Speech recognition using a robot has difficulties because the robot makes noise by itself. Especially, it is difficult for a small robot to reduce the internal noise by robot design. In this paper, two new methods are proposed that suppresses internal noise of small robots. These methods are based on spectral subtraction (SS). The difference of the proposed methods from the original SS is that the proposed methods use the estimated noise spectrum depending on the motion of the robot. One method, called MDSS, prepares the noise spectrums for all motions. Another method, called NPSS, predicts the noise spectrum from angular velocities of all joints of the robot using a neural network. From the results of the comparison among the original SS and the proposed methods, the proposed methods outperformed the conventional SS. The MDSS method gave good results when the noise within one motion was stationary, while the NPSS worked well even when the noise of the motion was non-stationary.

This paper presents the result of the evaluation experiment by　"passing　each other"　and　"positional prediction"　in simulated interactive situation between person　and mobile robot. We have manufactured four prototype robots based on four proposing　methods for preliminarily announcing and indicating to people the speed and direction　of movement of mobile robots moving on a plane. There was a significant difference be-tween when there was a preliminary-announcement and indication function and when it　was not even by each experiment, and the effect of the preliminary-announcement and indication of upcoming operation was declared. In addition, the features of each type of preliminary-announcement function and the effective usage were clarified. The method of announcing the state just after the present is effective when a person has to decide to which direction he should move to avoid collision instantaneously due to the feature that simple information can be quickly presented. The method of indicating the operations from the present to some future time continuously is effective when a person wants to avoid contact surely and accurately due to the feature that complex information can be accurately represented. Finally the knowledge and consideration obtained through the research and development are organized into an outlined.

This paper examines humans' cognitive characteristics toward home robots within the home environment over the course of a 16-day-long experiment. We adopted a semi-structured interview format to three families after the experiment and a conversation log with robots which have voice recognition system was analyzed for one family. For the conversation log, time-series analysis has been conducted in order to investigate the change of cognitive characteristics toward the home robot over the passage of time. The results indicate that(a)users get accustomed over time to the home robot and attachment emotions are elicited by living with the robot,(b)users emotionally communicate with the home robot by the touch interface, and(c)users gradually come to understand how the robot works and to construct a companion model for the robot as a sort of a mental model, which is an internal model for how something works.

In the near future, users of multi-function robots will have to read thick owner's manuals to use them. If users can use these robots without reading difficult manuals, it will improve user efficiency. We propose Action Sloping as a way for users to naturally understand a robot's function. This concept programs robots with feedback behavior that gradually changes in intensity as the user carries out given actions. By changing its feedback behavior in response to a user's actions, a robot encourages him or her to perform an action that will make the robot function. We conducted two experiments in which we programmed a robot dog with three patterns of feedback behavior based on the Action Sloping concept and two patterns not based on it as control conditions. The participants in the experiments tasked with identifying the robot's function, and the identification latency times were measured. The results showed that, as compared to the non-feedback conditions, only a chirping sound condition significantly assisted the participants in identifying the triggering action. These findings partially supported the effectiveness of the Action Sloping concept.

In this paper, we propose a concept of RobotMeme and peripheral mutual adaptation between human and robot based on the meme theory. Moreover, we carried out interaction experiment I and II between human and robot. From the results of the experiment I, it was proved that human tended to imitate robot unconsciously. It was also emerged that human tended to evaluate robot's original behavior (RobotMeme) more rational by imitating robot. Furthermore, from the results of the experiment II, it was emerged that robot's original behavior (RobotMeme) spread to other people successively through humans' imitation. From these results of the experiments in this paper, it is suggested that robot's original behavior (RobotMeme) will spread to human society. Henceforth, in order to realize peripheral mutual adaptation between human and robot, we plan to consider the process of meme spreading and create robot systems, also plan to carry out the interaction experiments between human and robot.

This paper proposes a novel model which enables a humanoid robot infant to discover and imitate other individual(e.g. human parent)focusing on the develop-mental process of human infants. In this work, the authors define "other individual" as an actor which can be predicted by a self-model. For modeling the developmental process of imitation ability, the following three approaches are employed. (i)Projection of a self-model for predicting other individual's actions.(ii)Mediation by a physical object between self and other individual.(iii)Introduction of infant imitation by parent. For creating the self-model of a robot, we apply Recurrent Neural Network with Parametric Bias (RNNPB) model which can learn the robot's body dynamics. For the other-model of a human, conventional hierarchical neural networks are attached to the RNNPB model as "conversion modules". Our target task is an "emulation" of moving an object. For evaluation of our model, human imitation experiments by the robot projecting its self-model were conducted. The results proved the adequacy of our model to enable the robot to recognize a human as an imitable being, and to imitate his object manipulations from his perspective.

