Sleep tracking headband

Jeff_B · July 2018

Hi there,

I was thinking to sleep with a Ganglion board (the definition of a geek ?) and a headband fitted with 4 flat electrodes (& pulse oximetry).

And then use the OpenBCI Python driver to implement ICA and SleepNet (Python) for postprocessing.

Do any of you do it or have you done it for sleep research ?

What was your experimental setup (headband/mask/headset ? electrode type ? ganglion board/battery fixation ?..).

For processing in real time & neurofeedback during sleep, anybody tried

- to use available io pins on the microcontroler of the OpenBI boards to read analog inputs (from an oximeter for exemple) or set digital outputs for feedback cues (to blink a light, play a sound,...) ?

- to use the ARM Cortex M0 (Ganglion) or the PIC32 (Cyton) to compute real time FFT to detect slow waves ?

Best,

Jean-François

wjcroft · July 2018

Here are some previous threads, found using the search button in the right column,

https://www.google.com/search?as_q=sleep+study&as_sitesearch=openbci.com

wjcroft · July 2018

re: "FFT to detect slow waves".

Easier and less computation than an entire FFT in the microcontroller, would be a simple digital filter, bandpass filter from say .5 to 4 hz. Then the moving average of the absolute amplitude values would give you the rough power in say the delta band.

Take a look at Hassan's WakeCap project

https://www.google.com/search?as_q=wakecap&as_sitesearch=openbci.com

Jeff_B · July 2018

Hi William,

Yes, bandpass-filtering and estimating the delta power... ultimately, it is to try to implement a sound stimulation that would be in phase (phase locked) with (a part of) the delta EEG waveform, and I was wondering if a biquad (IIR) filter could distord this phase in the delta band... and if it's possible to revert to a FIR filter selective enougth with a reasonable order (I have to see this).. I also wanted to implement a sleep staging algorithm which needs the other EEG bands for the 4 channels but maybe it's too much for the Cortex M0.

Jeff

wjcroft · July 2018

Much has been published in the area of slow wave enhancement via sound,

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5340797/

https://www.google.com/search?q=auditory+closed+loop+stimulation

https://www.google.com/search?q=enhancing+sleep+slow+waves+with+natural+stimuli

https://www.google.com/search?q=j&j+dreem

https://www.google.com/search?q=philips+sleepsmart

Jeff_B · July 2018

Yes, I know these products, the SleepSmart will only be available in 2019 in Europe/France (~350 € in the US) and Dreem is a very nice device (~500€), but it is not easy to get access to raw data for the time being. I am also a KS backer of the Neuroon v2 (visual feedback)(access to raw data) but...

Jeff

wjcroft · July 2018

Jeff, interesting.

If you end up doing an open source or other type of sleep project using OpenBCI, please post here in this thread your Github repository or website. Would be fun to try this.

I'm also aware of some groups working on lucid dream induction using a combination of EEG tracking and tACS 40 hz gamma stimulation.

http://aladdindreamer.com/

https://www.google.com/search?q=ursula+voss+lucid+dreaming

Regards,

William

Jeff_B · July 2018

Hi William

I will stick to acoustic stimulations, but I would like their waveform (phase, freq., amplitude) to be adapted in real time and not be a fixed train of stimulations.

My immediate concern is to see how I can implement on board in real time an efficient sleep staging algorithm like:

https://arxiv.org/pdf/1707.03321.pdf

The most demanding computations are the 4 STFT (N=512, overlap = 256, windowing @1Hz)(I have forgotten the idea of a real time ICA), the features extractions over 30s, and the multiple convolutions (+ReLu+max pooling) for the CNN and I will also need memory / computation ressources to optimize the stimulation waverform.

Jeff

Sleep tracking headband

Comments