Real-time imaging of human brain function by near-infrared spectroscopy using an adaptive general linear model

Many researchers use electroencephalograms (EEGs) to study brain activity in the context of seizures, epilepsy, and lie detection. It is desirable to eliminate EEG artifacts to improve signal collection. In this paper, we propose an emotion recognition system for human brain signals using EEG signals. We measure EEG signals relating to emotion, divide them into five frequency ranges on the basis of power spectrum density, and eliminate low frequencies from 0 to 4 Hz to eliminate EEG artifacts. The resulting calculations of the frequency ranges are based on the percentage of the selected range relative to the total range. The calculated values are then compared to standard values from a Bayesian network, calculated from databases. Finally, we show the emotion results as a human face avatar.

Emotion recognition using EEG signals with relative power values and Bayesian network

Decoding and classification of objects through task-oriented electroencephalographic (EEG) signals are the most crucial goals of recent researches conducted mainly for brain–computer interface applications. In this study we aimed to classify single-trial 12 categories of recorded EEG signals. Ten subjects participated in this study. The task was to select target images among 12 basic object categories including animals, flowers, fruits, transportation devices, body organs, clothing, food, stationery, buildings, electronic devices, dolls and jewelry. In order to decode object categories, we have considered several units namely artifact removing, feature extraction, feature selection, and classification. Data were divided into training, validation, and test sets following the artifact removal process. Features were extracted using three different wavelets namely Daubechies4, Haar, and Symlet2. Features were selected among training data and were reduced afterward via scalar feature selection using three criteria including T test, entropy, and Bhattacharyya distance. Selected features were classified by the one-against-one support vector machine (SVM) multi-class classifier. The parameters of SVM were optimized based on training and validation sets. The classification performance (measured by means of accuracy) was obtained approximately 80 % for animal and stationery categories. Moreover, Symlet2 and T test were selected as better wavelet and selection criteria, respectively.

Decoding objects of basic categories from electroencephalographic signals using wavelet transform and support vector machines.

http://dsp.ucsd.edu/~zhilin/papers/McICA.pdf

Morphologically constrained ICA for extracting weak temporally correlated signals

Expert systems (ES) are a branch of applied artificial intelligence. The basic idea behind ES is simply that expertise, which is the vast body of task-specific knowledge, is transferred from a human to a computer. ES provide powerful and flexible means for obtaining solutions to a variety of problems that often cannot be dealt with by other, more traditional and orthodox methods. Thus, their use is proliferating to many sectors of our social and technological life, where their applications are proving to be critical in the process of decision support and problem solving. Earthquake professionals for many decades have recognized the benefits to society from reliable earthquake predictions, but uncertainties regarding source initiation, rupture phenomena, and accuracy of both the timing and magnitude of the earthquake occurrence have often times seemed either very difficult or impossible to overcome. This research proposes and implements an expert system to predict earthquakes from previous data. This is achieved by applying association rule mining on earthquake data from 1972 to 2013. These associations are polished using predicate-logic techniques to draw stimulating production-rules to be used with a rule-based expert system. The proposed expert system was able to predict all earthquakes which actually occurred within 12h at-most.

Developing an expert system based on association rules and predicate logic for earthquake prediction

An electroencephalograph (EEG)-based brain computer interface (BCI) requires rapid and reliable extraction of features in EEG signal. Recently, the rhythmic component extraction (RCE) method has been proposed to extract features of multi-channel EEG. RCE can extract a signal component with a certain frequency from multi-sensor signals. In this paper, we applied RCE to extract a feature corresponding to hand movement imagery tasks from signals measured by EEG. This feature from a single trial EEG signal is classified between imaginary left/right hand movement EEG using machine learning. On two subjects, our experiment shows that the combination of RCE and fisher discriminant analysis outperforms common spatial patterns (CSP) in classification accuracy. It is also reported that other major classifiers together with RCE give better performance than CSP. Additionally, we consider the relationship between data length and classification accuracy. It is shown that the accuracy tends to decrease as the data length becomes small.

