Pedestrian Walking Distance Estimation Based on Smartphone Mode Recognition
Round 1
Reviewer 1 Report
Paper presents a walking-distance estimation method consists of a smartphone mode recognition and stride-length estimation based on the regression model using the inertial sensors of the smartphone. The advantages of the method are confirmed by simulation and the results of the experiment.
Unfortunately, the text requires significant corrections. The following comments are required.
1. The numbers of tables and figures are confused. In particular, there are two figures numbered 1 and two tables numbered 1.
2. Line 131 in the table 1, You must specify the units.
3. Probably some of the text is lost, starting with line 164. It is desirable to give a scheme here.
4. Line 199. What does BLE AIR in Fig1?.
5. Lines 199-200. What is the difference between Gyroccope sensor and Gyroccope?
6. Line 224. You must specify the time interval between measurements.
7. It is recommended to explain why the set (4) - (6) are used, for example, why it is not used. standard deviation for acceleration.
8. The phrase: "It is the difference between 75th percentile and 25th percentile of the signal" is not clear
9. Line 259 - FFT, DC-?
10. Lines 288-289 - it is necessary to write the title of the ordinate axis.
11. Lines 294-295; 302; 334- it is advisable to point the references to the literature
12. Figure 12 needs clarification.
13. It is necessary to describe the axes on the figure 14а
Author Response
Cover Letter
Dear Editors and Reviewers:
Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Pedestrian Walking Distance Estimation based on Smartphone Mode Recognition” (Manuscript ID remotesensing-497569). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made a major revision in the resubmitted manuscript which we hope to meet with approval.
The main corrections in the paper and the responds to the reviewer’s comments are as flowing:
Revision Description and Responds to the reviewer’s comments:
Paper presents a walking-distance estimation method consists of a smartphone mode recognition and stride-length estimation based on the regression model using the inertial sensors of the smartphone. The advantages of the method are confirmed by simulation and the results of the experiment.
Unfortunately, the text requires significant corrections. The following comments are required.:
Response: Thank you for your work.
The numbers of tables and figures are confused. In particular, there are two figures numbered 1 and two tables numbered 1.
Response: We have checked that the numbers of tables and figures are right in our submitted version. We also find that the numbers of tables and figures are confused in the version downloaded from the MDPI review system. In addition, the downloaded version has a slight change in Typesetting formats. Therefore, we guess that the review system has led to numbering confusion. We update the numbers of tables and figures again in the revised version.
Line 131 in the table 1, You must specify the units.
Response: In this revision, we have specified the units.
Probably some of the text is lost, starting with line 164. It is desirable to give a scheme here.
Response: yes, “Figure 1 depicts the architecture”. The number of figures is missing, this mistake maybe caused by the MPDI system. We update it in the revised version.
Line 199. What does BLE AIR in Fig1?
Response: The foot-mounted INS module communicates with smartphone via BLE AIR (Bluetooth Low Energy air) interface. We have added the definition of the abbreviation in the revised version.
Lines 199-200. What is the difference between Gyroscope sensor and Gyroscope?
Response: Gyroscope sensor and gyroscope are the same. We have unified it in the revised version.
Line 224. You must specify the time interval between measurements.
Response: The time interval between measurements is 0.01s (100 Hz) (the sample rate is specified in benchmark dataset section, line 187).
It is recommended to explain why the set (4) - (6) are used, for example, why it is not used. standard deviation for acceleration.
Response: The average magnitude of acceleration in a sliding window is used to determine whether smartphone is moving. The standard deviation of the gyroscope and the standard deviation of magnetic field magnitude are used to reduce the influence from irregular motion (say, false detection due to shaking or rotating smartphones) and improve the walk detection accuracy. In addition, we have redrawn Fig 7 (in old version, it is Fig 6) to improve readability in the revised version.
The phrase: "It is the difference between 75th percentile and 25th percentile of the signal" is not clear.
Response: In this revision, we have detailed the interquartile range.
Line 259 - FFT, DC-?
Response: FFT DC represents the direct component of fast Fourier Transform. We have refined the description in the revised version.
Lines 288-289 - it is necessary to write the title of the ordinate axis.
Response: In this revision, we have replotted Fig 9 and Fig 10 (in old version, it is Fig 8 and Fig 9) to improve readability.
