C# body gesture recognizer with CNTK (GPU) and Microsoft Kinect

The project consists in completing an already started project in C# language, WPF program.

The system is working in this way:
1) The Kinect is getting skeleton data from user

2) After pressing a button, a timer starts and some features are acquired from the user:
a. 32 features, 9 of them are angles of joints, the others are X, Y and Z coordinates of joints
b. Data is stored into ArrayLists (and can be saved in files)
After some seconds the time is out so the system stops to acquire data. This is the window of information acquired.

3) I can use this data for training the model or for testing it. In the first case I should add these data to a file (append). In the other case they can be sent directly to the CNTK evaluator for predicting the action performed. This part has not been developed.

4) A LSTM learner is already implemented. It loads data from a specific example data format. The learner is not set with the features in 2.a and I still need to create the data serializer that saves these train features like in the format of this file.

5) The predictor should be created too.

So, the thing that have to be done are:
1) Create the features serializer for allowing the LSTM to read them
2) Set the LSTM for load the specific train and the test data
3) Create the predictor (Can’t find the function name… It probably consists in one line of code…)

All other things are working.

Looking forward to build a long term relation.