A Feature Extraction Method Based on Differential Entropy and Linear Discriminant Analysis for Emotion Recognition

Round  1

Reviewer 1 Report

Paper is interesting.  But I have many concerns and they are given below:

Title: Looks complex. I think, it is a classification work?

Written English needs polishing.

The novelty of the part is not clear and explained?

Why other nonlinear methods are not used. I have provided few references:

Power spectral entropy analysis of EEG signal based-on BCI

Feature Extraction of EEG Signals Using Power Spectral Entropy

Characterization of focal EEG signals: a review

Analysis of EEG Signals using Nonlinear Dynamics and Chaos: A review

A Novel Depression Diagnosis Index Using Nonlinear Features in EEG Signals

Linear and Nonlinear methods for Brain Computer Interfaces

Please provide more details about the data? Is it from the public database?

Why LDA was used for dimensionality reduction? Why not PCA or other methods?

Tuning parameters of the classifiers need to be provided.

Authors should provide proposed system block diagram.

Looks like 70% and 30% cross validation was done. In my view: they should do LOOCV.

Feature extraction is not novel. Most of the features are already there.

Discussion table need to be included. Please discuss the features and results.

Include  Discussion table. Authors need to comment on the number of features and number of subjects used. Also, highlight on the application of deep learning as future work.

References:

Deep learning-based classification for brain-computer interfaces .

Deep learning with convolutional neural networks for brain mapping and decoding of movement-related information from the human EEG

Deep learning for healthcare applications based on physiological signals: a review

Discuss results, highlight advantages and disadvantages of your work.

Please provide future work and clinical implication?

Author Response

Please see the response to your comments in the pdf file.

Author Response File: Author Response.pdf

Reviewer 2 Report

li.li1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 12.0px 'Helvetica Neue'; -webkit-text-stroke: #000000} span.s1 {-webkit-text-stroke: 0px #000000} span.s2 {font-kerning: none} ol.ol1 {list-style-type: decimal}

Authors have proposed differential entropy and Linear discriminant analysis based feature extraction algorithm to recognize emotions using EEG signals. Though the application is interesting, the manuscript lacks novelty and the experimental design is very weak. Authors should consider addressing the following concerns:

What is the motivation of using differential entropy for feature extraction? Did authors explore any other entropy measures like sample entropy, modified multi scale sample entropy, wavelet entropy, etc.? Similarly, the motivation of using LDA for feature extraction is not clear. 

Were all 62 EEG channels used for extracting features? Did authors evaluate which channel provides the most discriminative information? A feature importance analysis will be insightful.

What parameters are used for training SVM and Random forest classifiers? Also, parameters for k-NN, logistic regression and other reported classifiers should be mentioned. 

Did authors tune the classifier parameters for each experiment reported in Table 1-5?

Did authors explore deep learning approach for emotion recognition? It will be interesting to provide a classification performance comparison of classic machine learning and deep learning approach using the same set of features.

Author Response

Please see the response to your comments in the pdf file.

Author Response File: Author Response.pdf

Round  2

Reviewer 1 Report

Authors have addressed my comments satisfactorily. So, I propose to accept the paper in its present form. Thank you.