Five male subjects (mean age 24.2 ± 6.3 years) without musculoskeletal or neurologic impairments took part in the experiment after giving informed consent. The study protocol was approved by the institutional ethics committee.

The subjects performed the STS task from a standardized sitting posture on a chair in two speed (fast and natural) conditions. They were told to cross their arms in front of their chest throughout the duration of the task. In the natural speed condition, the subjects were asked to stand at a comfortable speed after receiving a signal from the experimenter. In the fast speed condition, the subjects stood as fast as possible. The subjects gazed at a point presented in front of them during the tasks. They were also instructed to stand still at the end of the movement task until receiving another signal from the experimenter. This second signal indicated the start of the BTS movement task at the same speed condition. The height of the seat was set to two different positions, i.e., Low and High positions that yield knee angles at 80 degrees and 50 degrees flexion, respectively. Angle of the ankle joint was 10 degrees dorsiflexion for both positions. All subjects started STS tasks in the Low-natural speed conditions, followed by the fast speed condition. Tasks in the High position were always examined after all the tasks in the Low position were completed. The subjects performed 10 trials in each condition, thus a total of 40 STS trials and 40 BTS trials (2 speed conditions by 2 seat-height positions) were recorded. There was a 3minutes rest period between each test. The subjects were allowed remain seated during the rest.

The schematic diagram for the motion analysis process was shown in Figure 1. The positions of reflexive markers attached to the bony landmarks were captured using seven VICON MX cameras at a sampling frequency of 200 Hz (Vicon, UK). The markers were taped bilaterally on the acromion, greater trochanter, lateral epicondyle of the femur, lateral malleolus, and the head of the fifth metatarsal. A three-link rigid body model

composed of HAT segment (Head, Arms and Trunk), thigh, and shank in the sagittal plane was assumed to calculate the anatomical joint angle of the hip, knee, and ankle joints and position of the COM of the subject. The angular velocity and acceleration as well as the linear velocity of the COM were also calculated by numerical differentiation. The length and weight of the links for each subject were estimated using the height and weight of each subject [17] . Movement time was determined as the duration when the velocity of the COM exceeded 5% of its peak velocity. For the convenience of data processing, the data length of kinematics was normalized by the movement time (one hundred points). Raw kinematic data were low-pass filtered using a second-order Butterworth filter with a cut-off frequency of 5 Hz [18] .

The ground reaction force (GRF) under the seat was monitored using a Kistler 9287B force plate (Kistler, Switzerland) at a sampling frequency of 1 kHz to determine the moment of seat-off during STS, i.e., when the GRF reduces to zero, and the moment of seat-contact during BTS, i.e., when the GRF increased above two standard deviations of the unloaded period.

Two criteria were adopted to compute the trajectories. First, the cost function for the minimum jerk trajectory of the COM was determined as

where and represent the jerk (third derivative) of the COM position in the Cartesian coordinate system,

and T denotes the movement time. The optimal time sequence of COM positions, mini-

mizes the cost function under the boundary conditions and with

via-points,. As described later, timing of a via-point P was selected based on the kine-

matic events. Second, the cost function for minimum angle-jerk trajectory of the joints was determined as,

where denotes the angle jerk (third derivative of the joint angle), and subscript j indicates the joint number.

The optimal angular trajectories, minimizes the cost function under the boun-

dary conditions and with via-points

. In this study, the boundary conditions were obtained from measured data aver-

aged for ten trials in each speed condition. The problems were solved as the dynamic optimization with interior -

point equality constraints, as described by Flash and Hogan [2] .

To examine the temporal optimality of the movement kinematics, some position datasets during movement, i.e., via-points, were selected in the computation. Because the via-point is not evident a priori, the candidate positions were selected as the via-points based on the important components of the STS and BTS suggested by previous studies; i.e., hip flexion, and extension and COM movement velocity [19] - [21] . In this study, six candidate positions were identified:

・ Position at peak velocity of the hip flexion ();

・ Position at maximum angle of the hip ();

・ Position at peak velocity of the hip extension ();

・ Position at seat-off for STS or seat-contact for BTS (P-seat);

・ Position at peak horizontal velocity of the COM ();

・ Position at peak vertical velocity of the COM ().

The optimal time to pass through a via-point or via-posture was determined in the optimization process [2] .

To quantify the correspondence between the optimal COM trajectory and measured COM trajectory, the norm of the difference in each data point was calculated as the residual,

where superscript opt denotes the data of the computed optimal trajectory, and superscript meas denotes the data from the measured trajectory averaged over 10 trials. Subscript n means the data number within a total number of m (m = 100). The correspondence between angular kinematics can also be quantified by a similar equation,

where j represents joint number. All numerical computations were performed using custom programs of MATLAB (The MathWorks, Inc., USA).

