Emotion detection from Videos using Neural Networks -- 2

I have videos dataset and I want to extract features from faces and eventually decide if the face is confused or not. (you can use any method to extract face features but we need to use the classifiers in point Number 3 below) the requirement are as follows: Confusion detection system based on facial expression: 1. Detecting face from sequence of video clips, by using viola-jones face detector. 2. Extracting features from the face image (kindly check the attached image for the features features needed) 3. Using logistic regression(LR) and quadratic discriminator (QD) and LSTM (long short term memory) as a classifier to train the data (frames from video clip each of 1 sec length that have been collected) and test the data, and classify the data to (confused, not confused) with high detection accuracy. 4. Using 20 action units to train the neural network LSTM, QD and LR. I need it in 5-7 days and release the money in one week after verifying its ok and no modifications required. please check the picture Thank you