This invention relates generally to joint modeling, and more particularly, to a component framework that offers complex articulation and biomechanical accuracy for representing joints.
Accurate representation of skeletal articulations is important for many applications including biomechanics and computer graphics. In biomechanics, for example, accurate joint models can be used for injury prevention and rehabilitation. For computer graphics, accurate joint models can lead to improved realism in character animation.
One conventional approach for representing human characters with an articulated joint model uses the Denavit-Hartenberg link parameter notation from robotics to represent figures with articulated limbs. One such conventional system is described in M. Girard et al., “Computational modeling for the computer animation of legged figures,” Computer Graphics (SIGGRAPH '85 Proceedings), vol. 19, pp. 263-270, 1985, which is incorporated by reference herein in its entirety. Although the parameter notation relates coordinate frames between adjacent segments with four parameters, each parameter set describes only a single degree of freedom between two segments. Multiple sets of parameters can be combined to achieve multiple degree of freedom (DOF) joints, but additional complexity is added for the user to manipulate or to make use of the resulting joint expressions.
Other conventional approaches include the use of Euler angles to express segment orientations, as well as quaternions and exponential maps that have desirable interpolation properties and avoid singularities inherent with Euler angles. Euler angles have degrees of freedom that are natural analogs to motion descriptions such as twist, flexion-extension, and abduction-adduction in human movement. One disadvantage of Euler angles, however, is that the choice of parameterization is restricted for particular orientations.
In addition, specialized or specifically targeted models are conventionally used to represent complex biomechanical characteristics associated with particular types of joints. Physiological joints have been shown to have many complexities that are often neglected in graphical models. For example, biomechanists routinely specify joints with several non-orthogonal, arbitrary axes of rotation that are better aligned to bone articulation. Many joints have translational components and changing centers of rotation, including the knee that is traditionally simplified as a single DOF hinge joint. In joints like the shoulder, the closed loop consisting of the clavicle, scapula, and thoracic surface of the rib cage creates a coupling between the articulations of all these joints. Several conventional approaches model, this situation by enforcing a constraint on the scapula to stay on the surface of an ellipsoid approximating the rib cage. Techniques that use specialized models to represent biomechanical or physiological complexity can be difficult for users to configure and to control.
Further, the majority of commercial software that provides visualization features for three-dimensional (3-D) modeling does not support the complexity of multiple DOF joints or specialized models. The motion of a single joint is generally restricted to the relative motion between two adjacent segments, rather than a coordinated set of motions over several segments.
One example of a system that lacks coordination over several segments is described in A. Maciel et al., “Anatomy-based joint models for virtual human skeletons,” Proceedings of Computer Animation 2002, pp. 165-172. Maciel et al. describes a model that incorporates joints that can translate and rotate together on a plane and have joint limits that dynamically change with the DOF of any joint. Each DOF is associated with an axis of rotation or translation for a single segment.
Other conventional approaches lack the generality needed to include changing joint centers, surface constraints, and joint sinus cones for joint limits on ball-and-socket joints. For example, the Peabody system described in N. Badler et al., “Virtual humans and simulated agents,” Oxford University Press, 1992, which is incorporated by reference herein in its entirety, collects joints into joint groups that have group angles to configure the joint group's segments, but lacks this generality. One problem with conventional systems that provide a high level organization to coordinate individual joints is that they are not sufficiently generalized to represent all joints accurately.
Another problem with conventional systems is that joint models are not easily exchanged or interchanged among software environments. Although standardized humanoid joint hierarchy have been defined for the purpose of avatar representation, custom joints can be added only if they do not interfere with the movement of the standard joints. Although a standard human representation is important for avatar exchange and compatibility in different software, the flexibility to define new coordinated articulations is hampered by constraints enforced by the hierarchy.
What is needed is a joint component model framework that provides modeling of joint expressions over several bone segments and provides a high level organization to coordinate the joint components. What is further needed is a framework that accommodates biomechanically accurate joints with non-orthogonal rotation axes, changing joint centers, closed loops, and intuitive configuration parameters and controls. What is additionally needed is a framework that includes intuitive controls and that provides for exchange or interchange among software environments.
