Baseball pitching involves a complicated and rapid movement that has been investigated to prevent injuries and to improve the pitching performance. It is difficult to assess the pitching motion with high accuracy and a high sampling rate, due to the limitations of preexistent camera systems. The current camera system, however, has enough capacity to measure pitching motions with high accuracy and a high sampling rate. We have been diagnosing shoulder joint injuries caused by pitching. The patients (N = 939) felt pain during the pitching sequence as follows: top position (32.5%), maximum external rotation (27.2%), and ball release (14.5%). Hence the top position is one of the most important postures to investigate the mechanisms of shoulder joint injury in pitchers. There have been no studies that focused on the top position, however. The main purpose of this study was to develop a system to assess the pitching motion accurately. Another purpose was to estimate the instant of the top position and evaluate the kinematics of the shoulder and elbow joints. Pitching movement was assessed using a motion capture system (ProReflex MCU500, Qualisys Inc., Sweden) in a studio that has an official pitcher’s mound and home base. This system can record the positions of reflective markers at 500 Hz using seven CCD cameras. Thirty-two markers were mounted on the joints and body landmarks of each subject. Two markers were mounted on the ball. The pitching motions of eleven subjects were assessed, after a period for warm up. Kinematics parameters were calculated using three-axis gyroscopic Euler angle. The instant of the top position was observed for all subjects before the lead foot touched the ground. The interval from the top position to ball release was 0.242 ± 0.0438 [s] (n = 11). The subjects were divided into two groups by the type of posture at the instant of the top position, as follows: internal rotation group (n = 5), 11.8 ± 6.08 degrees, and external rotation group (n = 6), 38.1 ± 19.97. Other kinematics parameters at the top position were adduction of the shoulder at 74.2 ± 19.84 degrees, horizontal adduction of the shoulder at 37.3 ± 14.10 degrees, and extension of the elbow at 92.1 ± 21.63 degrees. The timing and posture of the estimated top position were almost the same as those of the conventional top position. From the top position to lead foot contact on the pitching sequence, there were three patterns of elbow leading. The three patterns did not depend upon experience. We interpreted them as individual variations.
The aim of this study was to examine the effect of aging on body reactions to perturbation during walking. The subject group comprised 15 young and 15 elderly individuals. Perturbation was produced by abruptly decelerating one side of the walking-belt of a two-belt treadmill for 500 ms during walking. Each subject received 30 perturbations during a 10-minute walking period. Muscle activity in the lower limb muscles and acceleration at the pelvis were recorded using surface electromyographs and accelerometers. Peak values of acceleration following perturbation were significantly greater than in normal walking. Posterior peak values of acceleration appeared prior to anterior peak values of acceleration following perturbation. These outcomes seem to demonstrate that external forces by perturbations act on the center of gravity of the whole body and that external forces work from the rear to the front. In addition, the tibialis anterior muscles demonstrated bilaterally reduced latency of muscle reactions. Latency of the tibialis anterior on the stimulation side was significantly shorter than that on the non-stimulation side. This suggests that perturbation in one leg triggers a righting reaction of that leg within the first 120 ms, facilitating the subsequent stepping reaction of the contralateral leg. Comparisons according to age demonstrated that both the appearance of anterior peak values of acceleration and latency of the tibialis anterior on the non-stimulation side were significantly delayed after perturbation for older subjects. These results suggest that the capacity for postural adjustment during stepping might be deteriorated in elderly individuals.
Electro-stimulation of muscles is utilized in various fields, such as functional electric stimulation (FES), therapeutic electric stimulation (TES), chronaxie testing of muscle, etc. The chronaxie is examined to estimate the relative excitability of muscles in clinical practice. ChronaxTM (OG Giken Co., Ltd., CX-3) is a commercial instrument for this type of diagnosis. However, the time required to obtain a strength-duration curve and the chronaxie is relatively long. It is also difficult to get reproducible measurement results because the start of muscle vibration is detected visually and subjectively. We previously developed a computer-based chronaxie meter. It consisted of a notebook-type computer, an electrical interface circuit, and a set of stimulating electrode and accelerometer. The motor point of a muscle was stimulated with negative rectangular pulses. As the muscle vibration caused by stimulation was detected using an accelerometer, the measurement reproducibility of the meter became better. The measurement time also decreased, compared with that of the ChronaxTM. The application of our instrument to many subjects and muscles was difficult for two reasons: a) individual characteristics of muscles; and b) heart synchronous pulsatile noise was superimposed on the muscle vibrations. As a consequence, our former instrument was inadequate because of the long measurement time required, and the accuracy was not sufficient for clinical use. In the present study, the former chronaxie meter was improved for an increase in measurement accuracy and a decrease in measurement time. The improved chronaxie measurement system is characterized by a measurement/control programming design and a transducer for detecting a pulsatile blood flow, newly appended to the former meter. In order to apply this system to many kinds of muscles, two characteristic frequencies and the rheobase current of the S-D curve of the muscle are examined prior to the usual S-D curve measurement. In order to avoid muscle vibration with superposed cardiac synchronous pulsatile noise, the muscle is electrically stimulated just after the onset of the systolic phase of the photoplethysmogram, measured transdigitally. With this trigger method, the muscle vibration can be detected free from noise. The measurement time for an S-D curve has decreased to approximately 50%, compared with the ChronaxTM. The dispersion in the obtained chronaxies of ten measurements has become 3.3%, and in the rheobases, 1.3% (3.8% and 2.3%, respectively in the former meter). The improved measurement system will be applicable to many subjects and kinds of muscles in clinical examination.
