Block Sparse Bayesian Learning Based Joint User Activity Detection and Channel Estimation in Grant-Free MIMO-NOMA

Round 1

Reviewer 1 Report

 

In this paper, the sparsity of signals and the spatial correlation of multiple antennas in an unauthorized MIMO-NOMA system are fully considered, a spatially correlated block sparse Bayes learning algorithm (SC-BSBL) is proposed to solve joint UAD and CE problems. Simulation results show that SC-BSBL can make full use of block sparsity and spatial correlation, in an unauthorized MIMO-NOMA system, UAD and CE can be completed accurately when the number of users is large and the sparsity of users is unknown.

Please add a comparison to illustrate the advantages of this algorithm over other algorithms

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1. The main challenge in the joint design of user activity detection and channel estimation can be highlighted. 

2. In the main contributions, the rationale for the developed method needs to be explained. For example, why does the use of the block sparse structure of the signal facilitate performance improvement in terms of active user detection? 

3. Does the proposed method still work under a dynamic scenario? For example, the vehicular networks, as introduced in e.g.,“Learning-based Predictive Beamforming for Integrated Sensing and Communication in Vehicular Networks” and “Integrated Sensing and Communications (ISAC) for Vehicular Communication Networks (VCN)”. Please add related references to justify the simulation scenarios.  

4. In section 4.2, what are the channel model and the adopted channel estimation method? Some existing channel estimation methods can be added for discussion to improve the readability of the paper. For example, “Channel-Estimation-Aware Joint Radar-Communications Designs” and “Deep residual learning for channel estimation in intelligent reflecting surface-assisted multiuser communications”.  

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Authors presented algorithm for joint user activity detection and channel estimation based on block sparse Bayesian learning approach. This study is focused on grant-free MIMO NOMA.

They claimed that the proposed algorithm accurately complete UAD and CE under various user activation probabilities, SNR and number of antennas.

Technically, paper looks good. However, language, flow of writing do not motivate and encourage reader. Reader has to repeat several times to understand the message. For example, read the abstract itself, last 3-sentence. In this, first two sentence highlight about what they did and they are saying finally results show.....while reading it looks that finally, authors must be saying another thing what they did. But, it is about result. Mistakes are in the entire text and therefore suggest proofread.

Also, normally if we follow the structure of typical paper, we find 5 paragraph in the introduction where 4th para highlights the contribution and novelty. Here, authors wanted to include a section for 3rd para and then contributions. It might be ok but review section can still move after contributions.

Normally, in the abstract, we have to quantify the result. It is not sufficient to say better performance or better accuracy or so. If accuracy, what %age?

In the result section, even this is not presented. What are the performance parameters for detection and estimation? I can only see error detection rate? Using MMSE and LSE will provide different performance. But, on which parameters UAD and estiamtion depends? Have we evaluated them? We do not find them. THese have to be identified and included.

Simulation setting parameters are limited. It is fine what is done, however, have authors checked for different modulation schemes, have authors considere more no. of users, have author consider more no. of antennas??? If not, can we generalize the model? With QPSK, algorithm can converge faster, what about with 256QAM? What about BPSK? In order to generalize, we have to show them.

Conclusion normally, has 3 points. CHallenges/observations is missing. Future work/scope is missing.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Title - "Block Sparse Bayesian Learning Based Joint User Activity Detection and Channel Estimation in Grant-Free MIMO-NOMA"

Can we say - Block Sparse Bayesian Learning based Joint Detection and Channel Estimation in Grant-free MIMO-NOMA?

By this, we can remove two word without changing the meaning.

Whether grant-free MIMO-NOMA is applicable only to machine type of communication? If not, why are we emphasizing on mMTC? Yes, it might work well in mMTC but not exclusively. Starting sentence with mMTC is not a good flow.

I am not in opinion to have 1.1 and 1.2 as separate subsection. The contents just with change of paragraph, they can be there.

Typo - Line 143 -So, it can makes .....

 Line 157 -The system’s models are shown in Section 2. The proposed BSBL algorithm grant-free MIMO-NOMA are shown in Section 3. The simulation results are provided in Section 4, and Section 5 provides the conclusion.

Shown...can be changed to presented, discussed, etc.

Sentence -Line 166 -Suppose K users send data to a central base station. Each user is activated with the probability of Pa, and the system has N subcarriers.

Last segment of the sentence does not fit well in the full sentence.

What is the meaning of each user is activated with probability of Pa? Who activates them? (base station)?

Should 2.3 be part of section 2?

Look at this -Line 288 - THe channel is set to block the fading channel, and its elements obey independent complex Gaussian distribution 289 CN (0, 1).

What is the meaning of block the fading channel? Is it blocking/stopping or block fading?

Authors claim is to detect active user (as one of the claims). What are the different performance measure for this? Should not we find probability of detection? Should not we find false alarm? Should not we find number of active user detection when number of users are varying? What else? UAD vs SNR is obvious, is it a performance para?

Similarly, for channel estimation - what are the performance parameters? calculating NMSE or convergence ....do they serve the purpose?

I hope, performance parameters must be identified and analyzed.

Language is still an issue. Improvement is there but flow, sentence is not clear. It can be done only by technical person who have done the work and who knows the work.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf