies, Artificial Intelligence-based Decision-Making Algorithms, and Big Geospatial Data Analytics in Cognitive Internet of Things,” Analysis and Metaphysics 20: 159–173. doi: 10.22381/am20202111.
Andronie, M., Lăzăroiu, G., Iatagan, M., Hurloiu, I., Ștefănescu, R., Dijmărescu, A., and Dijmărescu, I. (2023). “Big Data Management Algorithms, Deep Learning-Based Object Detection Technologies, and Geospatial Simulation and Sensor Fusion Tools in the Internet of Robotic Things,” ISPRS International Journal of Geo-Information 12(2): 35. doi: 10.3390/ijgi12020035.

Some portions of the text are poorly edited. E.g., ‘z. On the other hand，Since sample’.

Author Response

We have made modifications according to your points. Please refer to the file.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript addresses the problem of small sample sizes associated with intelligent underwater noise (UN) classification. The authors proposed a generative model for data augmentation in addition to a CNN model to classify nine types of typical UN. The work is interesting, however, I have the following minor comments:

1- In Figure 2, the authors show a Flow chart of underwater noise generation. However,  where the sum of the product happens in this flowchart, it is not clear

2- In line 154, what calculations do the authors mean? are those your own or from a specific reference? elaborate

3- In line 205, should the upper-limit frequency, and the lower-limit frequency be swapped with respect to the symbols at the previous line?

4- In equation 10, since you are using "I" here as an index,  the term "I" bar should contain this index too

Author Response

We have made modifications according to your points. Please refer to the file.

Author Response File: Author Response.doc

Reviewer 3 Report

 

 The article under consideration aims to provide a machine-learning (CNN) solution for the problem of underwater noise classification. However, the following points are required to be addressed:

1.      In the abstract, the motivation for the research problem (underwater noise modeling and its application in noise classification) must be mentioned.

2.      In the abstract, the limitations of existing practices (research gap) to address the target research problem must be mentioned.

3.      Abstract must clearly state the salient feature of the proposed method that will help to address the identified research gap.

4.      Introduction is required to be rewritten as the information about some critical aspects has not been provided. The following information must be provided in distinct paragraphs:

4.1.   Necessary background and motivation for the research problem (underwater noise modeling for classification with small-sized samples)

4.2.   Research gap (Limitations of the state-of-the-art for addressing the research problem)

4.3.   Major steps of the proposed Underwater noise modeling method for classification with small-sized samples

4.4.   How the proposed method will address the research gap?

4.5.   How the proposed method has been validated and What are the achieved results?

4.6.   Summary of contributions (based on results)

4.7.   Organization of the entire article.

5.      It’s strange to find “Experimental Data and Labeling” just after the introduction. Authors must provide a background section first. The purpose is to introduce all fundamental concepts and terminologies that are required to understand this article. The motivation behind the presentation of each concept should be explicitly stated.

6.      Background section must be followed by a related work section.

7.      Related work section should provide a comparison table to compare existing methods in terms of various attributes.

8.      Section 4 should provide the methodology section. This section should provide a holistic view of the entire method. Currently, various activities in the work have been presented in a haphazard way without any coherency. Therefore, a systematic presentation of related concepts (labeling, modeling classification) in a unified way is critical.

9.      Authors are suggested to follow a top-down approach. The design should be clearly explained in terms of structure and behavior.

10.  Design should not be mixed with the implementation details. Once the design is finished, the implementation can be described.

11.  One of the major limitations is the lack of performance comparison (with state of the art methods).

 

 

 

Author Response

We have made modifications according to your points. Please refer to the file.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

This revised version can be published.

Author Response

Thanks very much, and we appreciate any assistance that you provide to us.

Author Response File: Author Response.doc

Reviewer 3 Report

This revised version has addressed some concerns. However, some issues are required to be addressed:

 

Introduction: 

1.      Authors provide contributions on Lines 71 of page. However, they must first provide the major steps in the proposed work. It is also important to mention that how the proposed work will address the limitations of existing methods (research gap).  

2.      The description of the major steps of the proposed work must be followed by the validation information. It implies how the proposed method has been tested (verified). It includes a brief description of case studies and/or benchmarks. Why these benchmarks/case studies are important and interesting? What is the motivation behind the selection of these benchmarks?  Therefore, authors are suggested to provide appropriate validation details.

3.      Finally, the summary of the achieved results (quantitative and/or qualitative) must be provided at the end of the Introduction section. It will allow us to judge the significance of the proposed method. The significance of the achieved results can be described quantitatively as compared to state-of-the-art methods. 

Related Work:

1.      This Section must be closed with a concluding paragraph, summarizing the limitations of existing visual tracking methods. Here, the authors should also state how their solution will address the shortcomings.

 

Methodology:

1.      The methodology section on page 4 must provide a holistic overview of the entire method at the beginning of the Section. The holistic overview must be supported with an end-to-end diagram. In other words, the structure of the entire framework must be presented in a clear and detailed way.

2.      The methodology section should be restricted to design detail only. Authors are suggested to first describe the structural representation of the framework. Then a behavioral description should be provided to understand the overall mechanism.

3.      To summarize, the details in this section are required to be organized in a logical way. In its current form, it is very hard to judge and understand the technical worth of this proposal.

 Results and discussion:

1.      Authors are required to compare the results with state-of-art solutions. Good data visualization techniques must be used to illustrate the findings. Various performance parameters should be used for comparisons. However, it is very important to highlight the strengths of your article as compared to existing methods. Discuss the reasons why you were able to obtain good results. Similarly, it is equally important to mention the shortcomings of your solution with respect to current methods. Discuss the reasons for these limitations and shortcomings.

 

 

Author Response

Thanks very much. Please refer to the document.

Author Response File: Author Response.pdf

Round 3

Reviewer 3 Report

All the comments have been addressed carefully. 

The article can be published in its current form.