Two classes of movements in motor control

Abstract

This work investigated whether fundamental differences emerged between segments of complex movement sequences performed at different instructed speeds. To this end, we tested 5 novices and 1 karate expert as they performed beginner’s martial arts routines. We found that if one blindly took these segments and separated them according to the variability of trajectory parameters, one could unambiguously group two classes of movements between the same two space regions: one type that remained quite conserved despite speed changes and another type that changed with speed level. These groups corresponded to functionally different movements (strike segments explicitly directed to a set of goals and spontaneously retracting segments supplementing the goals). The curvature of the goal-directed segments remained quite conserved despite speed changes, yet the supplemental movements spanned families of trajectories with different curvature according to the speed. Likewise, the values of the hand’s peak velocity across trials were more variable in supplemental segments, and for each participant, there were different statistical signatures of variability between the two movement classes. This dichotomy between coexisting movement classes of our natural actions calls for a theoretical characterization. The present experimental results strongly suggest that two separate sets of principles may govern these movement classes in complex natural behaviors, since under different dynamics the hand did not describe a unique family of trajectories between the same two points in the 3D space.

Appendix

The first isolated technique is called a Jab. The Jab starts with the front hand extending toward the imaginary opponent’s nose (J1), keeping the hand in a tight fist, making sure that the elbow does not hyperextend; the hand should be retracted while it is still slightly bent (Fig. 2a). At the same time that the Jab is being retracted (J2), the Cross is being extended forward (C1). Again the imaginary target is used and the Cross is directed toward the nose. Simultaneously, the body is twisting, beginning with the back foot, then the torso and ending with the back of the hand extending forward. Because the body is already twisted, this motion naturally sets up the staged portion of the Hook (H1) aimed at the opponent. As the Cross (right hand) reverts back to its original position (C2), the left forearm is made into a C-shape with the hand in a fist and the palm facing down, and the body untwists itself, using the momentum of the body rather than the force of the hand to achieve the intended goal (to reach the opponent’s face). As the body untwists itself in a supplemental H2, the knees bend slightly in preparation for the intended Uppercut (U1). After the knees are bent and the left hand is returning to its original positioning to protect the face, the right hand fist shoots up in a motion that resembles throwing a bowling ball, but the hand is kept tighter aligned to the body and the palm facing the body. The supplemental portion U2 brings the hand back, and the body adopts the defense position again (Fig. 7 bottom panel-end of the cycle). It is important to note that all routines where done in the presence of an expert instructor in order to minimize risk of injury. For further information and descriptions see our website for a detailed video tutorial on these techniques.

Fig. 7

Methods–Martial arts routine—Jab–Cross–Hook–Uppercut

Three dimensional digital rendering frames from the expert’s performance of one trial of J–C–H–U beginner’s white belt technique using the real-time sensor outputs. Arrows mark the locations of 15 electromagnetic sensors recording at 240 Hz. The motion capture system provides the choice of outputting the raw accelerations and velocities (linear and angular) or various filtering and smoothing options. The system deals directly with potential spurious or noisy data due to estimation of higher order derivatives of position. Because of its reliability, this software-system is routinely used as a standard interface in sports training of the kind studied here. The update rate of 240 Hz per sensor and the latency of 3.5 ms permit real-time monitoring of the body motions. Each sensor has 6° of freedom (DOF) with static positional accuracy of 0.03in, static orientation accuracy of 0.15 deg RMS, a positional resolution of 4 × 10⁻⁵ cm at 30 cm range and a resolution of 1.2 × 10⁻³ deg of orientation. The range from the standard source is up to 1.52 m and the extended range is up to 4.6 m. Our experiments took place well within the standard source range. Several standard filtering algorithms are used by the professional software that comes with the Motion Monitor Sports Inn. We used a Butterworth filter (Butterworth 1930) with a cut-off frequency of 6 Hz. Further details about the electromagnetic system that our Motion Monitor uses can be obtained at the Polhemus company website.

Green traces in the rendered figure mark the hand motions. Forward segments were away from the body and staged against an imaginary opponent. They coexisted with the supplemental transitions of the other limb simultaneously moving away from the opponent. (A) Jab1 in the forward direction, away from the body. (B-D) Jab 2 back toward the body simultaneously with Cross 1 forward. (E–F) Cross 2 back simultaneously with Hook 1 forward. (G-J) Hook 2 back simultaneously with U1, ending the routine with both hands back to protect the face. Bottom panel focuses on the expert’s hands (Google my lab’s website for more details and videos).