K-mer-Based Human Gesture Recognition (KHGR) Using Curved Piezoelectric Sensor
, Tsung-Han Huang 2, Wei-Song Hung 2, Ching-Yuan Chang 3, Yu-Wei Wu 4,5,6,* and Po Ting Lin 1,7,*Round 1
Reviewer 1 Report
The authors focus their study on the field of human activity recognition and they classify the human gestures using data collected from a curved piezoelectric sensor. The authors have performed a detailed machine learning based analysis in order to perform the classification and they have utilized three machine learning algorithms which are the support vector machines, the random forest, and the K nearest neighbor algorithms. The authors have provided a very detailed theoretical analysis of the proposed model, as well as a detailed set of numerical results in order to mainly show the pure operation and the performance of the proposed framework. This reviewer has no concerns regarding the theoretical analysis of the paper, however the manuscript has some major weaknesses and the authors need to address them in the revised version in order to improve the scientific depth of their manuscript. Furthermore, there are some minor comments that the authors need to consider in order to improve the quality of presentation of their manuscript. Initially, the authors need to provide an introduction to the basic machine learning algorithms that they are using, by referring to the state-of-the-art, such as Huang, Xin-Lin, Xiaomin Ma, and Fei Hu. "Machine learning and intelligent communications." Mobile Networks and Applications 23.1 (2018): 68-70, in section one or two in order to enable the average reader who is not expert in the field to follow the rest of the analysis. Furthermore, the main concern of the reviewer is that the authors have not analyzed the computational complexity of the proposed implementation and they need to include the theoretical and also the computational complexity of the proposed approach in Section 4. Based on the previous comment, the authors need to provide some indicative numerical results capturing the computational complexity of the proposed framework in a realistic implementation. Finally, the manuscript has several typos and syntax errors that the authors need to address in the revised version.
Author Response
Thank you so much for your valuable suggestions and comments. In the attached response sheet we have properly addressed all the comments raised by you, which significantly helped a lot to improve the quality of manuscript.
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper presents a K-mer strategy for Human Gesture Recognition using Curved Piezoelectric Sensor. The paper is well presented and described in detail the experiments carried out.
My comments to improve the paper:
- I miss more information about the dataset. I’d suggest including a table with the main figures: number of recordings, subjects, subjects’ characteristics, …
- I’d suggest including more details about the sensors. Can you show the signals recorded from these sensors?
- Tables with results must have the same number of decimals.
- How was the data split in folds? Did you consider a subject-wise strategy? I mean in testing the recordings should belong to different subjects compared to those that were used to train the system. This aspect is important to see the significance of the results.
- All the results must include confidence intervals to see the significancy of the results.
- In figure 3, there is not a validation subset, how did you adjust the hyperparameter of every machine learning algorithm?
- Increase font in graphs.
- Regarding the ML used, why not using deep learning strategies?
Author Response
Thank you so much for your valuable suggestions and comments. In the attached response sheet we have properly addressed all the comments raised by you, which significantly helped a lot to improve the quality of manuscript.
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper has a method that is very straightforward without any novelty in it. it is no more than a class project.
No clear research gap. No comparison with previous methods, and more
Author Response
Thank you so much for your valuable suggestions and comments. In the attached response sheet we have properly addressed all the comments raised by you, which significantly helped a lot to improve the quality of manuscript.
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have addressed the reviewers comments.
Author Response
Thank you so much for your valuable suggestions and comments. We are greatful for review, which helped a lot to improve this manuscript.
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The authors have addressed all my comments. I'd suggest including a future work section in the conclusions.
Author Response
Thank you so much for your valuable suggestions and comments. We are greatful to you for your raised comments, It helped tp improve the manuscript and we have properly addressed all the comments raised by you in the revised manuscript.
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The issue with paper is the novelty. Good results with slight improvement from previous work s are not enough to be published. A suitable method that can be used in different domains is a novel. I would suggest improving the method, applying it to different datasets, test its robustness of it in different scenarios.
Author Response
Thank you so much for your valuable suggestions and comments. In the attached response sheet we have properly addressed all the comments raised by you, which significantly helped a lot to improve the quality of manuscript.
Please see the attachment.
Author Response File: Author Response.pdf
Round 3
Reviewer 3 Report
It is now okay to be considered for publication. In future, I suggest thinking about fusion techniques to increase the robustness.
