While muscle fatigue appears to be a common phenomenon and experienced in everyday life, the mechanism of its sensation is elusive. In this study, we investigated the hypothesis that fatigue sensation is not only generated by physiological changes and sensory information signaling muscle metabolic state but also modified by information processing of sensorimotor system. In the experiment, subjects answered the change in the sense of fatigue during continuous cyclic finger movement with visual feedback. Results revealed that the muscle fatigue sensation was enhanced with delay of visual feedback of their finger movement, and that this sensation was not ascribed to simple misinterpretation of delay perception because both score tunings against delay were quite different. It is also unlikely that changes in movement intensity caused the enhanced sensation of fatigue because the delay affected neither muscle activity level nor movement phase. Accordingly, we suggest that prediction of motor output based on motor command generation is involved in sensation of muscle fatigue.

The stressful sensation of resistance induced by a delayed visual feedback is a major problem of visuomotor control via telecommunication network. In this study, we introduce our method which reduces this sense of resistance associated with delayed visual feedback by showing visual motion stimuli on the users' display. The method compensates for the delay by a new approach of modifying the users' motion, and is easier to implement than a predictor display; the traditional solution for delay. We will introduce the effects of our method on a letter writing task, and discuss the mechanism underlying the phenomena. Possibilities for modifying behavior under delayed conditions are also shown along with our future plans.

When stepping onto a stopped escalator, we often feel odd sensation which is never felt for stepping onto stairs. Previous studies suggested that implicit motor behavior (forward body sway and mistaken leg landing movement) driven by an endogenous motor program for 'moving escalator' during stepping onto stopped escalator is directly associated with the odd sensation. In this study, we investigated whether an exogenous postural change caused by visual motion lead to the odd sensation. Most of subjects reported strong odd sensation in initial several trials of stepping onto the stopped escalator, but that sensation monotonically decayed with successive trials. After sufficient number of trials, we applied different patterns of sudden visual motion during their trials. When a body sway was induced by visual motion during stepping onto an escalator mock-up, they reported a strong odd sensation. In addition, when a postural change induced by visual motion was similar to that occurred during the initial several trials for the stopped escalator without visual motion, subjects reported that the odd sensation induced by visual motion was similar to that felt for the stopped escalator. These results suggest that implicit or reflexive motor response itself is an essential trigger for generating odd sensation rather than the endogenous motor program.

An inverse dynamics based estimation for human articulatory motor control has been presented. As the plant for this control application, a dynamic model of human articulators consisting of lip soft tissue with related muscles and bones is developed. The continuum of lip soft tissue is represented as a discrete model approximation composed of networked point masses interconnected via viscoelastic elements. By activating an appropriately selected muscle set, the dynamic model performs a mimic of actual human speech motion, called pseudo-speech motion. Pseudo-speech motion is a forward dynamics simulation using the articulator dynamic model. As a mathematical representation of human speech acquisition, simulations of articulatory motor control based on estimation of muscle motor command are presented. Inverse dynamics driven iterative estimation makes the dynamic model put into practice an articulatory motion. The solution of this inverse dynamics problem is a set of activated muscle motor command as the foundation of induced articulatory motion. Correlation between model-based motor command estimations and EMG signals recorded during actual speech motions is discussed.

Previous studies have suggested that a gradually introduced visuomotor discrepancy results in more complete visuomotor adaptation for goal-directed reaching than a suddenly introduced one. However, it is unclear what mechanism allows such a complete adaptation. Here, we investigated the computational mechanism of the visuomotor adaptation for suddenly or gradually introduced visuomotor discrepancy. Subjects practiced a fast center-out reaching with the hand position cursor rotated around the start position. One group of subjects was suddenly exposed to the 60 deg cursor rotation, and that rotation was kept throughout the training. After the training, approximate straight hand trajectory was observed with an elongated reaction time. On the other hand, another group of subjects, who were exposed to the cursor rotations updated incrementally up to 60 deg, generated the curved trajectory without changes in the reaction time after the training. When the hand cursor was turned off, the reaching endpoint was away from the target position. Additionally, the analysis of the response to the visual perturbation reveals that the visual feedback gain increased after the training. Taken together, these results suggest that, in the visuomotor rotation task, the motor planning process adapts to reduce the sudden large error, and that the motor execution process, including the visual feedback controller, adapts to reduce the small error without changes in the motor planning process.