The power spectrum is well known to shift to lower frequencies and increased amplitude with progressing of fatigue. The increase in amplitude depends on motor unit recruitment. The frequency shift depends on the reduction in muscle fiber conduction velocity. Neuromuscular function was evaluated noninvasively by observing electromyography (EMG) during fatiguing contraction. We studied the effects of factors such as exercise, aging, and short-term immobilization on EMG during fatiguing contraction and evaluated the feasibility of EMG as a noninvasive index for determining neuromuscular function. Subjects exerted maximum voluntary contraction (MVC) isometrically and conducted 60% MVC fatiguing contraction. Surface EMG was detected using multichannel electrodes, and EMG variables such as average rectified value (ARV), mean power frequency (MPF), median frequency (MDF), and muscle fiber conduction velocity (MFCV) were calculated. We studied the relationship between EMG during fatiguing contraction and muscle performance, such as MVC and endurance time. We studied the effects of fatiguing exercise on EMG using knee flexion in 14 men aged 21 to 29 years (mean: 24.9 years), divided based on twitch interpolation into high and low voluntary activation groups. Four sedentary subjects were included in the low-activation group. MVC was significantly greater and endurance significantly shorter in the high voluntary activation group. Changes in ARV and MPF calculated from EMG detected in the vastus lateralis muscle during fatigue were also smaller in the high voluntary activation group. EMG results during fatigue contraction suggest that the low voluntary activation group, which included sedentary subjects, has low metabolic capacity in the muscle but low voluntary activation during MVC. Endurance was thus prolonged during the fatigue task. We examined the effects of aging on EMG during fatigue in dorsiflexion of the ankle joint in 25 women aged 20 to 77 years (mean: 47.1 years) and divided by age into older (66 to 77 years) and younger (20 to 24 years) groups. MVC was significantly smaller, and ARV, MDF, and the MFCV calculated from surface EMG detected in the tibialis anterior muscle during MVC were lower in the older group than in the younger. A similar trend was seen in the change in EMG during the fatigue task. These results suggest that older subjects have low motor unit activation and that their muscle fiber consists mainly of slow-twitch fibers. We also studied the effects of short-term immobilization on EMG during fatigue contraction, using abduction of the index finger in 10 men aged 20 to 29 years (mean: 24.4 years). MVC decreased and endurance was shortened when the finger was immobilized in a cast for 1 week. The decrease in MDF calculated from surface EMG detected in the first dorsal interosseous muscle was smaller, but the rate of change was greater after immobilization than before. These results suggest that the decrease in metabolic capacity caused by short-term immobilization led to a decline in performance factors such as MVC and endurance. These results further suggest that EMG during fatigue depends on individual factors, such as exercise, aging, and immobilization, reflecting the muscle fiber type and the degree of motor unit activation. We propose that EMG be applied during fatigue as a noninvasive EMG biopsy.
This study focused on stabilizing the sensations elicited by electrical stimulation applied to the skin through surface electrodes, a technique widely used in sensory feedback research. The relationship between absolute threshold (the minimum stimulus amplitude that causes sensation) and the parameters of an electrical equivalent circuit for skin impedance was examined with neurologically intact subjects. Electric current pulse stimulation at 100 Hz frequency and pulse widths between 0.08 and 0.5 ms was used to elicit a cutaneous sensation. The electrical equivalent circuit for skin impedance consisted of resistance Rp and capacitance Cp connected in parallel, plus resistance Rs connected in series to both. First, we estimated the parameters of the equivalent circuit by the skin impedance measured with sinusoidal signals (Rps and Cps), and we examined their relevance to the absolute threshold. For Ag-AgCl electrodes used with electroconductive paste, variations over time of the frequency characteristics of Rps and Cps were found to be similar to the changes caused by repeatedly stripping the skin. Values of Cps and Rps at 10 Hz were found to relate to the absolute threshold with high correlation coefficients. For a solid-gel Ag-AgCl electrode, changes of absolute threshold, Rps and Cps caused by increasing size of the solid-gel electrode were similar to variations over time obtained using electrodes with electroconductive paste. Correlations between the threshold and values of Rps and Cps at 10 Hz were also observed. Next, we estimated the parameters of the equivalent circuit by measuring stimulus pulse current and voltage waves of square-wave stimulus pulses using solid-gel electrodes (Rps and Cpr). We then examined their relevance to the absolute threshold. The absolute threshold increased with repeated electrical stimulation to the skin and decreased after adequate rest with no electrical stimulation. The value of Rpr decreased with repeated electrical stimulation and increased after rest, while the value of Cpr did not change. These variations became minor after preparation of the skin for removing the stratum corneum. A significant correlation between Rpr and the absolute threshold was observed. These results show that electrical stimulation to the stratum corneum affects both absolute threshold and skin impedance; they also indicate that real-time evaluation is important for stabilizing cutaneous sensation elicited by electrical stimulation. The relevance of Rp to the threshold was shown in both methods that were used to estimate the parameters. Possible reasons for decreasing Rp were an increase of ionic conductivity due to osmosis of electroconductive paste into the stratum corneum or an increase of electric charge in the stratum corneum due to charge transfer by electrical stimulation through the skin. The increase of the threshold amplitude was considered to be the result of broadened current path due to increases of ionic or electric conductivities in the stratum corneum. It is necessary to clarify the mechanism of the relationship between the threshold and impedance parameters in order to develop a technique for stabilizing electrocutaneous sensation.
Palatolingual contact stress (pressure) and pattern are indispensable parameters in investigation of the dynamic properties of human tongue motion during palatal consonant phonation. Palatolingual contact can be detected by dynamic palatometry, a practical technique using electrodes imbedded in a thin palatal plate that adheres to the hard palate. Dynamic palatometry using the electrical impedance of palatolingual contact detects binary palatolingual contact patterns, but it is not capable of measuring palatolingual contact pressure. In this paper, we first present a new type of cantilevered force-sensor mounted on a palatal plate for high-sensitivity, high-accuracy measurement of palatolingual contact pressure and tongue force. The system consists of three components: a cantilevered force-sensor mounted on a palatal plate, a multi-channel amplifier to supplement the strain gauge, and a computer for data acquisition and signal processing. The cantilevered force sensor, which consists of a strain gauge and a cantilever, is mounted on a palatal plate attached to the hard palate. The force sensor is 3.0 mm wide, 5.5 mm long, and 1.1 mm thick. Its measurement error is less than 2.8%. When a load of 5 gf applied to the protuberance of the force sensor is instantaneously removed, the time required for reducing the output level from 90% to 10% is 0.3 ms. The palatal plate is a thin plastic plate shaped in such a way that it attaches tightly to the subject’s hard palate. Multiple force sensors can be mounted at arbitrary positions on the palatal plate, and the palatolingual contact pressure and tongue force can be determined using such multiple sensors. Second, we propose an algorithm for estimating tongue force based on palatolingual contact pressure. The palatal regions are divided into triangular elements with vertices located at the three-dimensional positions of the force sensors, using the Delaunay triangulation method. Palatolingual contact pressure at arbitrary positions within the triangular elements of the hard palate is estimated using linear interpolation, based on measured values and positions of the force sensors. Tongue force can be calculated via surface integration over the distribution of palatolingual pressure values in the palatal region. To verify the sensor’s function, we simultaneously measured the weight of a tongue-like elastic test body using an electronic scale and our force-sensor palatal plate. The results indicated that the accuracy of tongue force measurements using the force-sensor palatal plate is 4.7%. The palatolingual contact pressures and patterns of five adult male subjects who had no history of speech disorder were measured during Japanese /CV/ (C=/t, d, n, s/ and V=/a, i, u, e, o/) phonation. The proposed system allows direct observation of the dynamic aspects of palatolingual contact pressure and tongue force during the phonation of consonants.
