Abstract - The study of cognitive states has attracted the attention of artificial intelligence researchers searching for mechanisms to enable brain-computer communication. With the advent of portable brain-computer interfaces, it is now possible to study human behaviors towards using cognitive states in gaming environments. NeuroSky’s Mindset is a device with the operating principle of enabling portable EGG sensors to allow the reading of brain frequencies in real time. We believe that this type of device may be a very adaptable option to videogames to create new experiences and allow a new control mechanism. This interaction would be easy and natural and based less on motion and physical effort. This paper reports an assessment of the Mindset reader, particularly in relation to classifying the cognitive states of attention and relaxation, behaviors associated to the brain waves read by the device, using supervised learning algorithms. The aim is to estimate behaviors using human brain frequencies as inputs. I.INTRODUCTION Attention processes are typically related to the user’s psychology [8] since these refer to cognitive processes reflected on the behaviour of individuals. Traditional approaches to study attention include the determination of the relationships between the functioning of the nervous system and different processes comprising sensory, motor and cognitive. The study of the cognitive state of attention as computer input is a novel area requiring the collection of accurate data in the context this data can be applied to. Attention is read since brain activation is higher in some areas including frontal and parietal lobe. In psychological settings, however, current measurements of attention is carried out by standardized tests that are necessarily subjective as the person can express something different from what they are experiencing. This paper reports on the use NeuroSky’s Mindset device, a brain computer interface capable of reading neural activity based on the frontal lobe brain using electroencephalogram (EEG) principles. Direct readings of brain electrical frequencies using devices such as the Mindset might be an objective way to estimate the individual’s attention. The Mindset already generates two measurements with its proprietary software but we decided to work with raw measurements to analyze and determine whether these states are detectable using artificial intelligence classifiers. Previous works [1,2,3,4,5,6] have shown that the classification of cognitive states is possible [7]. Therefore, it is important to establish whether the readings of the full-wave spectrum allow the reading of brain activations in association with expected behaviors. An advantage of using raw readings is that it is possible to analyze user’s records when exposed to different types of stimuli, to determine whether the readings provided can be mapped out to human cognitive states. II. METHODOLOGY The study reported on this paper followed a methodology shown on Fig 1. This methodology consisted of showing different stimuli to 20 undergraduate students whose ages ranged between 18 and 28. This allowed the exploration of different tasks while capturing neural activation via the Mindset device. The six waves considered on this study are Low Beta, Medium Beta, High Beta, Alpha, Delta and Gama. The resulting data was classified using supervised learning algorithms including KNN, LDA, C 4.5 and Naïve Bayes, The results obtained showed promising prospective of classification with accuracy percentages of above 80%. Fig. 1. General diagram of the methodology. Matlab was used to connect the Mindset device to the computer as well as for recording, filtering and organization of the data. The Neurosky Lab library was employed by Matlab and amended for the purposes of this research. Visual stimuli were generated using Psychtoolbox 3.0. A. Frequency analysis The procedure used after data acquisition was to transmit the raw signal through a band-pass filter for each type of wave Candy Obdulia Sosa Jimenez 1 , Héctor Gabriel Acosta Mesa 1 , Genaro Rebolledo-Mendez 2,3 , Sara de Freitas 3 {cansosa, heacosta, grebolledo}@uv.mx Sfreitas@cad.coventry.ac.uk 1 Maestría en Inteligencia Artificial, Universidad Veracruzana, Mexico 2 Facultad de Estadística e Informática, Universidad Veracruzana, Mexico 3 Serious Games Institute, Coventry University, UK Classification of cognitive states of attention and relaxation using supervised learning algorithms