Previous studies have suggested that a gradually introduced visuomotor discrepancy results in more complete visuomotor adaptation for goal-directed reaching than a suddenly introduced one. However, it is unclear what mechanism allows such a complete adaptation. Here, we investigated the computational mechanism of the visuomotor adaptation for suddenly or gradually introduced visuomotor discrepancy. Subjects practiced a fast center-out reaching with the hand position cursor rotated around the start position. One group of subjects was suddenly exposed to the 60 deg cursor rotation, and that rotation was kept throughout the training. After the training, approximate straight hand trajectory was observed with an elongated reaction time. On the other hand, another group of subjects, who were exposed to the cursor rotations updated incrementally up to 60 deg, generated the curved trajectory without changes in the reaction time after the training. When the hand cursor was turned off, the reaching endpoint was away from the target position. Additionally, the analysis of the response to the visual perturbation reveals that the visual feedback gain increased after the training. Taken together, these results suggest that, in the visuomotor rotation task, the motor planning process adapts to reduce the sudden large error, and that the motor execution process, including the visual feedback controller, adapts to reduce the small error without changes in the motor planning process.

Somatosensory reflex during movement is known to be modulated according to given tasks and environments anticipatorily, before movement initiation. It is unclear, however, whether the reflex gain setting is completed before movement start or the gain setting is developed in relation to a target event even after movement initiation. Here we investigated temporal feature of anticipatory reflex gain modulation when subjects performed planar reaching movement with different force fields in the later part of movement. Arm reflex response was induced by a small perturbation at several timings before the arm encountered the force field, after movement start. We found clearer reflex gain modulation according to the force field directions as the perturbation was applied closer the force fields. An additional experiment further showed that the reflex modulation was developed in relation to the temporal distance to the force fields rather than the spatial distance. These findings suggest that somatosensory reflex gain is set in time domain for upcoming dynamical environment, independently of movement execution.

The inverse dynamics based motor control strategy in human articulatory motion has been investigated. As the plant for this control application, dynamic model of human articulators consisting of lip soft tissue with related muscles and bones is introduced. The continuum of lip soft tissue is represented as a discrete model approximation composed of networked point masses interconnected via viscoelastic elements. Stiffness of every viscoelastic element is adjusted to ensure the compatibility in static deformation between discrete model and its continuum prototype. Simulation of forward dynamics by activating ingeniously selected muscle set induces a mimic of actual human speech articulation, called pseudo-speech motion. As a mathematical description of the human speech acquisition, articulatory control simulation based on iterative estimation of muscle motor command is presented. Inverse dynamics driving iterative estimation lets the articulator dynamic model put into practice an articulatory motion, induced by the estimated muscle motor command.

In our daily life, we frequently perform coordinated movements using right and left hands and fingers. Such movements are important to explore coordinated control mechanisms between hemispheres. In this study, subjects perform two patterns forefinger movements (in-phase and anti-phase movements). In Exp.1, to explore the cross talk information between inter-finger control systems, we gradually changed the motion visual-feedback gain or the force field gain of left finger movement. We observed a strong inter-finger cooperation in the in-phase movements. And we found a significant cross talk effect in the kinematics of forefinger movements. In Exp.2, subjects perform in-phase movement tasks using sudden change of these gains with different distance between right and left hands. We found that the change in the right finger's movement amplitude is stronger in the 'near' condition than in the 'far' condition.

When we step in stopped escalator, we strongly feel odd sensation called "IWAKAN" in Japanese. Previous study argued that one could not suppress the motor program for "moving escalator" regardless of full awareness of escalator condition (i.e. escalator is stopped) and that the discrepancy between predicted and actual sensory feedback triggered an odd sensation. Furthermore, some could argue that structural nonuniformity of stopped escalator (i.e. the first step is shorter than other steps) caused clumsiness, and this clumsiness induced IWAKAN. We examined these possibilities by comparing the performance to stopped escalator with those to moving escalator and to the wooden stairs which duplicated the structure of stopped escalator. Our results did not support these possibilities. Specifically we correctly recognized the state of the escalator and adequately performed our action before stepping in stopped escalator. Once stepped in stopped escalator, drastic posture forward sway was observed. This sway was not observed in the performance to wooden stairs. These results suggested a new possibility which implied the motor program for moving escalator subconsciously emerge only after we stepped in stopped escalator. We also analyzed the participants' report of IWAKAN after stepping stopped escalator and wooden stairs. Participants felt no IWAKAN after stepped in wooden stairs. On the contrary, when stepped in stopped escalator, they felt IWAKAN, and we identified the kinematic index correlated with participants' report of IWAKAN.

During target reaching, manual motor responses with short latency can be caused by two kinds of visual stimuli, one is a target shift (TS), and anther is a large-field background motion (BM). To clarify the differences in processing mechanisms between the responses, we observed hand responses to these visual stimuli with a broad age range subjects (24-79yrs.) and examined effects of perceptual motion sensitivity on behavioral characteristics of the responses. The results show that the latency of TS response was significantly greater than that of the BM response. And the contrast sensitivity was correlated with the latency of BM response, but not with the latency of TS response. Amplitudes of both responses did not significantly correlate with the motion sensitivity or age. These different correlations can be explained by a two threshold model for the visuomotor processing. Considering the experimental results, we argue the idea that the rapid responses to the background motion and the target shift are generated by different computational mechanisms.