The aim of this study was to develop new eyeglass frames that fit well on most of the target population, using our original analysis method for three-dimensional human body forms. The facial region for eyeglass frame design was from below the nose (subnasal area) to above the brow in the vertical direction, and from the tip of the nose (pronasal area) to the back of the ear in the anteroposterior direction. Face forms of 56 Japanese males were captured as plaster models because it was necessary to obtain the complete surface form, including the back of the ear. The face was represented as a three-dimensional shape model consisting of 211 data points (366 polygons) based on 61 landmarks. These landmarks were defined by anatomical/anthropological information and industrial design information for fitting eyeglass frames. Shape distances among individuals were calculated from the models for 56 × 56 pairs. A shape distribution map was calculated by analyzing the distance matrix using a multi-dimensional scaling method. The first dimension was interpreted to represent the depth of the face and the nose bridge, and the second dimension was related to the breadth of the face, height of the orbit, height of the nose, and facial flatness. Japanese male faces were classified into four groups based on the distribution map for purposes of the marketing, and the average face form was calculated for each of the four groups. A new eyeglass frame was designed by CAD to fit each of the imported average face forms. Prototype eyeglass frames were adjusted using the actual average face models with a special soft material attached to the regions around the nose bridge and the upper ear base. These actual average face models were developed from the physical average forms using laser lithography. The soft material, made of urethane foam and silicone skin, had material properties very close to those of the human skin. A fitting test of the new eyeglass frames was conducted for 38 male subjects. The four types of new eyeglass frames and one conventional frame were used for the experiment. Each subject put on the frames in random order and made a sensory evaluation of the fit of each frame. The compression force of the temple of the frame and the relative movements between the eyeglass frame and the face were measured for each frame. As a result, it was confirmed that the new eyeglass frame of the correct type for each subject was designed for small compression force (half of the conventional level) and minimal slip. The score for fitting comfort was highest for the new frame of the correct type, as selected theoretically. Subjects who had larger (broader or deeper) faces and usually wore eyeglasses, however, tended to prefer a tighter frame to the new frame of the correct type.
Two positron emission tomography (PET) studies were performed on 18 normal volunteers to investigate regional cortical and subcortical activation induced by the repetitive lifting-holding-replacing of an object using a precision grip between the index finger and thumb. A three-dimensional PET was performed using H215O as a tracer. The first study was aimed to investigate the brain areas involved in a grasp-lift-hold-replace action and the effect of object weight. Data were obtained for 10 subjects under three object weight conditions (4, 200, and 600 g) and a rest condition. Grip and lift forces on a similar object and the activity of selected muscles in the hand, arm, and shoulder were also recorded in separate lifting trials. An external auditory beep sounded every 3.5 seconds to initiate the lifting action. The subjects repeated the lifting action 20-22 times during each PET scanning. A comparison between all movement conditions and the rest condition revealed significant activation of the primary motor (M1), primary sensory (S1), dorsocaudal premotor (PM), caudal supplementary motor (SMA), and cingulate motor (CMA) cortices contralateral to the hand used. On the ipsilateral side, activation of the M1, caudal SMA, and inferior parietal (BA40) was found. In the subcortical areas, the hemispheres of the bilateral cerebellum, left basal ganglia, and thalamus were activated. Behavioral adaptation to a heavier object weight was revealed in a nearly proportional increase of both grip and lift forces, and a higher level of hand and arm muscle activities. An increase in the rCBF associated with these changes was noted in several cortical and subcortical areas. Consistent object-weight-dependent activation, however, was observed only in the M1/S1 contralateral to the hand used and to some extent in the ipsilateral cerebellum. The second study was aimed at investigating the brain areas involved in each of the preparatory and execution phases for lifting of an object. In addition, the difference in brain activation between right (dominant) and left (non-dominant) hands was also investigated. Eight subjects were scanned under four conditions. The first condition was preparing and executing the task by the right hand, and the second was preparing only. The third condition was preparing and executing the task by the left hand, and the fourth was resting as a control condition. The subjects were cued for preparation for grasp by a tape-recorded voice saying “prepare, ” which was followed by a beep at some interval (4-8 seconds). For the execute condition, the subjects grasped and lifted the object at the beep sound, while for the prepare-only condition, the subjects stopped the preparatory state. This was repeated 10 or 11 times during each PET scanning. For the prepare and execute condition, results were similar to the activated areas for lifting action in the first study. BA40, however, was activated in the left side this time. For the prepare-only condition, after subtracting the resting condition, left SMA and S1 were the activated areas, and for the execute-only condition (the prepare-execute condition minus the prepare-only condition), left M1 and thalamus, and right cerebellum hemisphere were activated. When the task was performed by the left hand, there was nearly a mirror image of the activated areas seen in the right-hand lifting condition. These findings provided fundamental data for understanding the neural network governing control of fingers in grip-lift-replacement actions.
The goal of our studies is to examine the mechanism of the muscle output of the shoulder joint. In this paper we compared and quantified electromyographic muscle activities and the onset order of the rotator cuff and shoulder superficial muscles in relation to the change in concentric torque generated by performing shoulder external rotation. Twenty-four healthy male subjects (mean age 23.3) without any history of shoulder injury participated in the experiments. A Cybex 770 dynamometer was used to control the load and the angle of motion; the range of loads was 0-18 Nm, the range of motion was 75-0 degrees of internal rotation, and the angular velocity of the movements was fixed at 15 deg/sec. Electromyographs (EMGs) were recorded using surface electrodes placed on the shoulder superficial muscles and intramuscular wire electrodes from the rotator cuff. The raw EMG signals for each load were separated into five equal parts (each covering 15 degrees), which were then separately rectified and integrated. The integrated EMG data, which was normalized against data for maximum voluntary contraction, was compared for each load and each angle condition. The results of this experiment clearly showed that the supraspinatus (SSP), middle deltoid (MD), and middle trapezius (MT), as well as infraspinatus (ISP), teres minor (TM), and posterior deltoid (PD) muscles took part in external rotational movements. The EMG-onsets of the rotator cuff were generally faster, and their muscle activities were greater, than in the superficial muscles. For each muscle, however, the degree of participation was markedly dependent on the load and angle of motion. The EMG-onset of the ISP was the fastest, and the muscle activity of the ISP was the highest of all muscles, irrespective of load or angle of motion. It is thus certain that the ISP plays an important role as both a stabilizer and a mover during external rotational movement. At low loads, the TM demonstrated a pattern of muscle activity similar to that of the ISP, but it was notable that muscle activity did not increase for loads greater than 30-40% MVC. On the other hand, the muscle activity of the SSP was lower than that of the TM in the area approaching full internal rotation and at low loads. As the load increased, however, muscle activity of the SSP became higher than that of the TM. The TM and the SSP, therefore, may have different functional roles depending on the situation; both muscles may contribute to external rotational movements as the supporting mover. The muscle activity of the MD was higher than that of the PD in the area approaching full internal rotation. Clearly, not only the PD but the MD also takes part in external rotational movements, depending on the angle of motion. The EMG-on-set of the MT was faster, and the muscle activity of the MT at low loads was higher, than seen in other superficial muscles. Moreover, the activity of the MT increased with increasing load and degree of external rotation. These results suggest that the MT assists the mover muscles by anchoring the scapula at low loads but adducting it at high loads and in the area of external rotation.