Classification of single trial EEG during imagined hand movement by rhythmic component extraction

Recent years, a Steady-State Visual Evoked Potential (SSVEP) is used as a basis for Brain Computer Interface (BCI)[1]. Various feature extraction and classification techniques are proposed to achieve BCI based on SSVEP. The feature extraction of SSVEP is developed in the frequency domain regardless of the limitation in flickering frequency of visual stimulus caused by hardware architecture. We introduce here the feature extraction using a spectrum intensity ratio. Results show that the detection ratio reaches 84% by using a spectrum intensity ratio with unsupervised classification. It also indicates the SSVEP is enhanced by proposed feature extraction with second harmonic.

Spectrum based feature extraction using spectrum intensity ratio for SSVEP detection

Eye movement origin electrooculogram (EOG) artifacts yield significant problems for the saccade-related electroencephalogram (EEG) and its analysis. The denoising of EOG artifacts is important task to analyze the relationship between a saccade and a brain function. In this paper, a tensor product expansion with absolute error (TPE-AE) is applied to reduce the EOG artifacts from the EEG signal. The TPE-AE has some difficulty to separate the EOG from the saccade-related EEG data due to a background noise from a spontaneous EEG activity. We show that the TPE-AE, which calculates two outer products, is useful to reduce EOG components related to the eye movement.

EOG-related noise rejection in EEG signal with eye movement task by tensor product expansion with absolute error

The steady-state visual evoked potential (SSVEP) is used for input signals of brain computer interfaces (BCIs). There are two types of stimulus for SSVEP. One is flushing visual stimuli and the other is flipping checker board patterns. We have been studying SSVEPs with checker board patterns. There are few studies described relationship between SSVEP and property of flipping checkerboard patterns. In this study, We pay my attention to the size of the grid squares in the checkerboard and the length on the checker board pattern.

/pdf/ssvep-by-checkerboard-related-to-grid-size-and-board-size-4y4t6jr4kg.pdf

SSVEP by checkerboard related to grid size and board size

We have been researching relationship between saccadic eye movements and EEG signals before saccadic eye movements in order to develop the Brain-Computer Interface (BCI). In previous studies, we confirmed sharply changed EEG signals were recorded just before saccade. The EEG signals in right occipital lobe have low potential in the case of saccade to the right side and the EEG signals in left occipital lobe have low potential in the case of saccade to the left side. However, these EEG signals were recorded with the visually guided saccade task and the saccade in the visually guided saccade task is not voluntary movements perfectly. Involuntary movements are not suitable for inputs of BCIs. Therefore, in this research, we record EEG signals with the memory guided saccade task and compare EEG signals in the visually guided saccade task with EEG signals in the memory guided saccade task. In this research, we instructed subjects to perform following tasks. First, a visual stimulus (LED) on the right side or a LED on the left side is turned on and subjects memorize which LED is turned on. Secondly, after a few second, beep sound is turned on and subjects move their eyes to LED position where subjects memorized. As results, we obtained sharply changed EEG signal in occipital lobe just before eye movements. The EEG signals have sharp changes in right occipital lobe when subjects move their eyes to the right, and in the case of eye movements to the left, left occipital lobe have sharp changes. These results are similar to results in visually guided saccade tasks. Therefore, in the voluntary and the involuntary saccadic eye movements, sharply changed EEG signals in occipital lobe are observed just before eye movements and sharply changed EEG signals are suitable for inputs of BCIs.

Arao Funase

Papers

Classification of single trial EEG during imagined hand movement by rhythmic component extraction

Spectrum based feature extraction using spectrum intensity ratio for SSVEP detection

EOG-related noise rejection in EEG signal with eye movement task by tensor product expansion with absolute error

SSVEP by checkerboard related to grid size and board size

Analysis of EEG Signals in Memory Guided Saccade Tasks