One embodiment of the present invention provides a general joint component model that is exhibiting complex behavior of joints in articulated figures. The joint components are defined as a mapping of inputs to outputs that are controlled by a list of parameters. The parameters configure the joint component to describe specific features such as joint limits and axes of rotation. The joint components are modular, reusable building blocks that can be combined together in a network to implement a joint set function. Further, the joint components are capable of handling non-orthogonal, non-intersecting axes of rotation, and changing joint centers that are often found in the kinematics of real anatomical joints.
In one embodiment of the present invention, generalized coordinates are provided as input to the joint set function. The generalized coordinates represent the degrees of freedom (DOF) of the joint set function and are used to adjust the joint articulation. The joint set function produces output transformation matrices that are used to animate the bone segments. Various linear and nonlinear joint dependencies are made implicit within the framework. Therefore, the adjustment of joint articulation is done with a relatively small set of intuitive parameters compared to the number of articulations in the motions they parameterize.
In another embodiment of the present invention, the inputs to the joint set function can be designed to have intuitive interpretations to a user such as an animator, and be sufficiently compact because the dependencies and the coupling of the joints are implicitly built into the joint set function. Joint components can be reused in different contexts or extended to incorporate increasing accuracy in the joint model.
In a further embodiment of the present invention, a method for building a joint set function for a target joint is provided. One embodiment of the method includes identifying bones and articulations in the target joint, partitioning the bones and articulations into bone groups, choosing a bone group, and selecting joint components to model the articulations for the chosen bone group. Each joint component includes at least one parameter. The method further includes setting the parameter of each joint component to define the behavior of the joint component, and coupling the joint components into a network to produce the joint set function.
Further features of the invention, its nature and various advantages will be more apparent from the accompanying drawings and the following detailed description.
The accompanying drawings illustrate several embodiments of the invention and, together with the description, serve to explain the principles of the invention.
The present invention is now described more fully with reference to the accompanying figures, in which several embodiments of the invention are shown. The present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the invention to those skilled in the art.
In computer graphics applications, realistic depiction and animation of human characters is an important element contributing to the appeal and success of interactive graphics. Three-dimensional humanoids are often the physical embodiment of the interactive interface, either serving as an avatar for the player, or as representations of other autonomous characters in the virtual environment. While recent research has focused on better geometric representations, anatomy-based modeling, and pose-based skin deformation for representing and animating the exterior skin and muscle deformations of humans, the underlying articulated body representation that drives these methods has generally been unchanged since its first use in computer animation in the 1980s.
Introducing accurate biomechanical joint models to the traditional hierarchy of joint transformations can lead to improved realism in human character animation. Interactive applications such as 3-D computer games and virtual reality have benefited from improvements in graphics hardware by depicting humanoid characters with greater levels of detail in their geometric surface models. However, studies in perception of human motion with different geometric models have suggested that observers may be more sensitive to motion changes if polygonal models are used compared to stick figures. As the majority of 3-D interactive applications use polygonal models with increasing detail, observers may be more sensitive to noticing unrealistic joint motions in animations, such as in the shoulder and torso regions of the body. This is especially relevant to sport simulations where perceived athleticism is tied to the coordination of movement in virtual human players. For animation techniques that deform an outer skin model based on skeleton motion, accurate joint transformations of bone segments can lead to better performance of these methods by improving the association between skin movement and the underlying bones in the case of pose-based deformations and modeling more accurate muscle deformations in response to skeletal movement in the case of anatomy-based models.
The processes, features, or functions of the present invention can be implemented by program instructions that execute in an appropriate computing device. Example computing devices include enterprise servers, application servers, workstations, personal computers, network computers, network appliances, personal digital assistants, game consoles, televisions, set-top boxes, premises automation equipment, point-of-sale terminals, automobiles, and personal communications devices (e.g., cellular handsets/radiotelephones).
The program instructions can be distributed on a computer readable medium or storage volume. The computer readable storage volume can be available via a public network, a private network, or the Internet. Program instructions can be in any appropriate form, such as source code, object code, or scripting code.