It is well known that contracting muscles have elastic and viscous properties, which change widely with the level of muscle activation. Zeffiro reported that the torque generated by contraction of the triceps brachii muscle in a monkey increased with an increase in elbow angle, just as does an elastic spring. Recently we have investigated torque-angle relations in flexor and extensor muscles of the human elbow under static conditions. The properties of the extensor were similar to Zeffiro’s results. The properties of the flexor, however, did not show a simple elastic property; torque increased and then decreased with increasing elbow angle. The viscous property of the muscle could be explained in terms of a force-velocity relation. We estimated the torque-angular velocity relation in elbow extensor muscles; torque decreased with increasing angular velocity of extension, which was in agreement with common force-velocity relations of the muscle. There were few previous investigations of torque-angle relations of the flexor muscle in voluntary flexion movements. The relation would be a key for understanding the control mechanism of upper arm posture. The purpose of the present study was to obtain and to examine the relations with the constant muscle activation. It is almost impossible for the subject to maintain constant muscle activation during flexion movements. We have utilized a new artificial network technique to overcome this difficulty. In the present study, the torque-angle relations of the elbow flexor muscles showed very fascinating aspects; the stiffness (torque/angle) was negative at the zero velocity and positive at non-zero velocity of flexion. The experiments were performed with three normal subjects (male, aged 22-25). The task was isovelocity flexion of the elbow joint in a horizontal plane at the height of the shoulder. The subject was asked to hold the elbow joint at a fully extended position (elbow angle was almost zero) against the load torque, and then to flex the elbow joint at a constant velocity to about 120 degrees. In the holding experiments, he was asked to hold the forearm at the desired angle against the load torque. Surface electromyograms (EMGs), elbow joint angle, and torque were measured. Applied torques were approximately 0, 5, 10, and 15% of MVC (maximum voluntary contraction at an angle of about 90 degrees). The flexion velocities were 30, 60, and 90 deg/s. The measurements were repeated at least 10 times for one experimental condition. EMGs were recorded from six muscles: brachialis, caput longum bicipitis brachii, caput breve bicipitis brachii, brachioradialis, caput laterale triceps brachii, and caput longum triceps brachii. A pair of Ag-AgCl surface electrodes (10 mmφ) were put on the skin over each muscle. EMGs were full-wave rectified and then low-pass filtered with a cutoff frequency of 35 Hz. In order to obtain IEMG, this filtered signal was further running averaged over the time span during which the joint angle changed approximately 0.5 degree. Note that in the present study, steady state behaviors were examined. A three-layer artificial neural network was constructed with inputs of the elbow joint angle, flexion velocity, and six-channel IEMGs, and the output was elbow joint torque. The activation function of the input units was linear, and that of the hidden and output units was sigmoid. The appropriate number of hidden units was determined by varying both the number and the initial connection weight with back propagation learning. (View PDF for the rest of the abstract.)
The purposes of this study were to assess differences among the index, middle, ring, and little fingers in dynamic motor function, using single-finger-tapping and double-finger-tapping tasks, and to investigate the effect of long-term training on dynamic motor function, using trained pianists. Four miniature strain gauge force transducers were used to measure the external force generated by each finger during tapping. Twelve healthy male adults and three female adult pianists tapped force transducers for 9 seconds as fast as possible, using their dominant right hands, while the other fingers contacted with the force transducers. In the single-finger tapping, we asked subjects to tap using each finger: index, middle, ring, and little finger. In the double-finger tapping, subjects performed alternating movements of a pair of fingers. A comparison among the four fingers in single-finger tapping revealed that the index and middle fingers attained significantly faster tapping, indicating the superiority of these two fingers over the other two fingers in terms of dynamic motor function. Changes in isometric forces generated by the non-tapping fingers accompanied the tapping finger movements; the correlation between tapping and non-tapping fingers was greater for the ring finger than for the other fingers. This suggests that the ring finger is more dependent on the movements of other fingers, especially neighboring fingers. Differences among the fingers in dynamic motor function could result from anatomical and neurophysiological factors. During two-finger movement, the dynamic motor function of each finger was affected by the combination of fingers. Finger combinations that are less used in ordinary life had lower tapping speeds than the other combinations. For pianists, tapping rates of the little and ring fingers in single-finger tapping were similar to those of the index and middle fingers. There was less accompanying force by non-tapping fingers of the pianists than of non-pianists. In double-finger tapping, tapping rates for any combination of fingers was nearly the same for the pianists. These suggested that long-term training could alter dynamic motor function of the fingers.
