Review Reports - Low-Memory Indoor Positioning System for Standalone Embedded Hardware

Round 1

Reviewer 1 Report

This paper proposes a lightweight embedded hardware and low-memory schemes based on the characteristics of geomagnetic sensor measurement and convergence of estimated positions of a target in order to perform indoor positioning using algorithms running in the embedded hardware.

The work is technically sound and follows developments of earlier work by the authors such as in reference [22].

However, the paper could benefit from a review in its presentation. The introduction is too long. I suggest to the authors should summarize the last three paragraphs in the introduction into a single one. Some of the information provided in those paragraphs fit better in other sections of the text, including the conclusion.

The authors use the expression “In this paper” five times. The context in which this expression has been used cause some confusion to the reader, leaving the impression that the authors could not summarize the contributions in one place.

I also suggest that the authors define some acronyms used for the first time in the introduction, e.g., LBS.

Equations in Section 3.5 (page 8) are not so well presented. I suggest reviewing that section for clarity as well.

Author Response

Thank you for the careful review. Revisions to the points mentioned are listed below. Also, we revised the paper by using the "Track Changes" function in Microsoft Word, so that changes are easily visible.

The introduction is too long. ~

- We have summarized the last three paragraphs in the introduction into a single one. Also, we move some information in the introduction such as detailed experiment environment and experimental results to other sections such as environment configurations section and results section.

The authors use the expression “In this paper” five times. ~

- We have changed some of the term “In this paper” to different expressions in introduction section and experiment section (L98,382,400)

I also suggest that the authors define some acronyms used for the first time in the introduction, e.g., LBS.

- We have added definitions of acronyms such as LBS, IPS, BLE, RF and IR at the introduction section, and we have revised some expressions that the acronyms are not applied.

Equations in Section 3.5 (page 8) are not so well presented.~

- We re-express the equations in section 3.5 for clarity. Also, we add detailed explanation of the equations in Section 3.5.

Reviewer 2 Report

This paper presents a low-memory geomagnetism-based indoor positioning system for embedded hardware. To reduce the memory consumption in embedded systems, the authors propose a compression scheme for magnetic maps. To evaluate the performance of the proposed system, the authors have conducted extensive experiments using two different devices. Experimental results show that the proposed scheme can reduce the memory consumption by a large margin.

However, the reviewer has the following concerns:

1) In Figure 3, the authors estimate the location based on the displacement from previous positions inferred from PDR. What information did you use in the PDR module? Do you estimate the stride length of users? If this is the case, how do you address the diversity of users?

2) Geomagnetic readings could be stable indoors due to a lack of magnetic disturbances. Although the proposed method can reduce the memory consumption by quantization, it could lead to degraded distinctiveness of magnetic readings, leading to increased localization errors. How do you address this?

3) The authors have conducted experiments in two different trial sites. However, the numerical evaluation is limited. The authors should conduct more experiments to evaluate compare the localization quantitatively.

Author Response

1) What information did you use in the PDR module? Do you estimate the stride length of users? If this is the case, how do you address the diversity of users?

- We use the step count, the step length and the orientation of the movement in the PDR module. And we use the step length estimation algorithm mentioned in the article “Using the ADXL202 in pedometer and personal navigation applications”, which can address the diversity of users by using the minimum and maximum acceleration measured in the Z axis in a single stride. We added the explanation and the reference of the step length estimation algorithm. (L265-266)

2) Although the proposed method can reduce the memory consumption by quantization, it could lead to degraded distinctiveness of magnetic readings, leading to increased localization errors

- In the experimental results of section 4.5, the smartphone-based experiment was conducted without memory compression and map load strategy. As shown in the section, there is less difference in the localization errors with and without magnetic field map compression and map load strategy. For clarity, we have added experimental setup with the smartphone in Section 4.5.

3) The numerical evaluation is limited. The authors should conduct more experiments to evaluate compare the localization quantitatively.

- We increase the number of experiments and in order to reflect the increased experiments, we add tables of the localization errors in Section 4.5.

Reviewer 3 Report

Interesting topic and this article is on its way to become a valuable contribution!

L82-87: I don't buy in this argumentation (in my perspective these scenarios are espescially the once, where the established wireless and visual tracker are most suitable and outperform your approach in any dimension. -> Remove or reframe!

Please add the evaluation of the location accuracy and the corresponding statistical evaluation for both, magnetic and BLE for the location-specific measurements. In my opinion, the article is only meaningful when combining a thorough eval. of both, memory usage and location accuracy.

Than, compare as well with reported accuracies of similar evaluation-designs. I'm forseeing the following two articles, to be a good fit in this regard (sorry for suggesting selfauthored articles, but I think they fit well regarding study environment/measurement setup and intended research question - there might be as well various other examples you might want to consider):

https://www.researchgate.net/profile/Sebastian-Fudickar/publication/273764185_Most_Accurate_Algorithms_for_RSS-based_Wi-Fi_Indoor_Localisation/links/550b2c020cf285564096ffc4/Most-Accurate-Algorithms-for-RSS-based-Wi-Fi-Indoor-Localisation.pdf

https://www.researchgate.net/profile/Sebastian_Fudickar/publication/273764241_Comparing_Suitability_of_Sub_1_GHz_and_Wi-Fi_Transceivers_for_RSS-based_Indoor_Localisation/links/550b2c970cf285564096fff5.pdf

However, relying on Hightower's accuracy results might be misleading due to his evaluation setup...

Author Response

L82-87: I don't buy in this argumentation in my perspective ~

- We have revised the paper accordingly. (L81-82)

Please add the evaluation of the location accuracy and the corresponding statistical evaluation ~

- For comparative evaluation, we add a Wi-Fi-based indoor positioning experiment and statistical evaluation. We also add tables of the localization errors for the indoor localization algorithms in Section 4.5.

Then, compare as well with reported accuracies of similar evaluation-designs. I'm forseeing the following two articles~

- Thank you for the recommendation. We have added two recommended papers in related works and references.

Round 2

Reviewer 2 Report

The reviewer is happy that authors have conducted more experiment to evaluate the performance of the proposed algorithm. It seems that the proposed approach can achieve sufficient accuracy. All the concerns of the reviewer is addressed.