A. System Overview
The joint set function inputs 130 correspond to the set of parameters that user modifies to control the articulation of the model. One example user of the joint set function 105 is an animator, although one skilled in the art will appreciate that programmatic methods can be used to provide joint set function inputs 130. One such programmatic method is a conventional animation algorithm.
One advantage of the joint set function 105 is that the complexity of the joint components 110, 112, 114, 116, 118, 120, 122 is hidden from the user. That is, the joint set function inputs 130 can be a set of intuitively designed parameters that effectively control complex articulations of the subject. Further features and examples of constructing models for particular joints are described below.
B. Joint Component Model
For clarity of the following description, some terminology and notation is now described. An articulated figure includes a set of segments s that can express only rigid body motion. The segments are related to each other by a hierarchy where the motion of segment s is expressed relative to its parent p in the form of a 4×4 transformation matrix spT (the superscript p can be removed when the parent segment is not referenced). A segment can have no parent, s0T, implying that its motion is relative to the world coordinate frame. The segments of a single articulated figure can be partitioned, therefore, into several hierarchical trees. This can be useful if an articulated figure contains free-floating segments.
A joint set J contains one or more segments whose configuration is described by independent degrees of freedom (DOF) or generalized coordinates (GC), qJ. For each segment in the joint set, its relative motion with its parent is described as an articulation or joint.
Mathematically, the transformation matrix for each segment's joint in J is expressed as a changing transformation matrix function, sT(qJ). For each joint set J, the joint set function 105 can be defined as a mapping of its generalized coordinates (i.e., joint set function inputs 130) to the transformation matrix functions for each segment 140. Expression 1 represents one definition of the joint set function 105.
fJ:{qJ}→{sT|s∈J} (1)
By partitioning the motion of segments of an articulated figure into several joint sets, one can control complex articulations kinematically with only a few DOFs. As described in further detail below, for example, a 24 vertebrae human spine can be implemented using three joint sets for the cervical, thoracic and lumbar regions, with each region having three DOFs for flexion/extension, side-bending and axial twisting.
According to one embodiment of the present invention, joint components 110, 112, 114, 116, 118, 120, 122 are modular, reusable building blocks that implement the joint set function 105. Each joint component 110, 112, 114, 116, 118, 120, 122 implements a cohesive function, facilitating its reuse in different contexts. The joint component 110, 112, 114, 116, 118, 120, 122 is defined in Expression 2 as a mapping where Λ is a list of parameters that configure a joint component to describe specific features such as joint limits and axes of rotation.
j79:Θ→Ω, (2)
The sets, Θ and Ω can have scalar, vector, or matrix elements and correspond to the inputs and outputs of the joint component, j79 respectively. The joint set function 105 includes a network of joint components 110, 112, 114, 116, 118, 120, 122 that is created by connecting the output(s) of one component to the input(s) of one or more other components. In another embodiment of the present invention, the joint set function 105 (and/or corresponding network) includes a single joint component 110, 112, 114, 116, 118, 120, 122. Generalized coordinates, such as joint set function inputs 130, feed into the network with transformation matrices for segments 140 produced as output. In one embodiment of the present invention, a relatively small number of simple joint components 110, 112, 114, 116, 118, 120, 122 can be combined to create a diverse array of behaviors. This framework allows new types of joint components 110, 112, 114, 116, 118, 120, 122 to be added and used with existing components with minimal coupling between modules.
C. Joint Components
Several specific examples of joint components 110, 112, 114, 116, 118, 120, 122 are described below. One skilled in the art will appreciate that one feature of the present invention is generalized, flexible representation of a network of joint components 110, 112, 114, 116, 118, 120, 122. Therefore, it is noted that joint components other than those described below can be added to one embodiment of the present invention. Also the joint components 110, 112, 114, 116, 118, 120, 122 can be modified to include additional features or functions not specifically described herein.
1. Matrix Multiplication Component
A matrix multiplication component takes as input a list of several matrices and multiplies them together to produce a single transformation matrix as output. The order of elements in the list determines the multiplication order. The output can either be the final transformation that will be applied to the corresponding joint (e.g., transformation matrix 140) or an intermediate result that will be used as input to other joint components 110, 112, 114, 116, 118, 120, 122.
