Archived project

Brain Hash Function

Goal: The goal is to create audio-visual stimulation capable to entrain cortex activity near measuring electrode into specific steady state generating reproducible response over particular frequency bands. Obtained response directed as input into trained auto-encoder will produce data as input for Brain Hash Function. The final output will be Brain Hash Value allowing to identify specific user with given confidence.

Updates
0 new
17
Recommendations
0 new
0
Followers
0 new
1
Reads
0 new
94

Project log

Iaroslav Omelianenko
added a research item
In this article we present the results of our research related to the study of correlations between specific visual stimulation and the elicited brain's electro-physiological response collected by EEG sensors from a group of participants. We will look at how the various characteristics of visual stimulation affect the measured electro-physiological response of the brain and describe the optimal parameters found that elicit a steady-state visually evoked potential (SSVEP) in certain parts of the cerebral cortex where it can be reliably perceived by the electrode of the EEG device. After that, we continue with a description of the advanced machine learning pipeline model that can perform confident classification of the collected EEG data in order to (a) reliably distinguish signal from noise (about 85% validation score) and (b) reliably distinguish between EEG records collected from different human participants (about 80% validation score). Finally, we demonstrate that the proposed method works reliably even with an inexpensive (less than $100) consumer-grade EEG sensing device and with participants who do not have previous experience with EEG technology (EEG illiterate). All this in combination opens up broad prospects for the development of new types of consumer devices, [e.g.] based on virtual reality helmets or augmented reality glasses where EEG sensor can be easily integrated. The proposed method can be used to improve an online user experience by providing [e.g.] password-less user identification for VR / AR applications. It can also find a more advanced application in intensive care units where collected EEG data can be used to classify the level of conscious awareness of patients during anesthesia or to automatically detect hardware failures by classifying the input signal as noise.
Iaroslav Omelianenko
added an update
The research article now available at arXiv preprint service:
 
Iaroslav Omelianenko
added a project goal
The goal is to create audio-visual stimulation capable to entrain cortex activity near measuring electrode into specific steady state generating reproducible response over particular frequency bands. Obtained response directed as input into trained auto-encoder will produce data as input for Brain Hash Function. The final output will be Brain Hash Value allowing to identify specific user with given confidence.
 
Iaroslav Omelianenko
added an update
The resulting research paper is deposited to the HAL repository:
 
Iaroslav Omelianenko
added an update
Further improved classification power of ML model leading to almost 85% classification score with exhaustive grid search over hyper parameters and extended input data as following: 10 Hz/5Hz, wisp, attention, 70, cA5 delta, theta, alpha low, alpha high, beta low, beta high, batch size = 1
 
Iaroslav Omelianenko
added an update
By applying exhaustive grid search it was found that Gaussian Naïve Bayes Classifier demonstrate best prediction power in classification of signal and noise records. As can be see from picture attached it even outperforms Multi-Layer Perceptron. Our assumption that its due to the small number of data samples available which is natural to any research related to neuroscience.
The results shown was acquired from records preprocessed with Contractive Auto-encoder with following hyper parameters: 10 Hz/5Hz, wisp, attention, 70, cA3, delta, theta, alpha low, alpha high, batch size = 1.
The achieved classification score of 0.833 can be considered strong enough giving high confidence level that analyzed record is actual EEG signal collected from conscious human rather than noise from EEG device.
 
Iaroslav Omelianenko
added an update
The results of experiments with three subjects.
Findings
  1. There are strong correlation as within each subject records as between different subjects records
  2. There are considerable difference between noise records and signal records allowing to separate actual EEG signal received from entrained mind and noise signal
  3. The same results found for both stimuli configurations
 
Iaroslav Omelianenko
added an update
The best correlation found for 8/4Hz and 10/5Hz with filtered bands correspondingly: theta/alpha-high, delta/alpha-high
cA 2
Compared against noise. The both results seems very promising.
 
Iaroslav Omelianenko
added an update
The correlations between 8Hz / 10Hz / Noise, cA 16
 
Iaroslav Omelianenko
added an update
Second session (8) best analysis output results (best reconstruction costs)
 
Iaroslav Omelianenko
added an update
The correlation map for first session (10/5). The session results log file.
 
Iaroslav Omelianenko
added an update
Project goal
The goal is to create audio-visual stimulation capable to entrain cortex activity near measuring electrode into specific steady state generating reproducible response over particular frequency bands. Obtained response directed as input into trained auto-encoder will produce data as input for Brain Hash Function. The final output will be Brain Hash Value allowing to identify specific user with given confidence.
Background and motivation
  • The primary goal is to build robust classification system capable to reliably discern users taking their EEG inputs. The output of the system should be akin to hash function output which will represent particular user’s reaction to specific stimuli. This data allows to uniquely identify user and discern him among others.
  • The created Brain Hash Algorithm should include confidence level
  • The secondary goal is to build or propose biometric authentication system based on achieved results
  • We assume to achieve performance allowing to apply methodology for real time user authentication. Which means that authentication should take matters of seconds and no more than tens of seconds. Of course, this will heavily depend on context of system usage - for highly secure access rarely used (e.g. online banking) its affordable to perform robust authentication taking more time, than for consumer web based systems such as access to web site resources. Its possible to complement EEG based authentication with additional user identification means such as behavioral patters, fingerprints of device, etc in order to further enhance prediction accuracy and avoid user’s secure credentials leak.