The spatial relationships among reaching objects and our body parts (arm, eye, head, etc) are quite important in producing arm movements. These relationships have been examined in many previous studies of voluntary movements from the view points of coordinate representation and those transformations, but it is still unclear how these relationships are represented and used in involuntary movements. To explore this issue, we examined the effects of the spatial relationships among gaze, reaching-target and background visual motion on the reflexive visuomotor response induced by a background visual motion, called Manual Following Response (MFR). The results showed that 1) the amplitude of the MFR monotonically decreased with the separation between gaze and reaching-target, and 2) the spatial relationship between reaching-target and background visual motion did not affect the MFR. Given these results, it is suggested that the spatial relationship between gaze and reaching-target modulates the reflexive visuomotor gain. We will explain this modulation by the gain field model of which maximum is obtained when gaze location is matched to the reaching target.

When we step in stopped escalator, we strongly feel odd sensation called "IWAKAN" in Japanese. Previous study argued that one could not suppress the motor program for "moving escalator" regardless of full awareness of escalator condition (i.e. escalator is stopped) and that the discrepancy between predicted and actual sensory feedback triggered an odd sensation. Furthermore, some could argue that structural nonuniformity of stopped escalator (i.e. the first step is shorter than other steps) caused clumsiness, and this clumsiness induced IWAKAN. We examined these possibilities by comparing the performance to stopped escalator with those to moving escalator and to the wooden stairs which duplicated the structure of stopped escalator. Our results did not support these possibilities. Specifically we correctly recognized the state of the escalator and adequately performed our action before stepping in stopped escalator. Once stepped in stopped escalator, drastic posture forward sway was observed. This sway was not observed in the performance to wooden stairs. These results suggested a new possibility which implied the motor program for moving escalator subconsciously emerge only after we stepped in stopped escalator. We also analyzed the participants' report of IWAKAN after stepping stopped escalator and wooden stairs. Participants felt no IWAKAN after stepped in wooden stairs. On the contrary, when stepped in stopped escalator, they felt IWAKAN, and we identified the kinematic index correlated with participants' report of IWAKAN.

During target reaching, manual motor responses with short latency can be caused by two kinds of visual stimuli, one is a target shift (TS), and anther is a large-field background motion (BM). To clarify the differences in processing mechanisms between the responses, we observed hand responses to these visual stimuli with a broad age range subjects (24-79yrs.) and examined effects of perceptual motion sensitivity on behavioral characteristics of the responses. The results show that the latency of TS response was significantly greater than that of the BM response. And the contrast sensitivity was correlated with the latency of BM response, but not with the latency of TS response. Amplitudes of both responses did not significantly correlate with the motion sensitivity or age. These different correlations can be explained by a two threshold model for the visuomotor processing. Considering the experimental results, we argue the idea that the rapid responses to the background motion and the target shift are generated by different computational mechanisms.

When the visual target is suddenly shifted during a reaching movement, we can adjust our hand movement to the shifted target position with short-latency (< 150ms). However, the computational mechanism to generate such on-line reaching adjustment is still unclear. In this study we investigated this mechanism from the viewpoint of visuomotor coordinate transformation. We observed the hand response to the target shift in one of eight radial directions. The experimental data show that the direction of the initial phase (150-200ms) of hand response acceleration was slightly biased from the corresponding target shift direction, whereas the direction of the late phase (250-300ms) was little biased. Additionally, when the reaching target shift was introduced by the stimulus without object motion, the response latency was significantly delayed and the directionally biased response was disappeared. These results suggest that the on-line reaching adjustment would be generated by two different mechanisms: a reflexive controller which is induced by visual motion with short-latency and generates spatially inaccurate response, and a voluntary controller which generates spatially accurate response with long-latency. We suggest that the reflexive reaching adjustment would be acted to generate the fast and accurate on-line reaching adjustment for the sudden target shift.

Muscle activating dynamic model and control strategy for human articulation has been presented. The model consists of upper and lower lips with surrounding soft tissues, combined by jaw. Tissues in this model are represented as a mathematical model composed of lumped point mass network, interlinked by viscoelastic elements. To emulate the continuum deformation of human facial tissues, adaptive element stiffness is derived and applied, which approximates forced deformation of equivalent continuum to discrete model. Also, to certify the incompressibility of human facial tissue continuum, artificial actuation force made of volumetric flux feedback is introduced. This force is made to be applied on the outer boundary surface of the model, suppressing net volumetric flux across outer boundary minimum. As a simplified simulation model for articulation acquisition by human being, gradient-based estimation method is applied to obtain motor command input for muscular force producing target articulatory motion.