Demand for fine human function modeling methodology is rising with the popularization of assistive devices. Systems engineering based research that delves into the functional cooperative relationship of the digits, hand, and arm is needed for design of substitutive mechanisms and for their control of the human body. The optimal tool for such research is the design process of a sensor-based robotic hand-arm system. This paper discusses the research issues and our proposed strategy. Our research goal is to develop a sensory-controlled mechanical system for performing versatile human-like prehension. As a design concept, we propose an effective model extracted from functional analysis of the upper limb. The key assumption in categorizing hand behavior is the arm’s driving function. Without proper integration of the two, the hand function can be neither analyzed nor assembled. Our approach uses this assumption as its base; we propose a method for classifying a non-redundant relation to control the dynamic and static use of the artificial upper limb. We began remodeling the degrees of freedom (DOF) of the human hand by identifying the transverse and longitudinal adjustable arch structure in the hand. Our new model is composed of 21 active DOF, which include movement in the palm. We classified movements and postures of the hand and arm with this DOF model located at the end of a seven-DOF arm. We then classified the hand modes as prehensile forms and sustentacular forms. Based on this model and our previous research experience in developing prosthetic upper limbs and anthropomorphic robotic hands, we devised a robotic hand-arm with a total of 24 degrees of freedom. Additionally, a new multitactile sensor has been developed for use on the robotic hands in our laboratory at TDU. For generating control strategy, the behavior of the hand is classified into two divisions. The first is cooperation of the digits and palm, and the second is cooperation of the hand and arm modules. The digit cooperation task is fourfold: formation, transformation, deformation, and hold. A motion planner drives the digit movements from the relation of the hand forms to the task that is proposed. A tactile sensor-based control strategy is presented. The digits are controlled according to the hand modes and the sensory feedback loop, with slippage detection rules. Two strategies are proposed for extending the control for fine manipulation: cooperative slippage sensing of adjacent digits, and slippage prediction applying hand orientation information measured by an inclinometer. The overall objective of our approach is the design of a dexterous control system for a multi-DOF robotic upper limb. This includes discussion of a modeling method for the human upper limb.
To clarify the control mechanism for jaw movements, we developed an autonomous jaw-movement simulator, JSN/1, consisting of upper and lower jaws, temporo-mandibular joint, muscular actuators, periodontal sensors, control unit, and a computer. The muscular actuator consists of a cable-tendon DC servo motor, equipped with a cable-tension sensor, and a rotary encoder capable of detecting cable length, thereby providing JSN/1 with masseter, temporalis, lateral pterygoid, and digastric artificial muscles. The actuators were controlled adaptively under an impedance-control mechanism, using bite force, and tooth contact data, plus the tension and length of the actuator cable. We also developed a physiological adaptive control scheme for open-close movement of JSN/1, using EMG data for related muscles. The previous simulator, however, had a defect: the temporalis actuators were activated under a non-physiological position control during closing in order to obtain high reproducibility of the movement. To eliminate that defect, we first incorporated artificial lateral ligaments into the temporomandibular joint of the simulator, in an attempt to provide the condylar movement with a passive restriction. Additionally, the drive signal for the anterior temporalis actuator was updated to cope with a delay of closing movement due to the reduction of the position-feedback gain. This update was accomplished not only by reducing the position-feedback gain, but also by dividing the signal into two different sinusoidal waveforms, each having a control parameter, thus enabling us to modify the signals for closing and biting phases independently. Accordingly, the drive signals for the masseter actuator and internal-pterygoid actuator, newly incorporated in the JSN/2B, were modified so as to be activated immediately after tooth contact. Their position-feedback gains were also reduced. The second aim of this study was to establish a simulator control scheme for chewing-like jaw movement without food, as a preliminary study for the simulation of actual chewing movement. We assumed that such movement can be obtained merely by modifying the values of several control parameters of the digastric, lateral-pterygoid, and posterior temporalis actuators, determined adaptively during open-close movement, and that the values of the parameters can be optimized using data relating to the amount of lateral shift of the mandible. Experiments using the updated JSN/2B verified that a lifelike, reproducible open-close movement could be obtained in the simulator even without high position-feedback in the actuator control, demonstrating both the necessity of joint ligaments for stability of condylar movement and the validity of our control scheme. In other experiments on chewing-like jaw movement, we introduced two different optimization functions to determine the values of the control parameters, using bite-force sensor data on the working side and tension sensor data for the lateral ligaments. Empirical results using both optimization functions demonstrated that a lifelike, reproducible chewing-like jaw movement could also be obtained by slightly modifying the values of the control parameters, which were first optimized during open-close movements.
Independent mobility is an important factor in determining the quality of life. The powered wheelchair is a significant device in ensuring independent mobility for persons with severe disabilities. The adaptation of wheelchairs to persons with severe disabilities, however, is very difficult and takes a long time because many trials are needed. The purpose of this study is to support this complex situation. This paper describes development and practical use of a simulator for a powered wheelchair. The simulator has two computer screens and a moving platform. The screens show computer graphics of a virtual driving course, which includes traffic signs, a railroad crossing, a pathway with a curb from the roadway, a roadway with a cross slope, rough surfaces with bricks, slopes, and a cross slope. Some pedestrians and some bicycles come along on some of the path- and roadways. The platform is connected to six linear servo actuators that work by electric power. The actuators generate accelerations and decelerations similar to those of a real powered wheelchair. This system has a mathematical model of a powered wheelchair in the computer. Since it is driven by torque of the right and left drive wheels, it gets slower on an uphill slope and faster on a downhill slope. The model also turns downhill on a cross slope. Fifteen subjects, including two powered wheelchair users, drove the simulator. Results of this evaluation revealed that drivers of the simulator had virtual experiences that were quite similar in many ways to driving a powered wheelchair. The subjects, however, felt it more difficult to operate the simulator than to drive a real powered wheelchair. This was caused by the absence of a side view. Another problem was motion sickness, which was experienced by about half of the subjects. Caution about motion sickness is needed during the first trial of this simulator. A person with physical and mental disabilities tried the simulator in order to learn operation of a powered wheelchair. He lacked comprehension, concentration, and motivation. The first trial, of operating only one switch for driving the simulator, showed that he could understand the system and concentrate his attention on the operation. The next trial, requiring operation of two switches for driving, showed that he could not stop when a risk of collision arose. At this stage, we didn’t think he could drive the powered wheelchair. The last trial, controlling direction using head movement, however, showed he turned the wheelchair along the line on the road. This was a surprising result. These evaluations revealed that he could drive a powered wheelchair if it compensated for safety by itself. The simulator is effective in determining if a person with very severe disability can drive a powered wheelchair. This attempt will extend independent mobility of persons who have been regarded as unable to drive powered wheelchairs.
