This is a MATLAB 14-17 feasible code.
Download any database of your wish, better to have 1000 images atleast for each emotion.
Download all 3 files and savein your directory. Here, open the classification.m.
Change your folder name in the place of "EMODATB" - That was my folder name.
Now run the code section wise. Search for that option on your MATLAB toolbar. Once you run the classifier learning app, an additional window will be opened. There check if all features are available. Choose the second option HOLDOUT and keep that as 20% or 25%. This is basically shuffling and dividing the dataset into test and train datasets. Your model will be trained on the training part of dataset and will be tested on test-part of your dataset.
Now, open each machine learning model like decision trees, Support vectormachines, KNN classification algorithms - Here u need to use classifiers for our project etc. Train each model and check for accuracy. Read up on classifier learning app for better understanding.
Now in KNN, go to advance settings and keep k value = 3,5,7,9 and test for your dataset, which is giving better results.
open confusion matrix and ROC curves to understand your results better. Learn how to analyse confusion matrix online.
U can open the code of each model u train in matlab and fiddle with those parameters to check if u could get better results.
for any doubts mail to tejastk.reddy@gmail.com