Lines 294-295; 302; 334- it is advisable to point the references to the literature
Response: We have added the corresponding references in the revised version.
Figure 12 needs clarification.
Response: In this revision, we have detailed the figure 13 (in old version, it is Fig 12) in Line 341-348.
It is necessary to describe the axes on the figure 14а.
Response: We have described the figure 15 (in old version, it is Fig 14a) in Line 392-401.
We appreciate for Editors/Reviewers’ warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Authors presented an interesting paper which combines smartphone modes recognition and stride-length estimation to provide an accurate walking-distance estimation. The main idea was to implement a stride-length estimation model for each smartphone mode, recognize the current mode, and then accurately estimate walking distance.
Although this paper shows promising results, the main drawback is somewhat limited testing of the proposed method in realistic scenarios. Tests itself should be more extensive and cover more everyday scenarios, where subjects follow different and non-predictable routes.
More details follow:
There is an error in referencing tables (the error is shown instead of table name)
In line, 189 authors stated, “The dataset is obtained from a group of healthy adults with natural motion patterns”. Please provide more information about the subjects: age, sex, height, etc.
Also, how do you define slow, normal and fast walking?
In line 194, Stride length range is [0,2]. Does this mean that strides longer than 2 m are omitted?
Could website link provide in line 199 be placed in the form of a footnote, or reference?
How do you explain bimodality of histograms of real stride length, fig 3?
Line 209, is 10th order Butterworth filter overkill for its intended application?
Also, the cut-off frequency of 3Hz for 3Hz signal does not seem right.
There is one old paper which dealing with this problem written by D.A. Winter, https://0-doi-org.brum.beds.ac.uk/10.1016/0021-9290(74)90056-6
Basically, cut-off frequency should not be less than 6Hz. It would be a good idea to provide frequency spectra for signals (raw acc and 7 or data) together in fig 5.
Line 218-219 is confusing: “may lead to unreliable detection results when the user is abnormal movements such as unexpectedly rotating or…”
Fig 9, although authors marked rectangles with different colors, they could be (and are) plotted with different line types, for better visibility in grayscale print. Also, legend or extra-label may improve readability.
In fig 14, main diagonal may include precision data (in terms of %), which would increase readability.
In section 3.5. Experimental Results of Walking-Distance, authors in the experiment used a single route. I would like to see results if different routes are used, especially if there is a route with significant inclination, where people usually compensate inclination with shorter stride lengths.
Details in fig 17 are hard to see, (route and labels), please redo it.
In conclusion, this is an interesting paper with promising results and realistic implementation.
The main drawback is limited testing set, and possibly loss of important data with wrongly selected filter cut-off frequency.
Author Response
Cover Letter
Dear Editors and Reviewers:
Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Pedestrian Walking Distance Estimation based on Smartphone Mode Recognition” (Manuscript ID remotesensing-497569). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made a major revision in the resubmitted manuscript which we hope to meet with approval.
The main corrections in the paper and the responds to the reviewer’s comments are as flowing:
Revision Description and Responds to the reviewer’s comments:
Authors presented an interesting paper which combines smartphone modes recognition and stride-length estimation to provide an accurate walking-distance estimation. The main idea was to implement a stride-length estimation model for each smartphone mode, recognize the current mode, and then accurately estimate walking distance.
Although this paper shows promising results, the main drawback is somewhat limited testing of the proposed method in realistic scenarios. Tests itself should be more extensive and cover more everyday scenarios, where subjects follow different and non-predictable routes.
Response: Thank you for your work. In this revision, we have supplemented experiments in an outdoor stadium and a road with significant inclination in line 461-471. In stadium, subject follow non-predictable routes.
There is an error in referencing tables (the error is shown instead of table name).
Response: We have checked that the numbers of tables and figures are right in our submitted version. We also found that the numbers of tables and figures are confused in the version downloaded from the MDPI review system. In addition, the downloaded version has a slight change in Typesetting formats. Therefore, we guess that the review system caused the numbering confusion. We update the numbers of tables and figures again in the revised version.
In line, 189 authors stated, “The dataset is obtained from a group of healthy adults with natural motion patterns”. Please provide more information about the subjects: age, sex, height, etc.
Response: In this revision, we detail the information about the subjects in table 3.
Also, how do you define slow, normal and fast walking?
