Clutter Elimination and Harmonic Suppression of Non-Stationary Life Signs for Long-Range and Through-Wall Human Subject Detection Using Spectral Kurtosis Analysis (SKA)-Based Windowed Fourier Transform (WFT) Method
Round 1
Reviewer 1 Report
This paper proposes a life sign detection system using impulse ultra-wideband (UWB) radar. Many techniques are applied to eliminate clutters, improve SNR as well as respiration signals. To estimate range of human subject and respiration signal frequency, the spectral kurtosis analysis based windowed Fourier transform method and an accumulator are employed in the proposed framework. Experimental results demonstrated good performance compared with well-known algorithms. This is a great work and the paper may attract a lot of interest from the community and should be considered for publication. The reviewer only has a couple of comments that the authors should clarify in the next revision.
1. In Section 4.1 (Experimental Setup), the authors describe six experiments carried out in the paper. It would nice to add something to link these experiments to the following sections, i.e sections 4.2 to 4.7. I found it difficult to connect the results presented in these sections to section 4.1, for example, what section for what experiment.
2. In equation (6), the authors show q[m,n] as non-static clutter, g[m,n] as clutter caused by system electronics, and z[m,n] as clutter caused by moving objects other than the target. However, the reviewer is not clear how these clutters are eliminated in the proposed framework. Please discuss more details on these.
3. The paper has “life signs” in the title but it only focused on respiration detection. How about heartbeat detection? Please elaborate more on this. It would be nice to add some discussions on the possibility of extending the proposed work to heartbeat detection, perhaps, what needs to be modified to enable heartbeat detection or both.
4. In the Conclusions, it is said that the future work will consider the detection of multiple human subjects. It would be nicer to add some discussions on why it is challenging and what should be done to target it either in the Conclusions or in previous sections.
I am confused about the sixth experiment where it has six human subjects. Please explain more about the purpose of this experiment.
5. It is not clear to me how the proposed approach deals with through-wall effect. Please elaborate on how the through-wall effects are accounted for to compensate the range displacement.
Author Response
We would like to thank you for the time spent in evaluating our submission. The comments have helped tremendously in improving this paper. Below you will find our point-by-point responses to all the comments/questions.
This paper proposes a life sign detection system using impulse ultra-wideband (UWB) radar. Many techniques are applied to eliminate clutters, improve SNR as well as respiration signals. To estimate range of human subject and respiration signal frequency, the spectral kurtosis analysis based windowed Fourier transform method and an accumulator are employed in the proposed framework. Experimental results demonstrated good performance compared with well-known algorithms. This is a great work and the paper may attract a lot of interest from the community and should be considered for publication. The reviewer only has a couple of comments that the authors should clarify in the next revision.
1. In Section 4.1 (Experimental Setup), the authors describe six experiments carried out in the paper. It would nice to add something to link these experiments to the following sections, i.e sections 4.2 to 4.7. I found it difficult to connect the results presented in these sections to section 4.1, for example, what section for what experiment.
All these experiments have been amended to make them easy to read and understand.
2. In equation (6), the authors show q[m,n] as non-static clutter, g[m,n] as clutter caused by system electronics, and z[m,n] as clutter caused by moving objects other than the target. However, the reviewer is not clear how these clutters are eliminated in the proposed framework. Please discuss more details on these.
In this paper, LTS method and a smoothing filter are employed to remove the non-static clutter. The automatic gain algorithm is used to suppress the clutter caused by system electronics. To suppress the clutter caused by moving objects other than the target, there is only detection target in the experiments. Meanwhile, a band-pass filter and a smoothing filter are used to suppress the clutter caused by some unknown moving targets.
3. The paper has “life signs” in the title but it only focused on respiration detection. How about heartbeat detection? Please elaborate more on this. It would be nice to add some discussions on the possibility of extending the proposed work to heartbeat detection, perhaps, what needs to be modified to enable heartbeat detection or both.
This paper focuses on human subject detection in through-wall and long range conditions. Thus human respiratory is considered as factor to detect human subject. The amplitude of heartbeat is too smaller to obtain in this situation using the developed UWB radar. But, in this paper, the actuator is used to imitate human heartbeat. The actuator signal has an amplitude of 3 mm and a frequency of 0.333 Hz. Now, we are updating the UWB radar to acquire human heartbeat.
4. In the Conclusions, it is said that the future work will consider the detection of multiple human subjects. It would be nicer to add some discussions on why it is challenging and what should be done to target it either in the Conclusions or in previous sections. I am confused about the sixth experiment where it has six human subjects. Please explain more about the purpose of this experiment.
The conclusions have been updated.
You maybe misunderstand the sixth experiment.
In this experiment, one human subject served as detection target and three datasets are acquired for each given range including 3 m, 6 m, 9 m, and 11 m. Thus all 12 datasets are obtained for the first volunteer. Then, the second volunteer served as the detection target in the conducted experiments, thus another 12 datasets are acquired. All six different volunteers served as the detection targets in this experiment cyclically.