One skilled in the art will recognize that mathematical operation of the matrix multiplication component is a conventional matrix multiplication procedure. As further described in K. Shoemake et al., “Matrix animation and polar decomposition,” Proceedings of Graphics Interface 1992, pp. 258-264, which is incorporated by reference herein in its entirety, many transformations can be expressed as matrix decompositions with intuitive parameters for the user.
2. One-to-Many Mapping Component
A one-to-many mapping component has a single scalar input and produces a vector of one or more joint variables, which can be interpreted as angles or translational units for revolute or translational joints. This mechanism allows a single scalar DOF to control the articulations of more than one joint. For example, in one embodiment of the present invention, a knee model can be implemented where a single generalized coordinate is the common parameter of several cubic spline functions that evaluate the Euler angle rotations and translations for the patella, fibula and tibia bones. As one skilled in the art will appreciate, non-linear piecewise cubic spline functions are beneficial for producing smooth motion.
Within the one-to-many mapping component, each element of the output vector can have its own linear or nonlinear function in its connection from the input scalar. These functions can implement different rates of change for several joint variables, unit conversions, and joint limits for one or more joints. For example, the domain [0,1] can be mapped to the limits [θmin, θmax] for a rotation angle θ.
3. Compensation Component
In a conventional hierarchical skeleton tree, the transformations of a parent segment are inherited by the child segment. However, this may produce undesirable behavior in some situations. For instance, if one wants to shrug the shoulders of a human model, the rotation in the clavicle 205 would propagate to the humerus 215, causing the humerus 215 to rotate away from the body. A user or animator may want to keep the orientations of the humerus 215 and clavicle 205 independent.
Conventional compensation components are designed to apply a corrective angle rotation that compensates for the parent's propagated rotation. In one embodiment of the present invention, the compensation component generalizes this concept to cancel out the orientation transformation (created by any rotation parameterization) of any ancestor segment (not just the direct parent) for a particular segment while maintaining connectivity at the joint with its segment by adjusting the translation of the segment. This allows a segment to have orientation with respect to the world frame.
The compensation component for a particular segment s takes as input a single transformation matrix of an ancestor segment and produces a matrix that cancels out the undesired orientation change caused by the rotation of that ancestor. Depending on how far up the skeleton hierarchy tree we wish to cancel out orientations, a segment can have a compensation component for each ancestor. The outputs of each compensation component are then multiplied using the matrix multiplication component to produce a matrix that compensates for all the undesired movements caused by a segment's ancestors up to a certain level in the skeleton tree. In the shoulder joint set illustrated in
In one embodiment of the present invention, the computation of the compensation matrix is done in two steps.
More specifically, the scapula 210, which is the parent of the humerus 215, is itself a child of the clavicle 205. In
4. Rotation Component
In one embodiment of the present invention, the rotation component can be designed to accommodate non-orthogonal, arbitrary axes of rotation with changing joint centers that are a function of generalized coordinate values. Joints with multiple DOF rotations can be created by combining rotation components, each of which produces a rotation matrix for a single axis rotation. One skilled in the art will recognize that it is assumed that the joint centers for each axis rotation are independent of the rotations about the other axes. The rotation component includes the following parameters: (1) a list of n consecutive angle intervals: [ao, a1), [a1, a2), . . . , [an−1, an]; (2) for each angle interval [ai−1, a1), a rotation center point ci=<cx, cy, cz>; and (3) a common rotation axis x.
In one embodiment of the present invention, each angle interval can have a different joint rotation center. The ability to model a changing joint rotation center is important to describe rotations accurately in the several joints including the knee and the humerus of the shoulder. Of course, a single angle interval (and corresponding joint center) can also be defined for a rotation component. With these rotation parameters and an input rotation angle a α∈[a0, an] (which can be derived from the output of other joint components), the final output rotation matrix R for the joint component is computed as defined in Expressions 3 where R, Mi, Ti, Ti−1, and Qi are all transformation matrices. Qi is computed from a quaternion representing the rotation of angle βi around axis x.