Statistical analysis was carried out using SPSS version 17. Differences in the residuals and effects of speed and seat height were tested using analysis of variance (ANOVA) tests. Statistical significance was set at p < 0.05.

Mean and standard deviation of movement times during STS in the Low condition were 0.98 ± 0.12 (s) and 1.58 ± 0.19 (s) for the fast and natural speed conditions, respectively. For BTS, the movement times for the fast and natural speed conditions were 1.26 ± 0.10 (s) and 1.88 ± 0.40 (s), respectively. Two-way ANOVA (speed, seat height) was performed separately for both task direction (STS or BTS). The main effects of speed were significant for task direction (p < 0.01). The main effects of height and interactions were not significant. These results confirm that the subjects performed the tasks at different speeds according to the instructions.

To examine the effect of the condition on the initial and final joint angles, two-way ANOVA (speed, seat height) was performed separately for each task direction (STS, BTS) and the joints (ankle, knee, and hip). A significant main effect of seat height was found on the knee and hip joints for both task directions (p < 0.01 for initial and final) but not on the ankle joint for both STS and BTS. The effect of speed was not significant for all joints for both task directions, indicating the boundary joint angles were independent of the speed conditions.

As shown in Table 1, the timings of the candidate positions, i.e., the occurrence of measured kinematic features

relative to movement time, are distinguishable for all conditions except for between (at peak hori-

zontal velocity of COM) and (at peak vertical velocity of COM) in the BTS task. According to

two-way ANOVA (speed and candidate positions), there were significant main effects candidate positions (p < 0.01 for both STS and BTS) on the timing. The main effect of speeds and interaction effect were not significant for the STS and BTS. Bonferroni post-hoc analysis confirmed a significant difference among all six candidate

positions for STS and BTS (p < 0.01) but not for the difference between and in the BTS.

Hence, six candidate positions for the STS task and five positions for the BTS task (, ,

, P-seat, and) were temporally discriminated as via-points. The was unused for

the BTS task.

Linear regression analysis revealed that there were highly consistent temporal relationships between the timings of the candidate positions and movement time. Table 2 summarizes the slope and the determinant coefficient of the linear regression lines, showing that the timings of candidate positions are highly dependent on the move-

ment times ().

Given the candidate positions for via-points, six different optimal trajectories for STS and five optimal angular trajectories for BTS could be computed by the minimum angle-jerk model for each speed and seat-height condition. Figure 2 shows an example of angular profiles and computed optimal trajectories for the Low-fast speed condition for one subject. Selected original points as via-points are indicated by square markers and computed optimal via-points (and its timing) are plotted using circle markers. An optimal via-point computed for the hip

joint at in STS is clearly different from the original via-point in terms of timing. On the other hand,

computed optimal via-points at as well as the optimal trajectories for these via-points in both STS

and BTS are relatively close to the measured ones. In addition to the angular trajectories, optimal trajectories of the COM are also computed for the via-points. Figure 3 shows the time profiles of the COM displacement and

the optimal trajectories computed by the minimum jerk model. The trajectory computed for (a

via-point at peak velocity of the hip flexion) in STS is found to be close to the measured trajectory for both the horizontal and vertical directions. As for the BTS trajectories shown in Figure 3(b), on the contrary, the computed trajectories for all via-points are not compatible with the measured kinematics.

Figure 4 shows the residuals of difference, or the degree of correspondence between the computed optimal trajectory and measured trajectory in terms of the angular () and COM () trajectories for all conditions and subjects. On average, was small when the optimal trajectories were computed using the via-point at the maximum hip flexion for both STS and BTS. This trend was also true for the residuals evaluated on the COM kinematics () except for the BTS. Two-way ANOVA (height and via-points) was performed separately for each direction (STS, BTS) and type of residuals (,). As for the in STS, the main effects of the height (p < 0.01), via-point (p < 0.01), and interactions (height * via-points, p < 0.01) were

significant. Bonferroni post-hoc analysis found that at was significantly smaller than s at

, P-seat, , and. Main effects of height (p < 0.01), via-points (p < 0.01), and inte-

raction (height * via-points, p < 0.01) were also significant for in STS. Differences among s of

, and were not significant, but these residuals were significantly smaller than the s

of and. As for the in BTS, the main effects of the height (p < 0.01), via-

point (p < 0.01), and interactions (height * via-points, p < 0.01) were significant. Differences among the s

of P-seat, , and were not significant, but the residuals were significantly smaller than

the residuals at and. The main effects of height was significant (p < 0.01) for in

BTS, but post-hoc analysis failed to show significant differences among residuals at the via-points. These results indicate that the optimal angular trajectory computed using the via-points at around the maximum hip flexion fit the measured angular kinematics well.