Daily living care for severely disabled people, transferring them between wheelchair and bed for example, is very hard work for a caregiver. Actually, over 80% of professional caregivers had experience of lower-back pain. Many types of lifting devices are available for transferring disabled people, but hanging the person in the sling requires a large moment around the lumbar joint of the caregiver. The most practical solution to reducing the risk of lower-back disorder, therefore, should be to select the most suitable method for transferring. The lumbar joint moment is a reasonable biomechanical evaluation index for workload because it relates directly to the worker’s backmuscle tension and lumbar disk pressure. The joint moment is calculated from posture data and applied force data during care motions using a rigid-link model of the human body. In the case of two-person transfers, the caregivers stand sufficiently far away from each other that a standard motion measurement system of force plates and opto-electric cameras can be used. The measurement of one-person transfer motion is difficult, however, because many external forces caused by a patient’s body weight are applied on the caregiver’s contact points, and the body parts of the caregiver are extensively covered by the patient’s body. In this study, a new measurement method was developed. The contact forces were measured with a fluid pressure sensor jointed to a thin rubber tube sandwiched between curved plastic plates. The absolute angles of body segments were measured with twelve small tiltmeters. These sensors had been sewn on a plain jacket beforehand to reduce preparatory time for the measurement. Hold-type transferring and carry-type transferring by one person, and front- and back-type transferring by two persons were measured for beginners and for specialists. In hold-type transfers, the caregiver holds the patient’s head under one arm and grasps the patient’s waist belt, pulling up around the knee joints as a fulcrum utilizing the caregiver’s body weight. The carry-type transfer is the ordinary carrying method on one’s shoulder. The front caregiver in a two-person transfer supports the patient’s knees and calves, and the back caregiver pulls up the patient’s upper body. Calculated results of lumbar load in each type of transferring show that the hold-type transfer by one person is the least risky method regardless of caregiver’s skill level and body physics. The worst case is a carry-type transfer by a tall caregiver. In this case, the lumbar load reaches three times that of a hold-type transfer by a short caregiver. The load of a two-person transfer is as great as that of a one-person transfer. The load can be reduced to 80% by raising the bed surface about twenty centimeters above the wheelchair seat. In conclusion, the hold-type transfer by one person is a suitable transfer method for avoiding caregiver’s lower-back pain. The portable measurement system developed and used in this study will be applicable to analysis of other kinds of care motions.
One of the most serious problems in current labor-intensive industries, such as clothing manufacture, is difficulty in maintaining special hand skills, especially because of aging of highly skilled workers. A systematic way to hand down special skills is therefore becoming urgently necessary. For the purpose of supporting transmission of such special skills, we propose an efficient method of clarifying the important points of highly skilled behavior. As target tasks for analysis, we selected four typical difficult operations in clothing manufacture with sewing machines. The results can be flexibly applied to any kind of sewing-machine operations. Through observation of these operations, we pointed out several factors in the differences between highly skilled operators and unskilled operators. For example, highly skilled operators always keep their elbows lower and their wrists higher, whereas unskilled operators use the opposite posture during the task. To clarify the hypothetical skill factors, quantitative measurement was performed, and an analysis method was developed to quantify the differences in behavior between highly skilled operators and unskilled operators. Task behaviors were measured with several sensing devices, such as a motion capturing device, data gloves, and a sewing-machine motor rotation counter. The proposed analysis method was applied to the experimental data, and adequacy of hypothetical skill factors was confirmed. The elbows of skilled workers are kept more than 40 mm lower than those of other workers. Qualitative analysis was also performed. Differences in motion sequence patterns were analyzed as follows: First, arithmetically characteristic points were automatically extracted from each set of time-series data for measured behavior. These characteristic points were associated with predefined symbols and the symbols were used to establish a sequence pattern. The sequence pattern of symbols obtained was identified with a task model, which is a previously defined series of motions. Thus, portions of the whole measured behavior were associated with parts of the predefined task in partial motion level. Finally, the sequence patterns of symbols obtained from highly skilled operators and unskilled operators were compared, and the differences were extracted. Corresponding parts of sequences for each motion were also compared and their differences in partial motion level were extracted.
There have already been many studies of bipedal walking, but most of them have concentrated on building a good controller and have failed to utilize the dynamic properties of robots. Recently a new approach to bipedal walking has appeared. Using this new approach, a robot can walk down a shallow slope without any actuators or controllers. This walking mode is called Passive Dynamic Walking (PDW). Because of its simplicity in respect to both energy and control, PDW is thought of as a “glider” in walking (Pratt, 1999). Since the appearance of the first paper on PDW by McGeer (McGeer, 1990), many researchers have studied various aspects of it. Previous studies, however, treat only the lower part of the body, not the torso. If the purpose of PDW studies is to make effective use of the body in a humanoid robot, it is natural to treat a robot with a torso. In this paper, we show the effects of the torso on PDW walking in two-dimensional computer simulation. For the first PDW trial with a torso, we used a very simple mode consisting of three links: a torso and two legs. This is a simple extension of the simplest PDW robot. First we examined the possibility of PDW by this robot without adding any torque. The center of mass of the torso and the initial condition of walking were changed variously. We could not, however, find any suitable parameters or initial conditions for stable walking with the center of mass of the torso higher than the hip point. We therefore added torque to support the torso and tried to suppress that torque. As we wanted to maintain simplicity of PDW, a simple feedback control scheme was applied as follows: τ=−kvθw+k(θwd−θw) Torque is applied only to the joint between the torso and the support leg. Given appropriate initial conditions, the robot can walk stably down a shallow slope. Moreover, the walking mode of this robot has the same properties as the original PDW robot suggested by McGeer. For certain parameter combinations, the PDW robot can walk with a stable limit cycle motion. To suppress the torque needed to stand the torso up, four body parameters were changed variously: changing the posture of the torso, adding soft leg tips, changing the curvature of the sole, and adding passive joint resistance. From experimental results, four qualitative points were obtained: (1) as the torso inclines backward, the torque becomes small, (2) a shock absorber is useful to reduce the torque at heel contact, (3) the shape of the sole is important in suppressing torque in the supporting phase, and (4) passive joint resistance is effective in accelerating the speed and duration of walking, but it cannot suppress the torque. Simple analysis shows that the torque at the hip joint to stand the torso up is closely related to the acceleration of the hip. Moreover, the torque at the contact phase is determined by the vertical component of acceleration of the hip, and torque in the supporting phase is determined by the horizontal component. This analysis improves understanding of the simulation results.