The angle βi is determined by Expression 4. The final rotation of the angle α is the cumulative result of a sequence of quaternion rotations of smaller angles βi. Each of these quaternion rotations has its own center represented in the Ti and Ti−1 translation matrices. This is one implementation of a changing joint rotation center of a joint in accordance with one embodiment of the present invention.
One skilled in the art will note that for a given α that falls into a certain interval [aj−1, aj), all Mk (0<k<j) will not depend on α since each of their βk is a constant equal to ak−ak−1. The matrices Mk can be precomputed and stored for the full angle interval rotation to reduce real-time computation demands. Furthermore, the cumulative matrix product R in Equation 1 for the jth interval can have Mj−1×Mj−2× . . . ×M1 precomputed and retrieved to update the articulated figure motion at interactive rates.
5. Dependency Component
The dependency component allows the modeling of coupling behavior between different joints within a joint set function 105. In each dependency component, a pair of joints are specified, one as the active joint a that drives the other passive joint p. The movement of a DOF of p is set to be dependent on a DOF of a through a mapping function. The actual nature of the mapping function used in the dependency component can be any linear or nonlinear function. Interpolating splines are often convenient to match dependency relations to experimental data samples. In one embodiment of the present invention, the DOF corresponds to Euler angles that define the rotation matrix of each joint. The dependency component takes two input Euler angles, one from each joint, and contains a mapping function to output a modified angle for the passive joint. Several types of mapping relationships can be implemented, including (1) one-to-one mapping; (2) one-sided bound mapping; and (3) two-sided bound mapping.
a) One-to-one Mapping
For any given DOF value of a, a DOF value for p is defined. For instance, the rotation of the scapula around an axis perpendicular to its outward surface tangent plane is almost linearly dependent on the abduction of upper arm. A linear one-to-one mapping can capture this relationship.
b) One-sided Bound Mapping
One-sided bound mapping is a one-to-one map where values of p are bounded on one side by a lower or upper limit that is a function of a DOF of a. An example of this is the dependency between the abduction of the humerus 215 and the elevation of the clavicle 205. The higher the upper arm is raised, the more restricted is the vertical movement of the shoulder's clavicle. The restriction is due to a lower limit placed on clavicle 205 elevation, which can be implemented as a one-sided bound that is dependent on the amount of abduction of the humerus 215.
c) Two-sided Bound Mapping
The value of a DOF of p is bounded on both sides by limits dependent on a DOF of a. Again using the shoulder as an example, when the left upper arm is rotating in the horizontal plane from the left side to the front right of the body, the horizontal movement of the shoulder (at the clavicle bone 205) becomes more restricted. A similar phenomenon occurs when the left upper arm is rotating to the back of the body. Because there are both upper and lower limits, a two-sided mapping is appropriate in this example case.
6. Joint Cone Component
Joint sinus cones have been conventionally used to provide a better mechanism for joint limits for ball-and-socket joints than pairs of Euler angle bounds for each joint DOF. Further details of conventional joint sinus cones are described in W. Maurel et al., “Human shoulder modeling including scapulo-thoracic constraint and joint sinus cones,” Computers and Graphics, vol. 24, no. 2, pp. 203-218, 2000, which is incorporated by reference herein in its entirety.
Subdivision rules can be used to refine and smooth the curve. For example, joint sinus cone 510 represents a refined and smoothed version of joint sinus cone 505. An additional vector vrest is defined and positioned at p so that it lies in the same direction as the bone's longitudinal axis at its rest configuration. This cone provides a way of bounding the movements of two DOF of a joint, such as abduction/adduction and flexion/extension in the humerus 215 at the shoulder. To limit the third twist DOF, an additional pair of angle bounds is associated with each control point on the curve c and the point vrest. During the refinement process, new interpolated bound pairs are computed for the new control points produced by subdivision. In one embodiment of the present invention, a pair of twist limits for any given configuration within the cone can be interpolated at run-time. For example, the joint sinus cone 515 is shaded to represent the twist limits on each point. Darker shades indicate a more restrictive limit range. Further details of refining and smoothing the curve are described in Jane Wilhelms, et al., “Fast and easy reach-cone joint limits,” Journal of Graphics Tools, vol. 6, no. 2, pp. 27-41, 2001, which is incorporated by reference herein in its entirety.