One of the speech articulators, tongue, has elastic structure, and can largely deform its shape for altering sound resonance in vocal tract. Human can mimic articulatoy gestures of other persons by obtaining limited observation of lip and tongue movements. In this study, we formalized a control method of articulatory dynamics model by introducing an imitation learning strategy. We also performed computer simulation of articulatory movements in order to confirm the efficacy of the proposed method. Only by using a limited number of targets for a complex articulatory movement, we have succeeded to obtained a similar articulatory movement pattern after iterative estimations.

Several studies suggested that, during arm movement with unstable external dynamics in a particular direction, arm stiffness increase in unstable direction and change in muscle reflexive activity contribute to alter the arm stiffness. It is unclear, however, that whether such change in reflexive activity depend on the extent of external dynamics and/or where it is represented in central nervous system. To examine these issues, in this study, we measured muscle reflexive activity induced by a small mechanical perturbation before the arm encountered a constant force field with different direction, with or without single transcranial magnetic stimulation of motor cortex, which was applied 50ms before the perturbation to disturb muscle reflexive activity. As a result, we found that 1) muscle reflexive activity was pertinently changed according to the force field direction, and that 2) its change was diminished by applying transcranial magnetic stimulation, suggesting that reflexive sensorimotor processing was predictively and functionally modulated according to expected dynamical interaction and that motor cortex contributes its processing.

In this study, we examined a reflex coordination of human arm during hitting movement, which is a typical example of dynamical interaction with external objects. Subjects were asked to hit a virtual pack presented on a screen during acceleration or deceleration phase of horizontal right-to-left movement. The subjects felt a hitting-force in the most of the trials, but did not feel any hitting force in some trials while pack was visually bounced by hitting action. Surprisingly, in the no hitting-force trials, the extensor muscle showed sharp EMG response at the hitting moment before voluntary reaction whereas such EMG response was not shown when subject knew no hitting-force in advance. Additionally, reflex response of the extensor muscle to the perturbation in a heavy pack task was greater than that in a light pack task. These results suggest that in the hitting movement, sensorimotor gain is modulated in the interaction with external objects.

It has been recently reported that large-field visual motion during arm movement induces quick arm response having similar short latency with the arm response triggered by a target motion. To explore the mechanism of the quick response caused by visual motion, in this study, we examined the effect of image contrast and spatiotemporal frequency on the arm response. We found that 1) arm response rapidly increased as contrast increased in the range of 2 - 10% , while saturated in the range of over 10 %, and that 2) amplitude of the arm response increased as temporal frequency increased in the range of under 10 Hz, but decreased over 10 Hz, and that 3) arm response was maximal at a low spatial frequency (0.1c/deg). A part of these characteristics is quite similar with those of neuronal activities in the magnocellular layers of the LGN and in the area MT, suggesting contribution of these brain areas to this arm response. Further, the stimulus tuning functions, as well as short response latency, are similar to those of the reflexive eye response induced by a large field visual motion known as 'ocular following response (OFR)'. This suggests that the reflexive arm response induced visual motion is driven by similar mechanism to that of OFR.

This paper describes a robust realtime estimation method of time varying human multijoint arm viscoelasticity during movement. The general idea of the present method will be described as follows: 1) The time varying human multijoint arm dynamics is modeled by two factors, simplified musculoskeletal dynamics and uncertainty factor consisting signal noises and modeling error of a rigid body dynamics. Additionally, pseudo-random perturbation is supplied so that its frequency beyond that of the voluntary component of the movement. Utilizing this feature, perturbed component of kinetic data can be extracted by band-pass filter. 2) By using the proposed model, a robust Kalman filter is designed for identifying the multijoint viscoelasticity of the human arm. The uncertainty factor can be considered by using some aspects of prior information of the noise and error. The proposed method is evaluated by demonstrating several simulation results of estimating arm viscoelasticity during movements.

This paper describes a robust realtime estimation method of time varying human multijoint arm viscoelasticity during movement. The general idea of the present method will be described as follows: 1) The time varying human multijoint arm dynamics is modeled by two factors, simplified musculoskeletal dynamics and uncertainty factor consisting signal noises and modeling error of a rigid body dynamics. Additionally, pseudo-random perturbation is supplied so that its frequency beyond that of the voluntary component of the movement. Utilizing this feature, perturbed component of kinetic data can be extracted by band-pass filter. 2) By using the proposed model, a robust Kalman filter is designed for identifying the multijoint viscoelasticity of the human arm. The uncertainty factor can be considered by using some aspects of prior information of the noise and error. The proposed method is evaluated by demonstrating several simulation results of estimating arm viscoelasticity during movements.

