AUT Main Nav
Centre Banner
Main Content


EEG Spatio –Temporal Data Analysis using Spiking Neural Networks in NeuCube Architecture

EEG data is related to Neuromarketing, Alzheimer progression and Facial expression as follows:

1. Neuromarketing:

Due to the vital role of unconscious processes in consumers’ decision making, understanding the human brain activity and neural performance scope is of crucial importance to predict the consumers’ decisions in the field of Neuromarketing. In this study, we investigated the pre-perceptual and pre-consciousness processes of the brain in posterior regions towards marketing stimuli. The main hypothesis tested here is: before the consumers are consciously aware of the marketing stimuli, their decision making is influenced by their brain pre-perceptual processes. We studied the hypothesis and proved it on a case study of EEG data that measured the brain dynamics during a cognitive task (familiar versus unfamiliar logos). In order to analyse the brain performance before decision, we selected the early components of ERP (the first 200 milliseconds of the EEG data) after the stimuli is presented. In the ERP waveform analysis and statistical results, slight differences can be regarded meaningful if they are presented in the ground average diagrams (Fig.1). To have better understanding about these findings, we used the NeuCube Spiking Neural Network (SNN) architecture for EEG-ERP data modelling, learning, and visualisation, to reveal significant information about the consumers’ brain processes. Tracing the NeuCube SNN-based model connectivity, enabled us to find out that consumers’ decision making may happen even before the consciousness.

Fig. 1. Ground Average Waveforms.  Parietal lobe: Familiar and Unfamiliar logos; Occipital lobe: Familiar and Unfamiliar logos. Black line represents the ERP waveform of familiar logos while the red line represents the unfamiliar logos


 Fig. 2 Model neuronal connectivity of the pre-perceptual processes while the subjects are still unconsciousness towards the marketing stimuli. The SNNc is trained by 200 milliseconds of EEG data after the stimuli presentation. (a) SNNc trained on EEG data of familiar logos; (b) SNNc trained on EEG data of unfamiliar logos; Initial connections between the input neurons are created based on the small world connectivity rule; the blue lines are positive (excitatory) connections, while the red lines are negative (inhibitory) connections. Thickness of the lines identifies neuronal enhanced connectivity after the learning process 

2. The Progression of Alzheimer Disorder:

Motivated by the dramatic rise of neurological disorders, we proposed an SNN technique to model electroencephalography (EEG) data collected from people affected by Alzheimer’s Disease (AD) and people diagnosed with mild cognitive impairment (MCI). An evolving spatio-temporal data machine (eSTDM), named the NeuCube architecture, is used to analyse changes of neural activity across different brain regions. The model developed allows for studying AD progression and finding whether the MCI patient is likely to be developed as AD over the time. This model also enabled us to have precise analysis about brain activates across EEG band-frequencies (Alpha, Beta, theta and Delta). Using the NeuCube SNN-based visualization, we can obtain a better understanding and interpretation of the physiological brain ageing of AD patients.



3. Neurocomputational EEG Analysis on Perception and Production of Facial Expressions:

We used the NeuCube architecture of SNN for mapping, learning and visualisation of EEG data recorded from subjects when they were performing a facial expression-related task.  Using the NeuCube, as a potential spatiotemporal data machine, enables us for the first time, to study the level of variation between perceiving versus mimicking facial expressions. Also the visualization of the SNN models trained on the EEG data proves the mirror neurons principle in the brain. We identified that role of mirror neurons can be dominant in sadness emotion, compared with the other emotions.


Fig. 3(a) Exposing the facial expressions on screen (b) Connectivity of a SNNc trained on EEG data related to perceiving the sadness emotion image (c) Connectivity of a SNNc trained on EEG related to mimicking the sadness emotion image (d) subtraction between perceiving and mimicking emotions to visualise the differences between SNNc of (a) and (b) 

Last updated: 17 May 2016 3:15pm

Auckland University of Technology, New Zealand | Copyright © | Privacy | Site map