Recently, Brain-Computer Interface (BCI) which allows a user to control external devices or to communicate with other people just by his/her thought has been studied. P300 speller is one of the BCI system for inputting letters which uses P300 as the feature quantity. One matrix interface is usually used for Japanese P300 speller. However, this interface has manycharacters on the display, and it may cause the increase of input time and the decrease of accuracy. In this paper, we propose a new interface which has two matrices in Japanese P300 speller to improve the performance of P300 speller and shows the comparison results between the conventional interface and the proposed one.
The present paper investigates eye-gaze data of professional train drivers with different years of experience using Markov Cluster Algorithm (MCL) in order to extract characteristic eye-gaze patterns fostering a better understanding of their visual perceptual skills. MCL distilled a basic eye-gaze pattern in their visual behavior indicating that all the drivers would repetitively move their gaze ahead soon after looking at another area of interest, but they were found different in the "strength" of the pattern depending on their level of expertise. It was also clarified that inexperienced drivers made frequent deviations from the basic eye-gaze pattern in particular segments of the route where they had to deal with multiple tasks in parallel imposing higher cognitive loads of them.
Human walking is affected by vision, vestibular, somatic and other various sensations that come through the sensory-motor loop. But detail of the sensory-motor loop is not clear. In this study, we examined a possible affect of self motion sensation by an optical-flow stimulus in peripheral vision with a decayed somato-sensory feeling by a vibration stimulus on leg and foot area. In this experiment, we presented the optical flow for forward direction to the peripheral vision, and then gave the self-motion sensation by changing the flow to left or right direction. We examine the change of walking direction toward opposite of the self-motion sensation because of the decayed somato-sensory feeling. In this paper, we discuss on the sensory fusion mechanism of visual and somatic sensations based on the experimental result.
The movement of human hands is originated from thought expressed via brain activation. For a handicap person, if his or her brain activation signals are captured and encoded, he or she could grab things by robot hands. Recently, for Brain Computer Interface (BCI), several technologies have been developed including EEG and fNIRS. fNIRS measures changes of oxygen called BOLD signal during activation of brain. In practice it is difficult to sense the BOLD signal stably because the signal continuously changes day by day and depends on individuals. This paper aims at developing a smart calibration system that calibrates parameters at high speed and keeps the generalization ability of the classifier. Using SVM we develop a discriminator that detects starting and ending of the BOLD signal. From this signal the discriminator estimates intended movement of the left and right hand. Because SVM is fast, the system can derive the subject-specific parameters in high speed. For calibration, we use the Differential Evolution method (DE). Trial experiments use the tapping task test and take the BOLD signal data in the motor cortex area. This paper also shows the calibration result and its possible parameters.
In this study, we propose an evolutionary fuzzy neural network based on structured learning for gesture recognition. In general, processing for gesture recognition consists of feature extraction part and gesture classification part. In most of the works, they are independently designed and evaluated by their own criteria. However, it is difficult to design the components without considering the relationship between each component. Structured learning can be a solution to the problem. One of the primary aims of structured learning is a mutual adjustment to improve the classifier's generalization ability. We use a neuro-fuzzy system for the classification of human gesture and apply an evolutionary approach to parameter tuning and pruning of membership functions.
The authors measured electroencephalograms (EEGs) from subjects on recognizing and on recalling 13types of playing card images (from Ace to King) presented on a CRT. Each presented image card is the club. The canonical discriminant analysis was applied to these single trial EEGs. Four channels of EEGs at the right frontal and temporal were used in the discrimination. They were Fp2, F4, C4 and F8 according to the international 10-20 system. Sampling data were taken from 400ms to 900ms at 25ms intervals from each single trial EEG. The number of variates is twenty one by four channels ; so the data are eighty four dimensional vectors. The number of external criteria is thirteen, from Ace to King, and a number of explanation variates is thus eighty four. Results of the canonical discriminant analysis by use of so called jack knife method were more than 90% for nine subjects. We could perform a playing cards estimation magic without a trick.
