Activity Recognition Using Different Sensor Modalities and Deep Learning

Round 1

Reviewer 1 Report

Thank you for inviting me to review this manuscript.
This article investigates using neural network for human activity system prototype. 

The paper is well written.  The training and test samples are large enough to indicate a significant scientific result. The research design is appropriate, all research stages are correct. 

I highly suggest that the authors add the following information: 

1. Emphasize the novelty of the development. It is not clear from the text of the article how your system differs from analogs

2. Explain how signal preprocessing is performed? FSR sensors are highly sensitive and there is a lot of noises and artifacts during active movements

3. How did you verify results?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Authors have presented work on "Activity Recognition using different sector modalities and Deep learning" an interesting work. I have few questions or suggestions prior to accept it.

1. Key findings of the work need to be added in abstract section. So I suggest to rewrite it.

2. Clear motivation of the work is missing. Add it in introduction part.

3. Deep enforcement learning to be performed for all the cases like foot contact states, acceleration and gyroscope.

4. What is the proposed recognition approach for HAR?

5. How the raw data processed and what is the reference validation taken in the study?

6. Compare the result with available literatures.

7. Add some recent literatures.

8. Future scope is missing add it in conclusion section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The research attempted to design a wireless smart insole with a low-cost, lightweight and compact solution. Considering the limitation of traditional and modern machine learning methods for activity recognition, they used a hybrid model, namely convolutional LSTM neural networks to realize superior performance. Their experimental results showed that the combination of all data presented the best performance compared with other categories for activity recognition. And the used network provided better results for foot contact states compared to the combination of accelerometer and gyroscope data.

But there are many shortages on the mc as follows:

1) there lack key realization ways and details, such as no introduction on LSTM, and no sample organization on the proposed method, and etc.

2) How to different Sensor Modalities are integrated toward better performance?  The interrelation between these modalities and deep learning fails to be explained.

3) The English writing should be further improved.

It is required to improve

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The researchers have presented a paper on Activity recognition using sensors and Deep Learning. Neural network is utilized to improve and enhance the quality of data collected from the sensors. There is merit to this article, but considerable changes are needed.

1. Improve the grammar and readability of the abstract. For example, the sentence "Continuously monitoring and early detection is important to prevent symptom aggravation and progressive irreversible deformities." has minor grammar issues and does not clearly convey which symptoms should be detected early to prevent irreversible deformities.

2. The fullforms of IMU and FSR are missing in line 38. Acronyms/abbreviations should be mentioned in full once in both the abstract and the rest of the manuscript.

3. Proofread the manuscript. There are lots of grammar errors.

4. The full forms for CNN, LSTM and LSTM NN are missing.

5. Add references for equations.

6. Write a few lines about the IMU sensor utilized and how it was secured to the subjects.

7. Many other studies on smart insoles and deep learning have been published. Kindly compare the advantages of your proposed design to the existing designs. Is the precision better?

8. Smart insoles are available for purchase from different manufacturers. Does your design have any advantages over them? If so, then mention the same.

9. In the discussion section, it is mentioned that the proposed design has no visible wires. However, it is not explained why this is an advantage. Kindly elaborate. 

10. In the discussion section, it is mentioned that carrying IMU sensors may be uncomfortable. Kindly elaborate. Add details about where IMU sensors are worn, their typical weight, and compare it to the weight of your soles and IMU sensor. Is the IMU sensor placed on the thighs lighter than the ones that other studies use?

Extensive editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

All the queries have been addressed properly. Now it can be accepted at its present form.