sEMG-Upper Limb Interaction Force Estimation Framework Based on Residual Network and Bidirectional Long Short-Term Memory Network

Round 1

Reviewer 1 Report

-        Page 1 line 31 – Maintain the consistency in describing various methods. Some of the techniques listed as electromyography where others are listed as electroencephalogram. Authors are advised to use graphy at the end if they are describing a technique or gram if its related to record of human physiological signals. It is suggested to change electroencephalogram and electrocardiograph to electroencephalography and electrocardiography.

-        Page 3 line 86 – insert a space between Figure and 1

-        Page 11 line 311 – Authors proposed to choose four healthy adults in their experiment. However, figure 12 results show data sets from five participants. Authors are advised to revise their statement and be consistent with number of participants.

-        Page 11 line 330 – Figure 8 experimental setup doesn’t show a one-dimensional force sensor on the wrist and there is no description provided on the third EMG sensor in the description (line 316 to 327). Authors are strongly advised to provide information about the force sensor as well.

-        Page 12 figure 9 – X axis name should be Time (sec).

-        Page 13 figure 10 – All the plots show x-axis title Time Series. However, looking at the graphs, it looks like number of samples. Authors are advised to revise their plots and make necessary changes. Authors are also advised to maintain consistency in the font size for x and y axis titles and placement of letters a to I. For an example: Figure 10 (k) shows smaller y axis font size then 10(I): K and I letter places after Time Series don’t use similar spacing. This figure needs to revise. It is advised to include waveforms from two participants on each plot for better understanding.

-        Page 14 line 362 – There is no figure 13 in the paper.

Author Response

please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The method is interesting, but the description of experiment and presentation of the results need to be improved. Wrong units were used. Setup of force measurement have to be provided.     References: Very similar topic was researched in paper " sEMG-Based Identification of Hand Motion Commands Using Wavelet Neural Network Combined With Discrete Wavelet Transform - Feng Duan, Lili Dai, Wennan Chang, Zengqiang Chen, Chi Zhu, and Wei Li". It should be discussed.   Detailed remarks: Figure 3: lowest delta t not aligned properly L317: In fig. 8 there are 3 EMG (not 2 as in text). What is the "reference"? It needs to be clarified. Additionally, it was not described how the signals from two EMGs were used (paper describes processing of single EMG?).   L321: configuration of the force sensor: direction of force measurement for EACH elbow angle need to be described (schematic picture?) . As well as type and the parameters of the force sensor user in tests.   L327: Three times each? So for 5s there were longer breaks between MVC? Needs more description.   Figure 9: Where disappeared 10s breaks between trials (described in text)? Or this is picture for 3s trial divided into single seconds? Or it is example not connectec with trials? It has to be clarified. In general in Line 327 1,3,5s periods were mentioned, while later there are 1,2,3s periods. Why? Timing and series description should be more clear. The horizontal axis is describes as "time series". probably should be "samples". Much better axis description would be time in [s].   L341: 80% trainng 20% test? With 4 persons, 12 different trials (table 1) and unknown number of trial repetitions (1?) how it was actually divided into 20/80? Need to be explained in details. Additionally, 20% of test seems to be not too much for teh evaluation of the method - why only 20%?   Fig. 10 g,h,i actual force j,k,l estimated force are not well described.  j,k,l have the legend "Actual, Estimate", while there is only one curve. Additionally the horizontal axes  have different scales (5000 samples and 1000 samples - it makes impossible to compare actual/estimated forces.  Actual and estimated forces HAVE to be on single graph (the same scales) for comparison.   The unit of force in SI unit system is not [Nm] but [N]. So it is probably a moment (of the force), not a force. It has to be corrected.   If the actual moment was calculated from the force measurement, it need to be described at the setup description - how the moment was calculated, what was the arm of force and so on.   Fig. 11.  How Coefficient of Determinate was calculated: for whole trial (including 10s breaks) or for active regions only? If for the whole trial, the passive regions (breaks) will artificially rise the CoD. Needs to be clarified.   Fig. 12: What are the "subjects 1-5"?

Author Response

Please see the attachment

Author Response File: Author Response.pdf