In this paper, we proposes a natural deictic communication model between communication robots and humans utilizing pointing gesture and simple verbal cues with the combination of reference terms(kore, sore, are)and object information. The proposed model consists of interpretation processes and facilitation processes. Interpretation processes are the processes that enables humanoid robots to interprete human deictic expression to the one certain object in surrounding environment as well as expressing their intentional object toward humans utilizing a deictic expression. On the other hand, facilitation processes are unique and fundamental processes that facilitate the naturalness of deictic communication between robots and humans. With these two processes, the proposed model enables humanoid robots to have natural deictic communication with humans, utilizing pointing gestures and deictic cues. We developed a deictic communication system with communication robot "Robovie" based on the proposed model, and verified its effectiveness.

In order to facilitate multi-party conversations in a shared-virtual-space　voice chat environment, we propose a shared-virtual-space voice chat system with an automatic avatar's gaze direction control function using user utterance information, based on an avatar's gaze behavior model of turn-taking in multi-party conversation. In our gaze behavior model, a conversation was divided into three states; during-utterance, right-after-utterance, and silence. The avatar's gaze behaviors were controlled based on a probabilistic state transition model in each state. The developed system attained easy-to-use automatic gaze control without eye-tracking camera or manual operation.

We investigate cursor trajectory and kinematics features in a pointing task with mouse. Experiments are conducted to investigate cursor trajectory and kinematics features in a large and high resolution display. We analyze effects of target attributes(direction, distance, and size)on cursor trajectory using kinematics features. As a result, we found that the(1)cursor trajectory tends to be curved,(2)it is not affected by the target size, and(3)it is featured by the target direction and distance. These data suggest that visual feedback information based on the target size is used to modify the latter half of a cursor trajectory.

It is difficult for a lecturer to interact with audiences during presentations, unlike face-to-face communication. In human face-to-face communication, not only the verbal message but also the nonverbal behavior such as nodding and body motions is rhythmically related and mutually synchronized between speakers. It is desirable for people to be able to share embodied rhythms resulting from the entrainment between human speech and nonverbal actions and motions during communication. In this pa-per, a new speech-driven embodied entrainment device "InterVibrator" is proposed for supporting embodied interaction and communication during presentations. InterVibrator is a vibration device with a vibratory response equivalent to the nodding response and its response time to the speech input is similar to that of a listener. It is expected to support interactive communication by synchronizing the embodied rhythms using a vibratory response. Two kinds of InterVibrators having different shapes are developed for lecturers and audiences. The effectiveness of the device is confirmed by a sensory evaluation experiment in an actual presentation.

A pedestrian would find it difficult to steadily acquire his bearings justusing a Global Positioning System(GPS). This paper proposes a novel practical methodfor acquiring azimuth instantaneously with a new wearable GPS user segment that workseven when the user is stationary. This is suitable for timely decision making out-of-doors.Possible systems are also proposed using this method. The device using this method canwork both by itself and in combination;when multiple devices are connected, the azimuthlimitation capability is improved. Improvements due to scalability are quantitatively con-firmed with computer simulation.

This paper describes new electronic whiteboard management system us-ing IC-tags and networks. Handwriting data which written on electronic whiteboards istagged by using IC-tags automatically and data is sending to the management server.In Japan, teachers move the classroom at the class. Thus, teachers'handwritten datais saved without division in a usual electronic whiteboard. In this case, it is necessaryto search out handwriting data at the file name and the written date. Therefore, it isdifficult to use handwriting data by the next class. The other side, someone understandseasily the person who wrote the handwriting data in this paper'ssystem, because thesystem tags handwriting data automatically using IC-tags. Moreover, saved handwritingdata can be used from any classroom, because the system unitary manages handwritingdata by using networks. In addition, students can use these handwriting data for self-study. The systems set up in six classrooms, and the systems are starting use by a usualclass. In the future, the frame-less handwritten character recognition technology will beused to enhance this system.