Response: There is one old paper which defining the walking speed written by Pham Van, https://0-doi-org.brum.beds.ac.uk/10.3390/s18103186. This paper emphasizes that users walk freely and have no limit on speed.
In line 194, Stride length range is [0,2]. Does this mean that strides longer than 2 m are omitted?
Response: In the benchmark dataset, 99.5% of strides is within 1.55m, and all strides are within 1.75. Therefore, this paper concludes that the stride-length of pedestrian is [0,2]. We set the regression prediction range of pedestrian stride-length between 0 and 2 meters. Special cases such as walking backward, lateral walking, running, and jumping will be studied in the future.
Could website link provide in line 199 be placed in the form of a footnote, or reference?
Response: we have placed the website link in the form of a footnote in the revised version.
How do you explain bimodality of histograms of real stride length, fig 3?
Response: In the revised version, fig 3 is updated to fig4. The higher peak is the stride distribution of free walking. Our dataset contains stair scenario. The stair restrains the stride-length of pedestrian about 0.6m, thus resulting in a secondary peak.
Line 209, is 10th order Butterworth filter overkill for its intended application?
Also, the cut-off frequency of 3Hz for 3Hz signal does not seem right.
There is one old paper which dealing with this problem written by D.A. Winter, https://0-doi-org.brum.beds.ac.uk/10.1016/0021-9290(74)90056-6
Basically, cut-off frequency should not be less than 6Hz. It would be a good idea to provide frequency spectra for signals (raw acc and 7 or data) together in fig 5.
Response: I apologize for my typos: 1th order->10th order. Normally, the step frequency is lower than 3 Hz (3 steps per second). Moreover, we only use a FM-INS model to measure the stride (a stride=two steps) length. Therefore, we set the cut-off frequency =3Hz.
Line 218-219 is confusing: “may lead to unreliable detection results when the user is abnormal movements such as unexpectedly rotating or…”.
Response: In this revision, we have refined the sentence in line 220-221.
Fig 9, although authors marked rectangles with different colors, they could be (and are) plotted with different line types, for better visibility in grayscale print. Also, legend or extra-label may improve readability.
Response: In this revision, we have replotted Fig 9 and Fig 10 (in old version, it is Fig 8 and Fig 9) to improve readability.
In fig 14, main diagonal may include precision data (in terms of %), which would increase readability.
Response: We have added the precision data (in terms of %) to the fig 15 (in old version, it is Fig 14a).
In section 3.5. Experimental Results of Walking-Distance, authors in the experiment used a single route. I would like to see results if different routes are used, especially if there is a route with significant inclination, where people usually compensate inclination with shorter stride lengths.
Response: In this revision, we have supplemented experiments in an outdoor stadium and a road with significant inclination in line 461-471.
Details in fig 17 are hard to see, (route and labels), please redo it.
Response: In this revision, we have redrawn Fig 18 (in old version, it is Fig 17) to improve readability.
We appreciate for Editors/Reviewers’ warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.pdf
Reviewer 3 Report
WELL DONE!!!!
Dear Collegues, my personal compliments for your work whose set up is absolutely good and the paper is absolutely balanced in each section (background, description of methodology, test section), never boring and extremely clear, therefore, in my opinion, it can be accepted in this form apart two little errors at lines 128 and 182.
Author Response
Cover Letter
Dear Editors and Reviewers:
Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Pedestrian Walking Distance Estimation based on Smartphone Mode Recognition” (Manuscript ID remotesensing-497569). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made a major revision in the resubmitted manuscript which we hope to meet with approval.
The main corrections in the paper and the responds to the reviewer’s comments are as flowing:
Revision Description and Responds to the reviewer’s comments:
Dear Collogues, my personal compliments for your work whose set up is absolutely good and the paper is absolutely balanced in each section (background, description of methodology, test section), never boring and extremely clear, therefore, in my opinion, it can be accepted in this form apart two little errors at lines 128 and 182.
The votes on the voices (Originality, Significance of content, etc.) are only average because effectively the idea is not fully new, the content is not particularly important, the soundness is high but more technical than scientific so as the interest of the Readers could be not particularly high anyway nothing detracts from the overall quality of the paper.
Response: Thank you for your work. All cross-referencing errors have been corrected in the revised version.
We appreciate for Editors/Reviewers’ warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I am satisfied with the answers. Thanks
Reviewer 2 Report
Authors have responded to all of reviewers’ comments and concerns.