Previous studies revealed that arm stiffness is altered according to environmental dynamics. It is not clear, however, whether such contextual change in stiffness is caused reactively by the interaction between arm and environment or whether it reflects a change of motor command by the CNS. To examine this issue, we measured arm stiffness by applying small perturbations just before the arm encountered an altered dynamics during transversal movements. The first portion of the movements was always unconstrained, but the second portion could be either a virtual valley (stable), a virtual ridge (unstable), or no constraint. As a result, it was found that stiffness just before the ridge decreased in stable direction, and that stiffness just before the valley increased in unstable direction. Additionally, in a ridge condition, reflexive electromyogram (EMG) response to the perturbations was equivalent to twofold its background EMG, suggesting that the CNS anticipatorily adjusts arm stiffness not only by changing muscle coordination but also by altering reflex innervations according to expected environmental dynamics.

It would be important to reveal how the central nervous system (CNS) regulates the mechanical impedance of musculoskeletal dynamics in interacting with physical environment. To examine this issue, we measured human multijoint arm stiffness by applying small perturbation just before the arm encountered an altered dynamics (valley, ridge, null force field) during two kind of longitudinal movements (proximal to distal, and distal to proximal positions). As a result, it was found the orientation of stiffness was greatly altered according to movement directions, and stiffness increased in the unstable direction of the external dynamics. Then we examined the effect of the reflex electromyogram (EMG) response against the perturbations on the joint torque change. The condition dependent variation of the joint stiffness due to the reflex was similar to that of the total joint stiffness. Additionally, reflex force represented in hand coordination was functionally altered according to the physical environments. These results suggest that stiffness modulation is greatly attributed to the reflex response which could be regulated by the CNS.

It would be important to reveal how the central nervous system (CNS) regulates the mechanical impedance of musculoskeletal dynamics in interacting with physical environment. To examine this issue, we measured human multijoint arm stiffness by applying small perturbation just before the arm encountered an altered dynamics (valley, ridge, null force field) during two kind of longitudinal movements (proximal to distal, and distal to proximal positions). As a result, it was found the orientation of stiffness was greatly altered according to movement directions, and stiffness increased in the unstable direction of the external dynamics. Then we examined the effect of the reflex electromyogram (EMG) response against the perturbations on the joint torque change. The condition dependent variation of the joint stiffness due to the reflex was similar to that of the total joint stiffness. Additionally, reflex force represented in hand coordination was functionally altered according to the physical environments. These results suggest that stiffness modulation is greatly attributed to the reflex response which could be regulated by the CNS.

骨格運動においては、関節にあるトルクを発生させるためにいくつもの可能な筋活動の組み合わせが存在し、その中から1つを選択している。腕の多関節運動のモデルとして、従来いくつかの最小原理(最適規範)モデルが提案されてきた。本研究では、肩水平面内の様々な方向に手先力を発生したときの各筋の活動変化から、各筋の筋活動優先方向を求め、実際の筋活動といくつかの最適原理によって予測した筋張力の変化とを優先方向を評価基準として比較した。その結果、筋ストレス二乗和最小規範がもっとも良く筋優先方向を近似し、また計測で観察された「姿勢による優先方向の変化」を再現した。この結果から、筋の大きさに応じて複数の筋に発生力を分散するメカニズムが示唆される。また本評価手法は、筋活動と筋張力の非線形性や試行によりばらつきが起る対抗筋活動の上昇などによる影響が低いため、筋活動の評価法として有用であると考える。

Our purpose is to clarify the control mechanism of the cooperated speech articulatory movements. In this study, we perturbed jaw movement associated with labial-mandibular coordination for/Φ/ and /p/ utterance. Labial distance was recovered quickly by the downward shift of the upper lip. This compensatory movement was caused by the change in passive stiffness covaried with the level of muscle activation which is regulated accoerding to articulatory tasks. Using an upper lip-jaw model, we estimated the relative stiffness change between upper lip and jaw for /Φ/ and /p/ utterance, and reproduced these passive compensatory movements by it computer simulation of using measured upper lip and jaw movements and estimated stiffness-change. These results suggest that the muscle stiffness plays the important role in compensatory articulation.

To investigate principles of muscle regulation, we measured variations of muscle activities during multijoint static force regulation tasks in a horizontal plane. The results show that activities of monoarticular muscles at the shoulder and elbow joints strongly covaried with the corresponding joint torque respectively, and that the activities of biarticular muscles were covaried to some extent with the elbow joint torque. In the light of muscle directional preferences, several minimum principles have been considered to describe the variation of muscle activities. The characteristics of muscle directional preferences have been nicely reproduced by minimizing sum of muscle stress squared, suggesting that this minimum principle is a dominant constraint of muscle coordination by the central nervous system.

