Tadashi Tsubone

Application to robot control using brain function measurement by near-infrared spectroscopy

In recent years, study of brain computer interface (BCI) is conducted actively and many researches of implementation using electro encephalic gram (EEG) are reported. On the other hand, some realization of BCI based on near- infrared spectroscopy (NIRS) also had been reported. Since a measurement instrument for NIRS is comparatively small- scale and it can perform noninvasive measurements, NIRS is expected as one of useful tool in order to realize versatile BCIs. In this paper, the estimation method is shown the possibility of applications to the ON/OFF control of BCI by NIRS. We measured regional cerebral blood flow during tapping movement of the right hand by NIRS and we propose a method to quantitatively estimate start and end timing of movement by using a neural network.

Identification of moving limb using near infrared spectroscopic signals for brain activation

A method is described for classifying near-infrared spectroscopy (NIRS) signals measured for motor imagery and/or execution using the left or right hand. The measurement time intervals and the signal channels are used as features. The signals are discriminated using a support vector machine. Experiments demonstrated that this method has a higher generalization capability than a previous method for classifying NIRS signals. Testing of its ability to classify the signals according to whether they are for right- or left-hand motor imagery and/or movement demonstrated that its classification of NIRS signals satisfies the two-category classification problem. A promising application is to brain-computer interfaces, a potential communication tool for paralyzed individuals.

Relation between NIRS signal and motor capability

Brain activities of three subjects performing a right-hand tapping task were measured by near-infrared spec-troscopy (NIRS). In experiments, the hemoglobin concentration change in the subjects’ brains while they learned a new movement was analyzed. The results of these tests show that the channels covering the left primary motor cortex recorded a decreasing tendency in oxyHb when the subjects were repeating the tapping task. In contrast, the channels covering the supplementary motor cortex recorded an increasing tendency of oxyHb. Thease results suggest that the functional load on the brain decreases and the brain’s active domain changes during motor learning.

Estimation of the direction of arm force by using NIRS signals

In this study, we tried to discriminate the direction of arm force from hemoglobin concentration changes measured by near-infrared spectroscopy (NIRS). A self-organizing map (SOM) was used to classify the force direction information obtained from the NIRS signals. In a human subject experiment, the subjects were required to perform isometric arm movements in four directions. We investigated the feature quantities extracted from the time series data of the NIRS signals during the movement task. The feature vectors were used as the input vectors to the SOM. We tried to estimate the arm force direction by using a simple method based on the clusters given by the SOM. The results confirm that the direction of force is discriminable from the NIRS signal. In spite of the simple classification approach, the discrimination test yielded an average discrimination rate of 87.5 % for two directions. The experiment results suggest that NIRS signal must contain information related to the force directions.

Chaos and Periodic Oscillation of a Multirate Sampled-Data Dynamical System

In this paper, we consider a simple nonlinear system which consists of a chaotic system and multirate sample-hold controllers. The proposed system exhibits some stabilized Unstable Periodic Orbits which are embedded on the chaos attractor of the original chaotic system. We provide a condition to stabilize Unstable Periodic Orbits and its domain of attraction. Some theoretical results are verified in the experimental circuit.

Estimation algorithm of tapping movement by NIRS

In recent years, study of brain computer interface (BCI) is conducted actively and many researches of implementation using electro encephalic gram (EEG) are reported. On the other hand, near-infrared spectroscopy (NIRS) is noninvasive measurements, but there are many problems remaining unsolved and little is reported about the actual procedures or implementation examples when applying NIRS to realize BCI. We propose a method to quantitatively estimate start and end timing of movement by using a neural network. We measured regional cerebral blood flow during tapping movement of the right hand by NIRS. The following tendencies of total-Hb were observed. Hb increased within 10 s from the movement start time, decreased within 10 s from the movement end time. In this paper, the estimation method showed the possibility of applications to the ON/OFF control of BCI

Considering a learning model for communication based on reinforcement learning

Humans communicate with each other using language, sign language, gesture, etc. However, even if movement patterns that represent intentions have already been acquired, one should select those movement patterns that correspond to the desired intention to be communicated. We propose a model based on reinforcement learning for learning about the connections between intentions and movement patterns that are important for communication. We experimented that our model can acquire one-to-one mapping and also demonstrated communication between a human and a dog, using AIBO. As a result, proposed model suggested that can correctly learning for correspondence between intentions and movement patterns

Estimation of Force Motor Command to Control Robot by NIRS-Based BCI

We consider a possibility for estimating force amplitude and start and end timing of movements based on hemoglobin density by using near-infrared spectroscopy (NIRS). In first experiments, subjects carried out isometric movements of three levels of force amplitude in order to measure EMG, force amplitude and hemoglobin density, and these relationships were investigated. We confirmed strong correlations between these measurements. From these relationships we propose two estimation models; one is to estimate the EMG from hemoglobin density and the other is to estimate the force amplitude from the estimated EMG. We can construct estimation models with high performance by minimizing AIC. Second, we examines the estimation of start and end timing of tapping movement by using NIRS signal around pyramidal area. We show the analysis of regional cerebral blood flow during maximum tapping effort movement and the method to quantitatively estimate start and end timing of movement. Finally, we show an example of a BMI system applying estimation models to control an arm robot.

Estimation of force motor command for NIRS-based BMI

We consider a possibility for estimating EMG and force amplitude based on hemoglobin density. In experiments, subjects carried out isometric movements of three levels of force amplitude in order to measure EMG, force amplitude and hemoglobin density, and these relationships were investigated. We confirmed strong correlations between these measurements. From these relationships we propose two estimation models; one is to estimate the EMG from hemoglobin density and the other is to estimate the force amplitude from the estimated EMG. We can construct estimation models with high performance by minimizing AIC. Finally, we show an example of a BMI system applying estimation models to control an arm robot.

Possibility of reinforcement learning based on event-related potential

We applied event-related potential (ERP) to reinforcement signals that are equivalent to reward and punishment signals.We conducted an electroencephalogram (EEG) in which volunteers identified the success or failure of a task. We confirmed that there were differences in the EEG depending on whether the task was successful or not and suggested that ERP might be used as a reward of reinforcement leaning. We used a support vector machine (SVM) for recognizing the P300. We selected the feature vector in SVM that was composed of averages of each 50 ms for each of the six channels (C3,Cz,C4,P3,Pz,P4) for a total of 700 ms. We can suggest that reinforcement learning using P300 can be performed accurately.