I am doing an EEG experiment about imagined speech recognition. Which frequencies will have the most relevant information? I chose the frequency band [4-40] Hz as a start, mainly because of signal processing considerations. Filtering out the remaining frequencies basically removes the DC voltage and most of the EOC artifacts.

Which frequencies should I select?

Also which channels will contain the signal (i.e., the most relevant data)?


A study by D’Zmura et al. (2009) in which two syllables were spoken in imagination showed that imagined speech information was present in EEG alpha, beta and theta bands. The beta band (13-18 Hz) proved most informative. The most informative electrodes were located mainly near the top of the head (vertex) where electromyographic artifacts had least influence.

They used various signal processing steps (standard low-pass filtering to remove 60 Hz line artifact, DC filtering, signal enveloping etc.) to clean up their signal. Please refer to the reference list for specific signal processing details.

- D’Zmura et al. Lecture notes in computer science 2009;40-8. HCI International
- Brigham & Kumar, iCBBE 2010


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.