7. Scapula Constraint Component
In one embodiment of the present invention, the scapula constraint component addresses the specific situation of the scapulo-thoracic constraint in the shoulder. The scapula constrain component illustrates one example of how the component framework of the present invention can be extended for special handling of an individual joint. The scapula bone 210 glides on a curved surface defined by ribs, muscles and fatty structures. To represent this in the joint component model, a conventional ellipsoidal constraint can be used. A description of the ellipsoid constraint can be found, for example, in B. Garner, “A kinematic model of the upper limb based on the visible human project (vhp) image dataset,” Computer Methods in Biomechanical and Biomedical Engineering, vol. 2, pp. 107-124, 1999, which is incorporated by reference herein in its entirety. As shown in
Using the example illustration of
Referring to the embodiment of the present invention illustrated in
The scapula 210 can then be rotated twice to constrain these contact points on the surface of the ellipsoid. In the first rotation, a predefined vector x1 going through the joint origin is used as the rotation axis. Rotation of an angle θ around x1 brings the first reference point 220 onto the ellipsoid surface. In the second rotation, the vector connecting the joint origin and the first reference point 220 is used as the rotation axis, x2. Similarly, rotation of an angle ψ around x2 brings the second point 225 onto the ellipsoid. One skilled in the art will note that the second rotation does not change the position of the first contact point because it is on the rotation axis x2. In one embodiment of the present invention, binary search is used to find both rotation angles θ and ψ.
Although it is generally desirable in embodiments of the present invention to create joint components that can be reused, the ability to create specialized constraints can be useful to create tailored, intuitive parameters to simplify the description of complex articulations unique to a particular joint. Of course, more biomechanical detail can be added to a joint component as deemed necessary for the application.
D. Constructing Joint Component Models
Having described several example joint components, a process for connecting them into a network to construct joint set functions 105 for the segments of a skeleton is now described.
The illustrated process begins 804 with choosing 808 a target joint to model. Bones and articulations are then identified 812 for the chosen joint. The process then partitions 816 bones and articulations into separate groups of related bones to generate a set of bone groups 820. The bone groups 820 are then processed iteratively. The bone groups 820 can also be processed concurrently or in different orders as known to one of skill in the art.
At the beginning of the loop in the iterative example, a determination 824 is made whether all bone groups 820 have been processed. If all bone groups 820 have not been processed, an unprocessed bone group 820 is chosen 828. Joint components are selected 832 to model the articulations identified 812. To configure the selected 832 joint components, the parameters for each joint component are defined and set 836. Next, the selected joint components are connected 840 into a network. One technique for coupling or connecting the selected joint components is passing the input and the output data values from one joint component to the next joint component until each of the selected joint components are networked 844.
The process then returns to determining 824 whether all bone groups 820 have been processed. If all bone groups 820 have not been processed, the process repeats steps 828, 832, 836, and 840 as described above. If all bone groups 820 have been processed, the process proceeds to connect 848 all completed joint networks 844 to form a large network for the joint set function 105. The joint set function 105 is then tested 852 to determine whether the desired articulated motion is produced. One technique for testing 852 the joint set function 105 is visual display 856 of the joint motion. The joint builder can then decide 860 if the joint set function 105 accurately represents the identified 812 articulations. Visual display 856 of the joint motion can also include the joint builder adjusting a slider, manipulator, or other user interface control (e.g., specifying a range of values) to test 852 the joint motion.
If the joint motion is not desirable, then the joint builder can modify 868 the parameters, choice of joint components, or couplings in the joint component network. The process then retests 852 the joint set function 105. If the joint motion is desirable, the joint component model can be stored 864 in a data file 872. After storing 864 the joint component model, the process ends 786. One skilled in the art will recognize that data file 872 can be a database record or other file type in any appropriate format.
In one embodiment of the present invention, joint component models are stored 864 in extensible markup language (XML) file format. XML format can be an efficient and convenient way of providing ease of interchange among software environments or application software packages. One example XML formatted joint component model is provided below in the Appendix. In the example, the joint components 110, 112, 114, 116, 118, 120, 122 include tags and parameters.