5. It is not clear to me how the proposed approach deals with through-wall effect. Please elaborate on how the through-wall effects are accounted for to compensate the range displacement.
The major factors caused by the wall affecting human subject detection are the static clutters with larger amplitudes and larger signal attenuation. To deal with the problems, TMS algorithm is used to suppress the static clutters. The automatic gain algorithm and a smoothing filter are used to enhance the amplitude of life signs.
Reviewer 2 Report
The paper presents a method for detection of thought-wall human detection using a UWB radar. The main innovation is the application of spectral kurtois analysis and a frequency accumulator to improve the signal-to-noise and clutter ratio.
The paper is supported by several experiments and the signal processing is well described in detail.
However, I have some comments to improve the quality of the paper:
1. In the legend of fig.12 should indicate the measured range (6m) and situation.
2. The reference [37] should be cited just before the equation 31. The parameters used in eqn.31 (Crammer-Rao bound) such as mu, A and the number of pulsed must be given in the text.
3. The section 4.7 should be explained better. For example, in the histograms of Fig.24 the units of “Data” must be included. I suppose that is in cm. Table 5 must be explained better. How is estimated the distance if there is not a human target? In table 5 there are more rows than sets without human (must be 10 rows?).
4. The paper studies the influence of the angle. The signal amplitude decreases when the person is back to the radar due to the small displacement of the chest for this case. Is it possible to obtain the breathing or presence of a person when it is in back orientation (for example at a lower distance)?
Author Response
We would like to thank you for the time spent in evaluating our submission. The comments have helped tremendously in improving this paper. Below you will find our point-by-point responses to all the comments/questions.
The paper presents a method for detection of thought-wall human detection using a UWB radar. The main innovation is the application of spectral kurtois analysis and a frequency accumulator to improve the signal-to-noise and clutter ratio. The paper is supported by several experiments and the signal processing is well described in detail.
However, I have some comments to improve the quality of the paper:
1. In the legend of fig.12 should indicate the measured range (6m) and situation.
The legend of fig.12 has been updated.
2. The reference [37] should be cited just before the equation 31. The parameters used in eqn.31 (Crammer-Rao bound) such as mu, A and the number of pulsed must be given in the text.
The reference [37] has been cited before the equation 31, and the parameters used in equation (31) have been given.
3. The section 4.7 should be explained better. For example, in the histograms of Fig. 24, the units of “Data” must be included. I suppose that is in cm. Table 5 must be explained better. How is estimated the distance if there is not a human target? In table 5 there are more rows than sets without human (must be 10 rows?).
The units of “Data” have been given in cm. Table 5 has been amended. Using (23) and (28), the range distances can be acquired whether a human subject is present or not in the detection environment.
4. The paper studies the influence of the angle. The signal amplitude decreases when the person is back to the radar due to the small displacement of the chest for this case. Is it possible to obtain the breathing or presence of a person when it is in back orientation (for example at a lower distance)?
Yes, it is possible to obtain the presence of a person when it is in back orientation.
Round 2
Reviewer 1 Report
I appreciate that the revised manuscript as well as the responses address most of my concerns. However, there are some things that I am still not clear.
1. My second comment for the first version of the manuscript was answered in the authors’ responses. But please also add the answer to the revised manuscript too.
2. My fourth comment was about multiple human detection. I can see the authors has updated the Conclusions where “multiple subjects” is removed. However, I am curious about how the proposed system work in the scenario of multiple humans simultaneously existing in the target space. Please address it.
3. Regarding my fifth comment about the range displacement, even though the TMS algorithm can be exploited to suppress the static clutter but the phases of scattered fields due to the human subject itself are also affected which can result in detecting the target’s range inaccurately. Please address this.
Author Response
1. My second comment for the first version of the manuscript was answered in the authors’ responses. But please also add the answer to the revised manuscript too.
These answers have been added in the paper.
2. My fourth comment was about multiple human detection. I can see the authors has updated the Conclusions where “multiple subjects” is removed. However, I am curious about how the proposed system work in the scenario of multiple humans simultaneously existing in the target space. Please address it.
The conclusions are updated in this paper due to the proposed system can better detect multiple humans simultaneously via analyzing the characteristics of life signs, which have been analyzed in one of our works as in [1]. Thus, future work will focus on obtaining respiration signals when the subject is in motion such as walking and running due to it is more challenging to acquire human respiration frequency when the subjects are moving or walking due to they are in the same band as human breath with larger amplitudes.
[1] Experimental Study of Wireless Monitoring of Human Respiratory Movements Using UWB Impulse Radar Systems, Sensors 2018, 18(9), 3065.
3. Regarding my fifth comment about the range displacement, even though the TMS algorithm can be exploited to suppress the static clutter but the phases of scattered fields due to the human subject itself are also affected which can result in detecting the target’s range inaccurately. Please address this.
Using the UWB radar developed in this paper, only the amplitudes of the pulses modulated by life signs are acquired. The phases are not considered in the employed UWB radar system. Thus, the range can be estimated accurately so long as the clutters are suppressed effectively.