It is effective to determine running pace in advance, based on individual ability, in order to demonstrate the highest performance in long-distance running. The evaluation indices for a long-distance runner are maximum oxygen uptake, lactate threshold (LT), and ventilatory threshold (VT). These, however, are mostly used stastistically, so results may differ from real ability in a personal equation. The purposes of this study were to construct an energy-metabolism model and to optimize the running pace of long-distance running using a genetic algorithm (GA). The energy-metabolism model constructed in the study was composed of an anaerobic energy feeder structure, an aerobic energy feeder structure, and the section to be run. These elements were expressed as differential equations and restricted inequality formulas. The running speed for each subject, calculated from the best time for 300 meters, the amount of oxygen uptake, and running speed at the VT in each subject were used as parameters for the energy-metabolism model. VT was measured by a gradually increasing speed exercise using a treadmill because it was difficult to measure during field running. There are many differences between treadmill running and field running, however. In this study, the subject ran continuously on a treadmill with traction to his back using a rubber tube. The running speed for treadmill running was adjusted to that in field running based on heart rate. The energy-metabolism model had two controlled variables, and running speed could be controlled by these variables. We tried to optimize the energy-metabolism model by determining the two controlled variables using a GA. The spurt start point was also determined during optimization. The GA determined the spurt start point based on the energy-metabolism model. The running speed in 5000-meter races was optimized as follows: (1) speed ascends immediately after the start of the race, and then descends by a constant degree; (2) speed ascends again at 1000 to 1400 meters before the goal; and (3) almost 1 minute later, running goes to maximum speed then descends again by a constant degree all the way to the goal. This optimization result corresponded closely to the actual racing of the subject, who trained for improved ability in long-distance running.
The purpose of this study was to clarify the relationship between jumping direction and torque of the hip, knee, and ankle joints. Two methods were employed: an inverse dynamic approach, and a direct dynamic approach. The small degree of knee joint torque in the horizontal movement coincided with the report of Oshima et al. on output force distribution of the functional effective muscle system. Using the inverse dynamic approach, joint torque and work were compared between a standing broad jump and a vertical jump. The standing broad jump is a horizontal movement task, whereas the vertical jump is a vertical movement task. The ratios of work at the hip, knee, and ankle joints to total work were compared between the standing broad jump and vertical jump. Less work was done at the knee joint for the standing broad jump than for the vertical jump. The work was calculated as the product of joint angular velocity and joint torque. The time curve of joint angular velocity was similar for the standing broad jump and vertical jump. Knee joint torque for the standing broad jump was less than that for the vertical jump and was also less than were hip and ankle joint torques. For the vertical jump, the amplitude of knee joint torque was the same as that for the hip and ankle joints. Using the direct dynamic approach, a computer simulation was run in which the hip, knee, and ankle joint torque amplitudes for the vertical jump were modified by predecided magnifications. The relationship between magnification of the modified joint torque and the body’s center of gravity was examined. The predecided magnifications for each joint torque were from 0.0 to 1.9, at intervals of 0.1. There were thus 8000 patterns of magnification sets. The smaller the magnification of the modified torque of the knee joint in comparison with that of the hip and ankle joints, the greater was the forward velocity of the body’s center of gravity. It became clear that a small degree of knee joint torque is necessary and sufficient for generating horizontal movement. It was predicted that such a small degree of knee joint torque was caused by co-contraction between the knee joint extensor muscle and the bi-articular muscles. Oshima et al., in reporting on output force distribution of the functional effective muscle system, stated that the mono-articular hip extensor, mono-articular knee extensor, and bi-articular knee flexor must be activated in order to push the foot against the ground both backward and downward. If these muscles are activated, knee joint extension torque will be small.
A multibody dynamics model of the humerus-shoulder complex system, driven by a musculoskeletal system, has been formulated based on an updated Lagrange approach, using the continuum mechanics of muscle with the evolutional constitutive law of Hatze’s model. Japan’s population is rapidly aging; within the next 10 years, the elderly are expected to reach 25 percent of the total population. Assistive technology for recovery from disease and increasing medical expenses require more rational approaches to the bone-muscle system in limb-trunk motions, such as interactions between the thoracolumbar-spine system and the upper and lower limb system, which had been insufficiently analyzed. This paper focuses on a theoretical model of a multiple bones-joints system, driven by muscle activations. First we formulated the continuum mechanics of skeletal muscle with the evolutional constitutive law of Hatze (1977). A joint muscle activation generates kinematics and dynamics. For example, activation includes (1) generating angular motion of linked systems, for a multibone and multijoint system; (2) constraint of muscle contraction due to neuro-activation generating a counter force applied to equilibrate the external joint torque; (3) a paired contraction of counter muscles in a joint, giving rise to increased joint stiffness; and others. We therefore formulated the evolutional constitutive law of a muscle model so that it includes independent variables for both motion and constraint forces, the sum of which determine the activation level. The muscle constitutive equations were first formulated using the link coordinate system of each muscle strip. Then they were assembled by coordinate transformations from those for each strip to a common global system. A multi-body dynamics approach for a multiple bone-joint system driven by muscle activation was then formulated. The link element coordinates were defined for each bone, and positional vectors for the center of gravity of each bone, distance between adjacent joints, and muscle positional vector were formulated and transformed to the global coordinate system. Using the positional vectors, translational and rotational motions of individual elements, we formulated the kinetic energies of the elements and the potential energies stored in individual muscles and joint elements, expressed in incremental forms. The governing equations of motion for each link, and the dynamic equilibrium equation for each joint, based on the updated Lagrange method, were formulated and solved by prescribed constraint and applied load conditions, using the developed computer simulation system. As an actual application to assistive technology problems, an elderly person’s stand-up motion, aided by humerus-shoulder forces for lifting one’s upper body weight was analyzed. The calculated results for activation dynamics were compared with those, obtained from an EMG measurement experiment. Satisfactory agreement between these was confirmed, verifying that the developed computer system can satisfactorily estimate activation dynamics of muscle, replacing the traditional EMG measurement approach.
I designed a robot arm that simulates the human elbow-forearm structure. The model consists of a one-axis upper arm part and a two-axis forearm part, connected by an elbow joint. With this model, I attempted to create a structure that can flex and extend at the joint and simultaneously rotate at the distal end of the forearm. Geodesic curves on the cone were used for the two axes of the forearm part. Mathematically it is generally known that · When a cone rolls on a plane, the conic vertex does not move [theorem 1], and · When a cone rolls on a plane, the geodesic curve on the cone moves on a straight line on a plane [theorem 2]. With these theorems, I produced two movement models. Model I: When two cones and the geodesic curves on the cones roll on the front and reverse sides of one plane, they always touch during motion on a straight line on the plane. Model II: This is the movement model produced by adding a straight line that connects two conic vertexes to model I. Two geodesic curves on the cone can turn around the straight line. It was shown that movement model II was a structure for which rotation at the distal end of the forearm part was possible simultaneously with flexion and extension at the elbow joint part. Movement model II was redefined from the geometric movement conditions near the point of contact of the two geodesic curves on the cone. Further modification of these geometric movement conditions resulted in a new model with more natural movement of a human elbow-forearm structure.
