My work for the pas few years has been into Brian Computer Interfaces.
You have to work with what you have, in my case I have access to a wireless Electro Encephalographic.
The Emotiv EPOC, 128 samples per second, wireless, 14 Channels, plus 2 gyro channels.
I created a program to record simultaneously from the EPOC and from the keyboard and mouse.
Data was saved to binary files to reduce file size.
was used as a data mining / machine learning program to try to extract data from files.
We used the built in artificial neural net(ANN) of WEKA.
We looked at mouse movements specifically. One seconds worth of data proceeding the movement was used for the learning task.
A true positive rate of 70% was archived. The data is a time series, so I am currently investigating a newer type of artificial neural net called a Liquid State Machine(LSM).
An LSM is in the newer class of ANN, They in a broader sense have been called reservoir computing.
The general set up is that there is an input that is fed into a dynamical system. Output is created by training simple classifiers on a state in the network that
represents your desired output class.
Currently I'm working on some final code to run an LSM and working on my thesis.