bci2000.org BBS

One of the great parts of Brain Computer Interface research is the collaboration between medical doctors / neuroscientists, and bioengineers. Obviously, they speak a different language. To improve collaboration and project outcome, it would help if both could learn (at least the basics) from each other's language.

Speaking for myself: I am an MD with a background in neurosurgery. Although I do have good IT expertise, I do not have engineering expertise. The "Brain" component from BCI is familiar, it is my field of knowledge. The "Computer" component seems also okay, because I can fit it to my general IT expertise. The problem is the "Interface" component. The paragraphs on signal processing in the BCI literature are very hard to understand for me, as a medical doctor.

Are there any good introductory resources on digital signal processing, focusing on the BCI background? I try to avoid complex formulas as they don't help me. And I don't need DSP for any other application. I want to grasp the "framework", the foundation, for linear classifiers (LDA, SVM), and some non-linear (MLP, LVQ, Kalman). I know the review from Lotte (2007) but it lacks the basics for somebody without an engineering background.

(btw: this also includes preprocessing of the data, like filters, feature extraction and feature selection)

Many thanks,

Pieter

http://dign.eu/bci

Pieter,

It is quite possible that the answer to your question is effective collaboration rather than trying to understand the details of a different field. While it would undoubtedly be advantageous to understand as much as possible about the different contributing domains in BCI research, making decisions in the more unfamiliar domains may not.


Gerv

Pieter,

I'd recommend starting with the Wolpaw & Wolpaw book parts 2 & 3 http://www.oup.com/us/catalog/general/? ... 0195388855. From there, you might want to check out coursera's machine learning course.
https://www.coursera.org/course/ml
This course is not catered to BCI but it will help you understand how to go from a superset of features to an optimized linear or non-linear classifier. This is mostly the domain of engineers and I don't disagree with Gerv that collaboration would be effective for this step.

-Chad