Computer Vision Machine learning to detect body language

Request details
You will need to build a program that can look at a video of a person/s and be able to label them with a behavior/emotion based on their body language.
The feature of motion must be represented to describe the movements in temporal and spatial sequences such as MHI or MEI. produce an SVM classifier and one other and train it to label these features and whenever a new feature is seen the prediction will be able to estimate which label the feature is. An evaluation method ie K-fold cross-validation needs to be used to test the performance of the system.

The basic functionalities would be a feature detection and extraction SIFT feature
Feature representation MHI MEI
Data training and recognition using Support Vector Machine SVM
and Experimental results analysis
If possible I would like to compare different features and different Classifiers with the neural network
And run the system with a different data set with some more analysis

Deliverables
A written report on any background material you have used along with pre-existing code referencing it and where it has been found.
Any formulas pseudo-code and diagrams that can help explain the system and what you have done. A description of the experimental protocols and any test data and numbers of training and test images. A summary of any insights drawn from the analysis and comparison