Artificial Neural Network with Electroencephalogram Sensors for Brainwave Interpretation: Brain-Observer-Indicator Development Challenges

This paper reports on challenges and opportunities associated with the development of an electroencephalogram (EEG) based personalized device for monitoring of brain activities pertaining large scale neural dynamics in the observed and providing relevant feedback to the observer. The envisioned device interprets signals and categorizes them on classes of typical responses. This could enable a speechless interaction between an observer and a participant wearing the device. This framework is different from the brain-computer-interface (BCI) framework as it focuses on indicators relevant to the human observer, brain-observer-indicator (BOI). Sensors detect resting states of the brain with associated patterns, synchrony between regions, and spectral changes in response to a cognitive event. A cognitive event results in notable changes in the associated patterns of electrical potentials. Recognition of these patterns has a broad application base, if the pattern-activity mechanism is characterized and recognized. The scope of the project includes development of a smart interaction support system BOI, relying on utilization of an EEG toolkit and an artificial neural network for personalization. The objective is to develop software that will support applications requiring feedback (i.e., training), along with a method for obtaining statistical data on the associated brain activity for engineering studies geared to improve signal acquisition and device performance. The findings from preliminary stages of the project are encouraging but indicate multiple challenges that must be addressed including a trade between a reduction of noise and complexity of classification software, definition of classes and recognition of classes and patterns, and development of an effective training data set acquisition strategy.