E. Computing System
The memory 920 can be a conventional memory device, for example, random access memory (RAM), read only memory (ROM), a fixed media, or a removable media (e.g., a diskette, a compact disc ROM (CD-ROM), a compact disc recordable disc (CD-R/W), or a digital versatile disc (DVD)).
The memory 920 includes a number of functional modules: a joint component module 925, a configuration module 930, a joint set function module 935, and a user interface module 940. Exemplary functionality for each of these modules 925, 930, 935, 940 is now described in further detail.
The joint component module 925 includes program instructions and/or data for implementing the joint components 110, 112, 114, 116, 118, 120, 122. More specifically, the joint component module 925 performs calculations and data manipulations for the matrix multiplication component, the one-to-many mapping component, the compensation component, the rotation component, the dependency component, the joint cone component, and the scapula constraint component.
The configuration module 930 includes program instructions and/or data for configuring the parameters of the joint components 110, 112,. 114, 116, 118, 120, 122 and for networking the joint components into the joint set function 105. The configuration module 930 can interface with each of the joint component module 925, the joint set function module 935, and the user interface module 940.
The joint set function module 935 includes program instructions and/or data for implementing one or more joint set functions 105. The joint set function module 935 receives the joint set function inputs 130 (e.g., generalized coordinates or degrees of freedom) as input and provides transformation matrices for segments of the joint set 140 as output.
The user interface module 940 includes program instructions and/or data for receiving and for processing user, developer, or joint builder input. For example, the user interface module 940 can presents a display to the user that the user can use to manipulate the jet set function inputs 130.
One skilled in the art will recognize that other functional units or modules can be used to implement the features or functionality of the present invention. That is, the functional units shown in
F. Example Results
Example component frameworks are now described for two complex joints: the spine and the shoulder. To produce the following example results, initial estimates of parameter data for the joints were determined from literature on joint physiology, such as I. Kapandji, “The Physiology of the Joints: The Trunk and the Vertebral Column, 2d ed., vol. 3, Churchill Livingstone, 1982, and I. Kapandji, “The Physiology of the Joints: Upper Limb, 1st ed., vol. 1, Churchill Livingstone, 1982, both of which are incorporated by reference herein in their entireties.
The joint model including joint components 110, 112, 114, 116, 118, 120, 122 and corresponding joint set functions 105 can be implemented as program code modules using the C++ programming language. The program code modules can be adapted for use as plug-ins for the Maya 3-D modeling software, which is commercially available from Alias/Wavefront of Toronto, Ontario, Canada. The Maya environment provides interactive placement of the bones and adjustment of joint parameters. Maya's advanced modeling environment allows articulation of joint sets to be evaluated interactively or visually. The parameters for the joint components can be exported in an XML-based file format that can be loading into other tools, including an OpenGL-based application software. In one embodiment of the present invention, the OpenGL-based application software can accept as input a joint component model described in an XML-based file format and build the joint component network described therein. Once the network is built, the joint model is displayed with 3-D graphics using OpenGL and the joint model can be manipulated by adjusting generalized coordinates with an input device (e.g., a mouse). In one embodiment of the present invention, the OpenGL-based application software can be used to test the interchangeability of the XML-based joint model between software environments. A conventional Intel Pentium II 933 MHz general-purpose computer with Nvidia GeForce4 graphics hardware can achieve interactive display rates with Maya.
1. Spine
There are twenty-four movable vertebrae in the spine of a human. According to their position and functionality, they are divided into three joint sets: the cervical region (seven vertebrae in the neck), the thoracic region (twelve vertebrae in the thorax), and the lumbar region (five vertebrae in the abdomen). For the thoracic joint group, all the ribs and the sternum are included in the joint model to create the rib cage. For all three spine joint sets, the same type of joint function is used. The difference between them is the joint parameters given for each joint group, where the amount of rotation in the thoracic vertebrae is considerably less than the cervical and lumbar regions.
