Yuichiro Anzai

A new approach to spike sorting for multi-neuronal activities recorded with a tetrode-how ICA can be practical

Multi-neuronal recording with a tetrode is a powerful technique to reveal neuronal interactions in local circuits. However, it is difficult to detect precise spike timings among closely neighboring neurons because the spike waveforms of individual neurons overlap on the electrode when more than two neurons fire simultaneously. In addition, the spike waveforms of single neurons, especially in the presence of complex spikes, are often non-stationary. These problems limit the ability of ordinary spike sorting to sort multi-neuronal activities recorded using tetrodes into their single-neuron components. Though sorting with independent component analysis (ICA) can solve these problems, it has one serious limitation that the number of separated neurons must be less than the number of electrodes. Using a combination of ICA and the efficiency of ordinary spike sorting technique (k-means clustering), we developed an automatic procedure to solve the spike-overlapping and the non-stationarity problems with no limitation on the number of separated neurons. The results for the procedure applied to real multi-neuronal data demonstrated that some outliers which may be assigned to distinct clusters if ordinary spike-sorting methods were used can be identified as overlapping spikes, and that there are functional connections between a putative pyramidal neuron and its putative dendrite. These findings suggest that the combination of ICA and k-means clustering can provide insights into the precise nature of functional circuits among neurons, i.e. cell assemblies.

Humanlike conversation with gestures and verbal cues based on a three-layer attention-drawing model

When describing a physical object, we indicate which object by pointing and using reference terms, such as ‘this’ and ‘that’, to inform the listener quickly of an indicated objects location. Therefore, this research proposes using a three-layer attention-drawing model for humanoid robots that incorporates such gestures and verbal cues. The proposed three-layer model consists of three sub-models: the Reference Term Model (RTM); the Limit Distance Model (LDM); and the Object Property Model (OPM). The RTM selects an appropriate reference term for distance, based on a quantitative analysis of human behaviour. The LDM decides whether to use a property of the object, such as colour, as an additional term for distinguishing the object from its neighbours. The OPM determines which property should be used for this additional reference. Based on this concept, an attention-drawing system was developed for a communication robot named ‘Robovie’, and its effectiveness was tested.

Cognitive Control of Real‐Time Event‐Driven Systems*

The paper discusses the development of cognitive skills in control of real-time, event-driven systems with long time lag. Using simulators of ship steering, and based on thinking-aloud protocol experiments, it is shown that unskilled persons may devote most of the initial period of learning control skills to understanding the causal behavior of the human-machine interface. Experts employ goal-oriented strategies, planning from a relatively global view-point. A possible explanation for the development of cognitive strategies is presented to account for the gathering of information about the control interface, the acquisition of various kinds of strategies, and the learning of control skills for real-time, event-driven systems with long time lag. A computer simulation model is presented that merges cognitive and control theory models, and that simulates the process of strategy development.

InfoPoint: A Device that Provides a Uniform User Interface to Allow Appliances to Work Together over a Network

Abstract: This paper proposes a new hand-held device called “InfoPoint” that allows appliances to work together over a network. We have applied the idea of “drag-and-drop” operation as provided in the GUIs of PC and workstation desktop environment. InfoPoint provides a unified interface that gives different types of appliances “drag-and-drop”-like behaviour for the transfer of data. Moreover, it can transfer data from/to non-appliances such as pieces of paper. As a result, InfoPoint allows appliances to work together, in the real-world environment, in terms of data transfer. A prototype of InfoPoint has been implemented and several experimental applications have been investigated. InfoPoint has shown its applicability in a variety of circumstances. We believe that the idea proposed in this paper will be a significant technology in the network of the future.

Active interface for human-robot interaction

In the near future, robots used by human-like personal computers will appear in office or at home. In this paper, we call these robots personal robots. A personal robot can be thought as a small autonomous mobile robot. First, we discuss a user interface design for personal robots in the face-to-face situation. Second, we consider the features of personal robots and their environments, and propose a new user interface concept for a personal robot an active interface. To show the effectiveness of the active interface, we design and implement a speech dialogue system called Chaser for human-robot interaction based on the active interface.

An adaptive sensor network system for complex environments

This paper describes the design and implementation of a sensor network system that can adapt to node failures and changes in node positions. The objective of this system is to support human actions in typical living and working environments such as buildings, offices, and homes. The adaptive sensor network system we designed and implemented can estimate the spatial position of each node, can detect a connection or disconnection between a node and the network, and can translate a position-based address into the appropriate set of sensor nodes. The evaluation results show the possibility of the system adapting to dynamic network changes.

Classification of neuronal activities from tetrode recordings using independent component analysis

Abstract Classifying spike shapes in multi-unit recordings has been important to extract single neuronal activities from nervous tissue. Although several methods for this purpose have been developed, most of them have had limitations in their ability to decompose single unit activities. When more than two neurons generate action potentials simultaneously, it is difficult to identify the spikes because of the overlap of the spike waveforms. In this paper, we suggest a procedure that solves this problem using independent component analysis. By testing for the refractory period of spikes in each independent component, the proposed procedure is efficient for the decomposition of neuronal activities.

A Mobile Agent-Based Framework for Configurable Sensor Networks

This paper presents a new framework for self-configurable sensor networks. The framework enables sensor networks to be dynamically reconfigured to suit the requirements of applications and changes in environments detected by sensors. The framework provides a middleware system with an architecture that is structured as a collection of components organized in a five-layer hierarchy. Since these components are implemented as mobile agents and their layers have common interfaces, we can deploy the components at remote sensor nodes and dynamically replace them with other components designed for the same layer without affecting the rest of the nodes. Therefore, the framework provides a powerful approach for developing adaptive and application-specific software for sensor nodes easily. This paper describes the architecture of the framework and the implementation of a prototype, which currently uses Java as the implementation language along with a component description language. Two interesting applications are briefly described to demonstrate the utility and flexibility of this framework.

Solving large scale puzzles with neural networks

The n-queens problem is solved with Boltzmann machines and a depth-first search (DFS) algorithm. In large-scale problems, the Boltzmann machines found a solution much faster than the DFS. The 1000-queens problem was solved using an energy minimization technique. The polyomino puzzles were also solved with Boltzman machines and a DFS algorithm. In small-scale problems, the DFS solved these puzzles faster than Boltzmann machines. Using Gaussian machines, large-size polyomino puzzles were solved successfully. For example, 36 unique solutions were obtained for the 1000-queens problem, and 5*8, 6*10, and 8*8 sized difficult polyomino puzzles were solved.<<ETX>>

Embodied cooperative behaviors by an autonomous humanoid robot

Previous research works in robotics and cognitive science have reported that humans utilize embodied cooperative behaviors in communication, such as nodding in response to anothers it utterance and looking at a certain object in a certain direction as others look or point at. We have developed a humanoid robot that utilizes such an embodied cooperative behavior for natural communication in a route guidance situation. It obtains numerical data on a humans body movement via a motion capturing system and then autonomously selects appropriate cooperative embodiment units from 18 implemented units. Each unit realizes a certain cooperative embodiment behaviors such as eye-contact by using the motion capturing system as well. As a result of a subject experiment, we have verified the effectiveness of the embodied cooperative behaviors of the robot for reliable and sympathetic communication. Moreover, we analyzed how the auditory expression and the embodiment contributed to the effect.