We has developed Air Brain system, wearable telemetry system for electroencephalogram. Features of Air Brain system is that it enables us to get electroencephalogram and to measure human behavior easily, at anywhere and in anytime. In this study, we attempted to utilize the system for BCI, detecting event-related potentials (ERPs) evoked by silent reading of the visual symbols (either upward arrow or downward arrow) presented on the monitor of a smartphone with experimental participants. We confirmed that there are relationships between the meanings of applied visual symbols and amplitude/polarity of ERPs at near the right frontal lobe of participants. Remarkable changes of the averaged EEG after visual stimulus were observed at around 400ms to 700ms from the start of the symbol-presentation. When upward arrow was presented to the participants, the positive peak of ERP expressed in 400ms to 700ms after visual stimulus. However, when downward arrow was presented to participants, the negative peak of ERP expressed in the similar timing. The rate of correct answer for silent reading of the direction of the presented symbols reached to 90%. These results are consistent to the previous reports, and it suggested that the Air Brain system, wearable telemetry system for EEG, was potentially applicable to BCI.
To elucidate higher order brain function, it is important to analyze spatio-temporal patterns of electrical activity of neuronal network. For this purpose, electrical activity in neuronal networks measured from dissociated neuronal culture system or acute slice of brain are widely studied. In such studies, accurate detection of neuronal electrical spikes from measured electrical-potential-data including noises is critically required. In addition, the on-line spike detection from the measured electrical signals is preferred to offline detection, especially for the application of the neuronal spike detection in brain-machine interface. In this study, we developed the novel and simple threshold-based-algorithm to detect neuronal electrical spikes, determining adequate threshold for spike detection, even though the frequency of spontaneous spikes and noises drastically changes. Using this novel method, numbers of detected spikes were improved to 96.4\% of correct number, while number of detected spikes were 91.8\% of correct number with previous method, suggesting that number of lost neuronalelectrical spikes decreased. In addition, we developed software to simultaneously perform the on-line spike detection and the recording electrical signals from 64-electrodes, which is convenient to control electrical devices according to the electrical activity of the neuronal network.
For development of bioinspired system or information processing system utilizing biological components, it is critical to understand the network electrical dynamics fundamental for information processing in living neuronal network. Basic activity such as spontaneous activity is considered to be basic unit composing internal states of the living neuronal network. Spontaneous activity is generated by mutual interaction between neuronal cells in brain and it is also observed in the cultured rat hippocampal neurons. The spontaneous activity has complex and dynamic spatiotemporal patterns. Therefore, it is likely that reproducibility of the electrical activity patterns, expressed after transient blockade of network activity, is not necessarily guaranteed. In this study, spontaneous activity in a dissociated neuronal culture was recorded by extracellular-potential-multisites-recording-system and we compared the activity patterns recorded before and after a transient pharmacological blockade of spontaneous activity. As a result, frequency of spontaneous activity was increased and temporal pattern of the activity became to be intermittent pattern. Modified temporal pattern of the network activity lasted for several hours and gradually recovered to the initial state. These results suggested that the equilibrium between neuronal activities was broken by a transient abolishment of spontaneous activity, and that the internal states in the system were changed. Thus, it is required to consider the dynamic features of electrical activity in living neuronal activity when we decode information from neuronal activity.
We discuss how to acquire logicality, orientation, and connectivity of action potential from living neuronal network in vitro. The action potential of rat hippocampal neurons organized into complex networks is detected in a culture dish by MED system carried with 64 planar microelectrodes. We propose an analysis method to acquire logicality and orientation of action potentials detected at three electrodes of the culture dish by fuzzy operators incorporating t-norm and tconorm operators. In addition, we propose a definition of connectivity of action potentials among electrodes by fuzzy inclusion degree. Finally, we discuss the relationship between logicality and connectivity of action potentials propagating with an orientation, e.g., transmission, absorption, and diffusion. We show that fuzzy concept is useful even in living neuronal network in vitro.
Since onomatopoeias are sensuous and ambiguous representation, information about these is often treated as multidimensional data. Recently, it has been analyzed based on visualization by mapping to a low dimensional space. In this study, we analyze the document data including onomatopoeias using self-organizing maps (SOM) that are widely used in the visualization of multivariate data. Further, the quantified information regarding onomatopoeia impression based on semantic orientations values is assigned to the map of the SOM. The effectiveness of the proposed analysis method was confirmed through visualization analysis for language resources on Twitter. The proposed analytical method can be useful as a tool to support the understanding of hidden linguistic properties of onomatopoeia.