Our purpose is to clarify the control mechanism of the cooperated speech articulatory movements. In this study, we perturbed jaw movement associated with labial-mandibular coordination for/Φ/ and /p/ utterance. Labial distance was recovered quickly by the downward shift of the upper lip. This compensatory movement was caused by the change in passive stiffness covaried with the level of muscle activation which is regulated accoerding to articulatory tasks. Using an upper lip-jaw model, we estimated the relative stiffness change between upper lip and jaw for /Φ/ and /p/ utterance, and reproduced these passive compensatory movements by it computer simulation of using measured upper lip and jaw movements and estimated stiffness-change. These results suggest that the muscle stiffness plays the important role in compensatory articulation.

It is well-known that the articulatory motion varies depending on speech tasks and that the articulatory organs are cooperated by the brain to realize desired utterances. In the previous studies, limited types of coarticulatory behaviors of lips and jaw were demonstrated in the experiments where instantaneous perturbations were introduced to the speech dynamical system. In this study, to investigate various coarticulatory behaviors organized by articulatory control mechanisms, we developed a jaw manipulandum which enable us to perturb jaw movements in open/close directions. We conducted jaw perturbation experiments during speech including labial fricative consonant /f/ and labial nasals consonant /m/, and during sustained utterances, and observed upper-lip compensatory movements in these conditions.

We recorded the activity of single neurons in the medial superior temporal (MST) area of the cortex, the dorsolateral pontine nucleus (DLPN), and the ventral parafloccular (VPFL) lobes of the cerebellum of alert monkeys during ocular following elicited by sudden movements of a large-field pattern. Their temporal firing patterns were analyzed by linear regression models using the eye movement and the retinal slip: a position, velocity and acceleration and a bias model. We found that for the most P-cells in the VPFL the reconstruted firing patterns using eye movement fitted the raw temporal firing patterns very well, regardless of dependent of the stimulus speeds or not. However, for neurons in the MST and DLPN, the raw temporal firing patterns could not be reconstructed well using the eye movement independent of the stimulus speeds. Their firing patterns were fitted by the reconstructed firing patterns using retinal slip at one stimulus speed better than those using eye movement. It is possible that the neurons in the MST and DLPN play roles in detecting the visual motion signal, and the information of their temporal firing patterns of all cells converge on P-cells in the VPFL, finally the temporal firing patterns of the P-cells represent the motor command utilized by the downstream structures for eliciting ocular following.

We recorded the activity of single neurons in the medial superior temporal (MST) area of the cortex, the dorsolateral pontine nucleus (DLPN), and the ventral parafloccular (VPFL) lobes of the cerebellum of alert monkeys during ocular following elicited by sudden movements of a large-field pattern. Their temporal firing patterns were analyzed by linear regression models using the eye movement and the retinal slip: a position, velocity and acceleration and a bias model. We found that for the most P-cells in the VPFL the reconstruted firing patterns using eye movement fitted the raw temporal firing patterns very well, regardless of dependent of the stimulus speeds or not. However, for neurons in the MST and DLPN, the raw temporal firing patterns could not be reconstructed well using the eye movement independent of the stimulus speeds. Their firing patterns were fitted by the reconstructed firing patterns using retinal slip at one stimulus speed better than those using eye movement. It is possible that the neurons in the MST and DLPN play roles in detecting the visual motion signal, and the information of their temporal firing patterns of all cells converge on P-cells in the VPFL, finally the temporal firing patterns of the P-cells represent the motor command utilized by the downstream structures for eliciting ocular following.

発声運動系は、所望の複雑な音響現象を実現するため、複数の発声器官を巧みに協調させていることが知られている。この協調制御のメカニズムを探ることは、音声生成のための脳情報処理を明らかにし、またそのモデルを構築する上で重要な課題である。本研究では、発声力学系に外部から変化を与えることによって起こる挙動変化を観察することにより、この協調運動のメカニズムを探る。今回、顎運動の開閉方向への摂動を可能とするマニピュランダムを開発し、下顎への摂動による唇、顎の挙動の変化を観察した。その結果、発話中に正確な隙間の形成を必要とする唇摩擦音/f/およびゆるやかな唇接触を伴う唇鼻音/m/、および/f/の持続音に対し、唇筋電図、唇顎運動などで協調動作を確認した。

We quantitatively compared actual point to point planer movements constrained to pass through a via-point to the movements predicted by three models ; 1)minimum hand jerk model, 2)minimum angle jerk model and 3)minimum torque change model. We defined the difference between the tangential velocity of unconstrained point to point movement and that of via-point movement as a distortion of tangential velocity. The distortion of tangential velocity varied with workspace and whether the position of the via-point was set near or far from body. Furthermore, we studied how learning effected distortion as the via-point varied in distance from the center midline separating the start and end points. If the via-point was at approximately the center between start and end point, then learning caused the tangential velocity profile to approach that found in unconstrained point to point movement. However, if the via-point was placed near the body, then the learned movements had a velocity profile that was dissimilar with that found in unconstrained point to point movements. This pattern of results is similar with that predicted by minimum torque change model.

