An Intelligent Error Correction Algorithm for Elderly Care Robots

Round 1

Reviewer 1 Report

In this manuscript, the authors propose an error correction algorithm for hand gesture recognition in elderly care with the Pepper robot. The manuscript is not very well written: it is very long and redundant; there are several typos that should be corrected; it lacks some fundamental references to the related work (e.g., what is the previous work done by the authors mentioned in the introductory section? how do you get the three-dimensional surface maps? etc.); finally, it lacks details to ensure the reproducibility of the study. Also, the resolution of the figure is really poor.

The central part of the article, that is the description of the error correction algorithm, is not easy to follow and it seems to me that there are some inconsistencies between the formulas and the tensor dimensions. Furthermore, there is no real connection between the proposed algorithm and the cognitive and biological reasons why an elderly person performs gesture recognition differently than a young person. The algorithms appear to examine the differences between the feature maps, so this makes it independent of the specific domain input images.

As for the experimental evaluation, it is not clear how the images of the hand gesture were extracted from the frames captured by the camera Pepper is equipped with.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors propose a method for correcting errors in recognition of gestures performed by elderly people. The method seems effective and helpful, which is proved by the results presented in Section 4.3. However, the paper has many flaws, which the authors need to address before publishing.

1. I suggest removing words "is presented" from the title. It should be "An Intelligent Error Correction Algorithm for Elderly Care Robots".
2. The title of section 3.1 should not start with "Based on". I think this title is wrong because it does not tell the reader what the subsection is about.
3. The authors should explain what is Pepper. If it is a robot, then I don't understand this sentence: "…and finally the humanoid robot developed by Softbank can only be operated by pepper". Is a robot operated by a robot? And why its name is written with lowercase letter?
4. The authors claim that the four gestures in Fig. 8 (00, 01, 02, 03) are very similar to 01. For me only the first two are similar to 01.
5. How many gesture classes are used in the experiments? Table 1 suggests that there are 7 classes, however, Table 2 suggests that there are 6 classes.
6. From what I understood, the dataset consists of 20 subjects performing gestures from 6 or 7 classes (see the comment 5) and the total number of gestures is 200. It should be clearly written in one place.
7. Please specify what is the training and the testing set in the experiments.
8. Please specify what kind of validation tests was used. Were they cross-validation tests?
9. How many people participated in surveys based on which the statistics presented in Figures 12 and 13 were made?
10. The conclusions are poorly written (grammatically).

Language:

• "Therefore, in order to improve the gesture recognition rate of the elderly accompanying robot, improve the recognition rate of the elderly accompanying robot." – I don't understand this sentence
• ", On the one hand, based on the behavior of the elderly, the recognition probability of each gesture is counted, the misreading-gesture database and the probability matrix are established, and it is concluded that the recognition error is caused by the non-standard gestures made by the elderly;" - this whole text should be in brackets without the first comma, without the capital letter O (on) and without the last semicolon: "(on the one hand (...) by the elderly)"
• In Table 1 "2 The palm of your hand steady" - should be "2 The palm of a hand is steady"
• What means "His fist" in Table 2? Shouldn't it be simply "Fist"?
• "(The reason for choosing the fifth layer will..." - first words of bracketed fragments should not start with capital letter
• "But fundamentally speaking, it is still two different gestures." - should be "...they are still two..."
• "...(143, ?12,i)}, Finding approximate formula (3)" - should be "finding" without capital letter
• In Algorithm 1 - "Find out who has won two games, because everyone has played two games, and two games are all wins" - there is something wrong with this sentence. What means "two games are all wins"?
• "matrixes" (multiple times) - the plural of "matrix" is "matrices"
• "fifth layer of convolution layer" (multiple times) - shouldn't it be "fifth layer of convolutional layers"?
• "In order to verify whether the error correction algorithm can improve the accuracy of gesture recognition, and also to enhance the user experience." - this sentence lacks predicate
• "To obtain the number of times that the robot finds errors and corrects errors to illustrate the effect of the algorithm." - I don't understand this sentence. It lacks predicate.
• "In the aspect of helpfulness evaluation, because the error correction algorithm based on game rules improves the gesture recognition rate, the probability of robot providing correct help is greatly improved, so the evaluation is relatively high." - should be "In the aspect of helpfulness evaluation, the probability of robot providing correct help is greatly improved because the error correction algorithm based on game rules improves the gesture recognition rate"
• "This paper has produced a patent, but the patent is still in the trial stage." - should be "This paper has produced a patent, but it is still in the trial stage."
• "The process of error correction is: according to the neural network identification number to determine whether to correct, if the identification number belongs to the error prone number, then extract the fifth layer volume base level feature value of the recognition object at this time, and calculate the three-dimensional surface peak value on Channel 6 and channel 58, finally determine the peak value of the gesture, and then output the error correction identification number." – this sentence is incomprehensible and too long.
• "In this paper, an intelligent error correction algorithm is proposed to solve the problems of ambiguity, the robot can not understand the real intention of the elderly, the accompanying process is not smooth, the elderly can not get good service and so on." - I don't understand this sentence.
• "can not" – should be "cannot"

The paper needs proofreading (especially Chapter 5).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

All my previous comments have been addressed.

Author Response

Thank you very much for your review.

Reviewer 2 Report

1. "Only six gestures are listed because the seventh gesture rarely misjudges each other with other gestures. The probability is negligible.", "According to table 2, we need to select the gestures with a recognition rate lower than 85%..." – So why gesture 04 was selected and presented in Table 3? Its recognition rate from Table 2 is above 85%.
2. "The method used in this paper is hold-out. The division of training set and testing set is shown in 7. This paper mainly explores the intelligent error correction algorithm. The training of network model is not the focus of work. And because the amount of data of each kind of gesture is large enough. So I chose the hold-out method." – Please write about using the hold-out method in the paper.

There are still some language issues:

• "Winning two games is all win and the only one" – I don't understand this sentence.
• "...the recognition results will be in the label." - I don't understand.
• Comments in Algorithm 1 are full sentences and therefore, they should end with dot.
• "the reason for choosing the fifth layer will to be given in the experiment section" - should be "will be given"
• "Whether error correction is based on neural network identification." – something is missing in this sentence. I don't understand it.

Author Response

Thank you very much for your review. Please see the attachment.

Author Response File: Author Response.pdf