Training a Classifier to Detect Instantaneous Musical Cognitive

Functional Magnetic Resonance Imaging is a brain im-aging technique that allows the observation of brain activ-ity in human subjects based on the increase in blood flow to the local vasculature that accompanies neural activity in the brain. It produces time-series data that represents brain activity in a collection of 2D slices of the brain. The collec-tion of the 2D slices form a 3D image of the brain contain-ing in the order of 12000 voxels, i.e. cubes of tissue about 3 millimetres on each side. Images are usually taken every 1-5 seconds. Despite the limitations in temporal resolution, fMRI is arguably the best technique for observing human brain activity that is currently available. Figure 1 shows three axial slices of the brain and the activity over time of the voxel in which the slices intersect. Figure 1: fMRI image showing the instantaneous activity of the brain: white voxels are those with highest activity while dark voxels are those with lowest activity. Greatly due to the advent of Functional Magnetic Reso-nance Imaging (fMRI), the study of human brain functions has dramatically increased in recent years. While fMRI has been extensively used to test hypothesis regarding the loca-tion of activation for different brain functions, the problem of automatically decoding cognitive states has been little explored [4]. The study of this problem is important be-cause it can provide a tool for detecting and tracking cogni-tive processes (i.e. sequences of cognitive states) in order to diagnose difficulties in performing a complex task