We observed changes of muscle activation and hand trajectories during motor learning. Subjects were asked to learn planar reaching movement which included a via-point while EMG and position data were recorded. Arm stiffness was estimated from the sum of flexor muscle torque and extensor muscle torque reconstructed from EMG. Joint torque was also calculated from hand position data. As learning proceeded, muscle activation decreased, although, in some case, there was little change in the joint torque profile. The results suggest that learning enables us to realize similar trajectories with lower arm stiffness.

We observed changes of muscle activation and hand trajectories during motor learning. Subjects were asked to learn planar reaching movement which included a via-point while EMG and position data were recorded. Arm stiffness was estimated from the sum of flexor muscle torque and extensor muscle torque reconstructed from EMG. Joint torque was also calculated from hand position data. As learning proceeded, muscle activation decreased, although, in some case, there was little change in the joint torque profile. The results suggest that learning enables us to realize similar trajectories with lower arm stiffness.

We quantitatively compared actual point to point planer movements constrained to pass through a via-point to the movements predicted by three models ; 1)minimum hand jerk model, 2)minimum angle jerk model and 3)minimum torque change model. We defined the difference between the tangential velocity of unconstrained point to point movement and that of via-point movement as a distortion of tangential velocity. The distortion of tangential velocity varied with workspace and whether the position of the via-point was set near or far from body. Furthermore, we studied how learning effected distortion as the via-point varied in distance from the center midline separating the start and end points. If the via-point was at approximately the center between start and end point, then learning caused the tangential velocity profile to approach that found in unconstrained point to point movement. However, if the via-point was placed near the body, then the learned movements had a velocity profile that was dissimilar with that found in unconstrained point to point movements. This pattern of results is similar with that predicted by minimum torque change model.

We quantitatively and statistically compared actual point to point movements to movements predicted by three models; 1) minimum hand Jerk model, 2) minimum angle Jerk model and 3) minimum torque change model. Hand trajectories were measured in various initial and final positions located within a large workspace. These actual trajectories were compared with simulated trajectories for each model by using a curvature index and the sum of square difference between actual and simulated movement. Results confirmed that the minimum torque change model predicted the trajectories closest to the measured ones with respect to spatial and temporal aspects. The results of this experiment supported the hypothesis that hand traJectory is planned in intrinsic-dynamic space. Furthermore, it is considered that hand traJectory is planned to minimize the torque change which depends on change of viscosity.

The sensitivities of the geometrical characteristics of arm stiffness at hand during force control in the horizontal plane were investigated, which are indexed by shape and orientation of stiffness ellipse and characterized by arm posture and ratios between shoulder, elbow, and double-joint stiffness. At the distal hand position, as previously studied, geometrical characteristics, shape and orientation of stiffness ellipse, derived from arm stiffness matrix, were not much altered under different requested conditions. At the proximal hand position, those characteristics, however, were much altered. From theoretical analyses, it has been revealed that the shape and orientation of ellipse are insensitive to the changes of the joint stiffness ratios at the distal hand position, and are sensitive to those changes at the proximal hand position. In other words, slight changes of geometrical characteristics of stiffness ellipse do not directly indicate slight changes of joint stiffness ratios.

The sensitivities of the geometrical characteristics of arm stiffness at hand during force control in the horizontal plane were investigated, which are indexed by shape and orientation of stiffness ellipse and characterized by arm posture and ratios between shoulder, elbow, and double-joint stiffness. At the distal hand position, as previously studied, geometrical characteristics, shape and orientation of stiffness ellipse, derived from arm stiffness matrix, were not much altered under different requested conditions. At the proximal hand position, those characteristics, however, were much altered. From theoretial analyses, it has been revealed that the shape and orientation of ellipse are insensitive to the changes of the joint stiffness ratios at the distal hand position, and are sensitive to those changes at the proximal hand position. In other words, slight changes of geometrical characteristics of stiffness ellipse do not directly indicate slight changes of joint stiffness ratios.

Human arm joint-stiffness and hand-stiffness during static force control under several instructions were investigated to reveal the human motion control strategy. Although the magnitude of joint-stiffness and the ratio between the joint stiffness components were altered under different conditions, the directional characteristics of hand stiffness (represented by stiffness ellipse) were not changed dramatically. This is because the ratio of double joint stiffness relative to the single joint stiffness was restricted in the range insensitive to the change of the stiffness ellipse direction. Additionally, the directions of equilibrium positions during force control in several directions were greatly different from the directions of the corresponding forces at the hand position, and the distance between the equilibrium and actual positions did not proportionate to the force magnitude, indicating that the equilibrium position in Cartesian coordinate might be difficult to be used as the control parameter for force control.