Human motions seem to be so generated as not to oppose body mobility resistance due to anatomical constraints such as limb kinematics, inertial properties of limbs, range of joint motion, and muscle size and attachment. This implies that the human nervous system (the brain) may somehow represent the anatomical constraints and that neuronal representation of body mobility may be utilized so as to spontaneously generate a natural reaching motion. In this study, the body mobility due to anatomical constraints was hypothesized to be represented by a series of self-organizing topographic maps, and we attempted to develop a neural network model that can spontaneously generate natural reaching motions based on those maps. The musculo-skeletal system of the human upper extremity is constructed as three rigid, two-dimensional links in a sagittal plane. A visco-elastic element is attached around each joint to represent passive joint structure. A total of eight muscles are attached around the joints, generating force according to muscle activation signals from the nervous system. The nervous system is modeled as an artificial neural network that incorporates topographically arranged arrays of neurons (topographic maps), the weights of connections of which represent negative gradient vectors of potential functions defining the body mobility due to anatomical constraints and task constraints of body motion. Thus the nervous system may autonomously generate muscular activation signals that tend to move the hand to a goal, while yielding to the mobility of the musculo-skeletal system, if the neuronal representation of body mobility is correctly organized. In this study, a biologically motivated self-organizing mapping algorithm was constructed in order to acquire the maps. As a result of the proposed learning algorithm, the maps are successfully self-organized from a series of random movements of the arm. The motion generated using the maps actually improves as learning proceeds. The simulation results imply that these neuronal maps representing body mobility due to anatomical constraints are probably obtained in the brain for generation of natural motions, and these maps can be self-organized by a relatively simple algorithm. The existence of these neuronal mechanisms seems to be indispensable for spontaneous motion generation, indicating that the proposed mechanisms may be incorporated in actual human motor control.
Recently, nonstationary properties of various chaotic phenomena in biological systems were studied extensively. It is well known that there are some nonlinearities in speech signals. With this fact in mind, we evaluated chaotic properties appearing in the five Japanese vowels. Taking our previous studies into account, here we report the method of nonlinear analysis, comprehensively. Following these results, we show that some positive components appear in the Lyapunov spectrum of all five Japanese vowels. Furthermore, we try to apply recurrence imaging analysis, which type of nonstationary analysis, to our biological system. In this paper, we propose a new technique based on resolution parameter to categorize the five Japanese vowels without any interruptions, by nonstationary signals in the vowels. We hope the report will be useful among the fields of speech-engineering and complex systems. (Study content and flow) 1. We studied whether chaotic dynamics appear in the five Japanese vowels by estimation of the Lyapunov exponent based on the report of our previous study. In the report, each vowel showed the chaotic property, since a positive Lyapunov exponent was observed in the Lyapunov component in each vowel. To construct the embedding space from the time series, we estimated a delay time and embedding dimension. False Nearest Neighbor analysis is used in determining dimension, and mutual information criterion was used in estimating the delay time. 2. Since the value of the delay time influenced the estimation of the embedding dimension, we tried to determine both values consistently. In fact, the previous approach is a method that is executed for correct embedding. Even though we accurately estimated d and τ, the speech data (dynamics) include linear stochastic properties, which would be calculated as spurious exponents under improper embedding. So we also realized that the positive component is not caused by internal gaussian noise but is generated by some nonlinear vocal source using surrogate data methods. 3. It is known that the method of generating of surrogate data employs a variety of algorithms. We exploited the algorithms for FT (Fourier transforms) and AAFT (amplitude adjusted Fourier transforms), and tested whether each algorithm can reject these null hypotheses. 4. In this work, we exploited Recurrence Plot (RP) imaging, which is a nonstationary analysis of the time series. As a result, we saw that each vowel has particular pattern in the RP image and can extract the attractor information from the image. The vowel data are based on the acoustic voice of a male subject (age 40) recorded in an ATR speech database. These data are a time series of sound pressures with a sampling rate of 20 kHz and 16-bit quantization.
The purpose of this study was to estimate the vocal-tract length based on the distribution of sound pressure in the three-dimensional vocal tract during production of the vowels /a/ and /i/. The distribution of sound pressure was estimated in the three-dimensional vocal tract using the Finite Element Method (FEM). The three-dimensional shapes of a vocal tract and a dental crown were measured using Magnetic Resonance Imaging (MRI). A male subject was asked to produce the vowels /a/ and /i/ while wearing a dental crown plate that contained a contrast medium for MRI processing. Three-dimensional MR images of the vocal tract for each sound were obtained while the subject’s tongue was kept still. The vocal tract shapes were obtained from profiles of their sagittal sections. First, sagittal MR images (4-mm interval) were estimated from coronal MR images using grey-level interpolation. Second, air-tissue boundaries of the sagittal sections were obtained from each MR image using a threshold operation in which the threshold value is the average of grey levels at the air-tissue border points. Adjacent air-tissue boundaries are connected by spline interpolation, since the vocal tract shapes are constructed by a cascade connection of the air-tissue boundaries. The three-dimensional vocal tract shapes during vowel production were divided with a tetrahedral element (/a/: 11, 179 elements, 3, 185 nodes; /i/: 12, 763 elements, 3, 558 nodes). At the glottis, the whole area of the cross-section was driven, and at the radiational surface, the radiational impedance was assumed to be that of a circular piston in a rigid sphere. The boundary condition of the wall was defined as the wall impedance. The distribution of sound pressure shows that a sound wave propagates in the vocal tract as a non-plane wave in the high-frequency region. In this study, the vocal tract model with cascading circular tubes was called the VT model. The vocal tract transfer characteristics for both the VT model and the three-dimensional FEM model were determined. A comparison of the VT model and FEM model showed that they share almost the same characteristics in the low-frequency region, but express a difference at high frequencies. The vocal-tract length was estimated based on the distribution of sound pressure in the three-dimensional vocal tract. As a result, in the case of /a/, the vocal-tract length is 188 mm for a driving frequency of 1000 Hz, and 226 mm for a driving frequency of 3000 Hz. In the case of /i/, the vocal-tract length is 163 mm for a driving frequency of 1000 Hz, and 173 mm for a driving frequency of 3000 Hz. These results suggest that the vocal-tract is long in the high-frequency region.