For example, in the cervical region, the flexion/extension maps to the following angle rotation ranges (in degrees) for the seven vertebrae: C7[−13.2,5.5], C6−7.5,5.5], C5[−4,5.5], C4[−4.6,5.5],C3[−8,6.5],C2[−5.5,6.5],C1[−18.5,6.5]. The conversion is a linear map between [−1,1 and a pair of vertebra-specific angle limits [min, max]. Each angle is then sent to a specific rotation component Rn for the vertebra to which it corresponds. The output matrices of the rotation components for each DOF are directed into a matrix multiplication component Mn to generate the final transformation for the vertebrae C1 through C7.
The implementation of joint set functions for the thoracic and lumbar are similar to the cervical vertebrae with the exception that the thoracic group contains ribs attached to each vertebra. Because the rib cage creates a closed chain with the spine and sternum, it tends to resist thoracic spine movement that otherwise causes the individual ribs to rotate away or into each other during lateral bending. The rotation of a rib can be set by parameters to be dependent on the amount of motion of its attached thoracic vertebra to maintain the overall shape of the rib cage. Intuitively, one skilled in the art will appreciate that the ribs are rotated in a direction opposite to that of the spine's rotation with the ribs rotating less than spine. Therefore, the axes of ribs can defined to be opposite to those for vertebrae and the axes of the ribs can have smaller rotational limits. One skilled in the art will note that for more accurate deformations of the rib cage, a custom joint component can be designed for the joint component model of the present invention.
In this example of a spine model, each joint set has three DOF of rotation for flexion/extension, lateral bending, and twist, making a total of nine DOF to control the entire spine and the rib cage. This is considerably fewer DOF than the total number of articulations achievable in our model because various dependencies are built into the model implicitly. The fewer DOF provide a lightweight and intuitive control mechanism, which balances accuracy with control simplicity.
2. Shoulder
The shoulder includes four articulations (the scapulo-humeral joint 260, the acromio-clavicular joint 255, the sterno-clavicular joint 250, and the sliding scapulo-thoracic joint 265). Several bones are involved in these joints: the clavicle 205, scapula 210, humerus 215, sternum and rib cage. In addition to the bone articulations, ligaments, cartilage and muscles also can play important roles in the shoulder to create coupling behavior and dependencies among the shoulder's joints.
A shoulder joint model was built that includes the four joints and three bones. Its joint set function has five DOF, of which three control the scapulo-humeral joint 260 (flexion/extension, abduction/adduction and twist of the humerus 215) and two control the stemo-clavicular joint 250 (vertical and horizontal rotation of the clavicle 205). The acromio-clavicular joint 255 has zero controllable DOF because its movement can be fully determined by movements of the other two joints and the ellipsoidal surface constraint with the rib cage. The joint set function 105 outputs three transformation matrices 140 for the three joints respectively. Inside, the joint set function 105 has three parts, each in charge of one joint.
The combined result of these two matrices, together with the twist angle, then go through a joint cone component, identified as L in
One skilled in the art will recognize that the three functional portions illustrated in
A component joint model in accordance with one embodiment of the present invention includes a simple control interface, and the joint components can be computed for real-time interactive applications using conventional computing devices. The component joint model can be used to produce realistic joint movements. A joint designer or builder is enabled through the framework to consider different competing criteria such as desired level of detail, joint parameter design, and computational complexity to develop suitable models for the desired target application.
One skilled in the art will appreciate that due to the complexity of several joint set functions 105, it may not always be possible to compute their analytic derivatives for inverse kinematics or gradient-based optimization. However, as the joint set functions 105 produce deterministic joint transformations, finite difference techniques can be used to estimate their derivatives. An embodiment of the framework can also be used to find subject-specific parameters that enable the same joint set functions 105 to be customized for different individuals or animals. The joint components described above do not represent a complete set for modeling all joints at every level of accuracy. In particular, through the design of new components or new joint sets, various other joint models can be incorporated into the framework of the present invention.
Having described embodiments of joint component framework for modeling complex joint behavior (which are intended to be illustrative and not limiting), it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in the particular embodiments of the invention disclosed that are within the scope and spirit of the invention as defined by the appended claims and equivalents.