We examined the relationship between human arm joint-stiffness and surface EMG signal during static force control in the horizontal plane. Joint-stiffness was roughly estimated from EMG signals recorded under (1) rest, (2) co-contraction, (3) force control, and (4) force control and co-contraction conditions by a six-muscle arm model using the least-square-error method. During the static force control condition, the contribution of single joint muscles to single joint stiffness increased whenever the contribution of double joint muscles to single joint stiffness increased; this made the single joint stiffness always higher than double joint stiffness. The EMG study also revealed that double joint muscles play an important role in changing static joint-stiffness.

We quantitatively and statistically compared actual point to point movements to movements predicted by three models; 1) minimum hand Jerk model, 2) minimum angle Jerk model and 3) minimum torque change model. Hand trajectories were measured in various initial and final positions located within a large workspace. These actual trajectories were compared with simulated trajectories for each model by using a curvature index and the sum of square difference between actual and simulated movement. Results confirmed that the minimum torque change model predicted the trajectories closest to the measured ones with respect to spatial and temporal aspects. The results of this experiment supported the hypothesis that hand traJectory is planned in intrinsic-dynamic space. Furthermore, it is considered that hand traJectory is planned to minimize the torque change which depends on change of viscosity.

We have studied the properties of simple-spike (SS) and complex-spike (CS) discharges of Purkinje cells (P-cells) in the cerebellar ventral paraflocculus in monkeys during ocular following responses (OFR), elicited by a sudden movement of a lange visual field. The temporal pattern of the SS and CS firing rate that occurred during OFR can be reconstructed by an inverse-dynamics representation of eye movements : the linear combination of position, velocity, and acceleration using a generalized linear model. By using the generalized linear model, we can estimate objectively the degree of model fit by the deviance function. The SS firing represented motor related signal rather than retinal slip. The CS firing rate represented both motor and retinal slip. The SS and CS firing did not represent acceleration of the retinal slip, but represented that of the eye movement. We compared the coefficients of acceleration, velocity, position of eye movement for CS and SS firing in each P-cell. It was revealed that there were close correlation between the coefficient of the CS and SS both in the absolute value and in the sign for each cell, suggesting a close relationship between the CS and SS responses.

We have studied the properties of simple-spike (SS) and complex-spike (CS) discharges of Purkinje cells (P-cells) in the cerebellar ventral paraflocculus in monkeys during ocular following responses (OFR), elicited by a sudden movement of a lange visual field. The temporal pattern of the SS and CS firing rate that occurred during OFR can be reconstructed by an inverse-dynamics representation of eye movements : the linear combination of position, velocity, and acceleration using a generalized linear model. By using the generalized linear model, we can estimate objectively the degree of model fit by the deviance function. The SS firing represented motor related signal rather than retinal slip. The CS firing rate represented both motor and retinal slip. The SS and CS firing did not represent acceleration of the retinal slip, but represented that of the eye movement. We compared the coefficients of acceleration, velocity, position of eye movement for CS and SS firing in each P-cell. It was revealed that there were close correlation between the coefficient of the CS and SS both in the absolute value and in the sign for each cell, suggesting a close relationship between the CS and SS responses.

In this paper, we propose a new learning scheme using feedback-error-learning for a Neural network model applied to an adaptive Nonlinear Feedback Controller (NNFC). This system uses a Conventional Feedback Controller (CFC) both as a usual feedback controller to guarantee global asymptotic stability and as a reference model of the response. The output of the conventional feedback controller is used as the error signal for a neural network model for adaptive nonlinear feedback control. The response of the controlled object follows the response of the reference model after the learning period.<BR>The convergence properties of this learning scheme are provided by using the averaged equation and the Liapunov method. This scheme was successfully applied to the control of an inverted pendulum by computer simulation. We also pointed out the relationship of this learning scheme to the cerebellum's posture and locomotion adaptive control mechanism in animals.

In this paper, we propose a new learning scheme using feedback-error-learning for a Neural network model applied to an adaptive Nonlinear Feedback Controller (NNFC). This system uses a Conventional Feedback Controller (CFC) both as a usual feedback controller to guarantee global asymptotic stability and as a reference model of the response. The output of the conventional feedback controller is used as the error signal for a neural network model for adaptive nonlinear feedback control. The response of the controlled object follows the response of the reference model after the learning period.<BR>The convergence properties of this learning scheme are provided by using the averaged equation and the Liapunov method. This scheme was successfully applied to the control of an inverted pendulum by computer simulation. We also pointed out the relationship of this learning scheme to the cerebellum's posture and locomotion adaptive control mechanism in animals.