Next Article in Journal
A Fast Deep Learning Method for Security Vulnerability Study of XOR PUFs
Next Article in Special Issue
Improving Classification Accuracy of Hand Gesture Recognition Based on 60 GHz FMCW Radar with Deep Learning Domain Adaptation
Previous Article in Journal
A Full-Duplex LED-to-LED Visible Light Communication System
Peer-Review Record

Improving Deep Learning-Based UWB LOS/NLOS Identification with Transfer Learning: An Empirical Approach

Electronics 2020, 9(10), 1714;
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Electronics 2020, 9(10), 1714;
Received: 20 September 2020 / Revised: 13 October 2020 / Accepted: 15 October 2020 / Published: 18 October 2020
(This article belongs to the Special Issue AI Applications in IoT and Mobile Wireless Networks)

Round 1

Reviewer 1 Report

Authors have definitely delivered us an important and timely work that combines the machine learning techniques with the UWB channel modeling/measurements. The NLOS/LOS identification accuracy can be therefore enhanced by the proposed scheme. This paper is generally well written, however, there are some concerns and questions:

First, in the introduction part, authors mentioned that the multipath immunity is one of the advantages UBW WLC holds, the reference [4] seems to have not support this conclusion, could you please double-check or elaborate more about it, or put a reference on it?

Moreover, the machine learning assisted NLOS/LOS detection is critical and can be widely extended to be used in a wider range of application scenarios, such as UAV comm, Maritime Comm, etc., authors may consider the following references to add to make the introduction more rich and interesting:

[1] Y. Huo, X. Dong, T. Lu, W. Xu and M. Yuen, "Distributed and Multilayer UAV Networks for Next-Generation Wireless Communication and Power Transfer: A Feasibility Study," in IEEE Internet of Things Journal, vol. 6, no. 4, pp. 7103-7115, Aug. 2019, doi: 10.1109/JIOT.2019.2914414.

[2] L. Li et al., "Millimeter-Wave Networking in Sky: A Machine Learning and Mean Field Game Approach for Joint Beamforming and Beam-Steering," in IEEE Transactions on Wireless Communications, doi: 10.1109/TWC.2020.3003284.

[3] Y. Huo, X. Dong and S. Beatty, "Cellular Communications in Ocean Waves for Maritime Internet of Things," in IEEE Internet of Things Journal, doi: 10.1109/JIOT.2020.2988634.

Finally, the EVK1000 dev. board may need some proper description and reference. The legend in Fig. 5 and Fig. 6 needs careful correction.  


Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper presents the identification scheme of the ultra-wide line of sight / non-line of sight that has been improved, which is based on a hybrid method (deep learning and transfer learning). In previous studies, the accuracy dropped sharply. To solve the problem, the authors propose an NLOS identification method for classifying these UWB conditions. A Decawave EVK1000 was used to measure the data. More details about on the innovative aspects should be highlighted.

This article is good, from the point of view of the structure and information that it gives us and who are helpful. I really like how formulas are arranged in the paper and how they are explained.

Try to look for more references, for an article of 11-12 pages, I consider that 19 references are not enough. For example, related work regarding other use cases could be added:

- He, Ying, Yan Chen, Yang Hu, and Bing Zeng. "WiFi Vision: Sensing, Recognition, and Detection with Commodity MIMO-OFDM WiFi." IEEE Internet of Things Journal (2020).

- Aileni, Raluca Maria, George Suciu, Carlos Alberto Valderrama Sukuyama, Sever Pasca, and Rajagopal Maheswar. "Internet of wearable low-power wide-area network devices for health self-monitoring." In LPWAN Technologies for IoT and M2M Applications, pp. 307-325. Academic Press, 2020.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors present a machine learning technique for the LOS/NLOS identification of unmeasured environments with improved detection rate and computational training time. This is achieved by applying the transfer learning technique by which the MPL and CNN are partially trained with new data, keeping information of previous trainings. This speeds up the processes and improves prediction.

The description of the state-of-the-art is adequate, as well as the methodology and results. I only have two minor comments about the paper:

1) Just for clarity, when the authors write "3100 data points for LOS, weak NLOS, and NLOS in each room." Does it mean 3100 for the three condition together, or for each condition at each room? Later in the same paragraph they say that it was "(3000 LOS data, 3000 Weak NLOS data, 3000 NLOS data, 100 LOS data for transfer learning, 100 Weak NLOS data for transfer learning, and 100 NLOS data for transfer learning) for both the lab and classroom environments." So my take is that it was 3100 data points for each condition in each room. I think it should be expressed slightly more clearly to avoid confusion in this point.

2) Could the authors provide some details on the implementation of the Neural Networks?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Back to TopTop