Paper Summaries Any takers? Articulated Figures III Motion Capture - PDF document

Paper Summaries � Any takers? Articulated Figures III Motion Capture Projects Projects � Question about presentations � Exam Week Presentation day: � We have approx 19 projects � Thursday, March 2nd � Presentations: � 15 minutes (max) per project � 12:30pm -- 2:30pm � Sign up for time via e-mail (first come / first served) � Midquarter report due next Wednesday � Room 70-3445 � Grad Reports � We have 10 � Please indicate topic by end of day (e-mail) � 20 minutes per presentation � Week 9 Assignments Plan for today � Assignment 1: Keyframing � Next 2 weeks: Articulated Figures � Due last Wednesday � Monday: Forward Kinematics � Grading in progress (mostly done) � Assignment 2: Dynamics � Wednesday: Inverse Kinematics � Due Wednesday, January 25th (Wed) � Monday: Motion Capture � Assignment 3: Group motion � Wednesday: Advanced algorithms � Particle systems / Flocking: Due Feb 6th � Assignment 4: moCap � Then � To be given today � Due February 20 � Monday: Character animation 1

Motivation Films Motivational Film � Early examples of motion capture � Brilliance (1985) � It’s all Apple’s fault! � Robert Abel & Associates � first entirely computer generated TV ad � Debuted at Super Bowl XI X (1985) � Who said motion capture was a new technology? Motivational Film Plan For Today � Don’t Touch Me (1989) � Topics � Motion Capture � Diana Walczak and Jeff � Walking Kleiser � Synthespian, (Synthetic � Assignment # 4 (maybe) Thesbians) � Dojo – First female Synthespian � http://www.kwcc.com/ Role of Animation Motion Capture � Degrees of freedom � The idea between motion capture � Number of parameters � You want realistic human motion? whose values must be � Go to the source defined in order to fully � No, not Newton this time… position the articulated � Use an actual human figure � Purpose of animation � Provide values to each of the DOF for each time step. 2

Rotoscoping Motion Capture in CG Used to trace motion of live � � First introduced by Abel and Associates actors, frame by frame into for “Brilliance” an animation Invented by Max Fleischer in � 1916 First used in Koko the Clown � cartoons Used extensively by Disney in � Snow White Motion Capture Motion Capture � What motion capture gives us: � Sampled values for each DOF in time. � Since captured directly from human motion � Subtleties of motion come for free. � Difficult for an animator to keyframe these subtleties Watt/Policarpo Motion Capture Optical Motion Capture � Motion Analysis Corp � Types of motion capture systems � I Robot � Optical � Incorporate directionally-reflective balls referred to as � Final Fantasy markers which attach to the performer. � Entirely motion capture � Three (at least) digital video cameras that track markers. � Polar Express � video � Provides most flexibility for performers. � Problem: Markers may be occluded from cameras views. 3

Motion Capture Prosthetic Motion Capture � Types of motion capture systems � Gypsy 4 � Prosthetic � By MetaMotion � set of armatures attached all over the performer’s body � The armatures are connected to each other by using a series of rotational and linear encoders. � Accurate, though cumbersome for the performer Motion Capture Systems Motion Capture Systems � Types of motion capture systems � Types of motion capture systems � Acoustic � ElectroMagnetic � An array of audio transmitters are strapped to � Much like acoustic except magnetic various parts of the performers body. transmitters/receivers used instead of acoustic � Three receivers are triangulated to provide a point in 3D space. � No occlusion problem. � Cables are cumbersome to performers � No occlusion problem. � Though now wireless solutions are available � Cables are cumbersome to performers � Metal / other magnetic fields may interfere. � Ambient sound may interfere Motion Capture Systems Electromagnetic Motion Capture � MotionStar 2 � Types of motion capture systems � Ascension � Fiber Optic Sensors Technologies � Flexible FO sensors strapped to various parts of the performers body. � Sensors can directly measure joint rotations � Used in conjunction with electromagentic sensor for head and torso. 4

Fiber Optics Motion Capture Capturing Human Motion � Shapewrap II � Minimal set of recording points � Measurand Frey, et. al Motion capture Systems Motion Capture Systems � Challenges: � Challenges: � Even if motion capture data was perfect, we still � Signal is not perfect have the following challenges: � Noisy � Re-use – use the motion for a slightly different purpose � missing data � Creating impossible motion – Motion capture won’t do it, but may be desired in animation � not perfectly aligned with joints � Change of intent – we can’t always predict what motion � Retargeting we will need � Data is only valid for virtual character who � Take Home Message: Motion Capture captures a possesses same scale as real character. particular, single motion. Motion Capture Systems Motion Capture Data � So what CAN we do with motion � Examples capture data? � From Eurographics Computer � We can Animation and Simulation EGCAS'96 � speed up � slow down � From The Polar Express � time warp � Motion warp � However, one must remember that Captured data is Sampled Data. 5

Sampling Theory Sampling Theory � Signal - function that conveys � Point Sampling information � start with continuous signal � Audio signal (1D - function of time) � calculate values of signal at discrete, � Image (2D - function of space) evenly spaced points (sampling) � Continuous vs. Discrete � convert back to continuous signal for � Continuous - defined for all values in range display or output (reconstruction) � Discrete - defined for a set of discrete points in range. Sampling Theory Sampling Theory � Sampling can be described as creating a set of values representing a function evaluated at evenly spaced samples = ∆ = K f n f ( i ) i 0 , 1 , 2 , , n ∆ = interval between samples = range / n. … Foley/VanDam 0 1 2 n Sampling Theory Sampling Theory � Sampling Rate = number of samples per unit � Example -- CD Audio � sampling rate of 44,100 samples/sec � ∆ = 1 sample every 2.26x10 -5 seconds = 1 f ∆ 6

Sampling Theory Sampling Theory � Rich mathematical foundation for � Spatial vs frequency domains sampling theory � Most well behaved functions can be described as a sum of sin waves (possibly � Hope to give an “intuitive” notion of offset) at various frequencies these mathematical concepts � Describing a function by the contribution (and offset) at each frequency is describing the function in the frequency domain � Higher frequencies equate to greater detail Sampling Theory Sampling Theory � Nyquist Theorum � A signal can be properly reconstructed if the signal is sampled at a frequency (rate) that is greater than twice the highest frequency component of the signal. Foley/VanDam Sampling Theory Sampling Theory � Nyquist Theory � Example -- CD Audio � Said another way, if you have a signal with � sampling rate of 44,100 samples/sec highest frequency component at f h , you � ∆ = 1 sample every 2.26x10 -5 seconds need at lease 2f h samples to represent this signal accurately. 7

Sampling Theory Sampling Theory � Nyquist Theory -- examples � Aliasing � CDs can accurately reproduce sounds with � Failure to follow the Nyquist Theorum frequencies as high as 22,050 Hz. results in aliasing . � Aliasing is when high frequency components of a signal appear as low frequency due to inadequate sampling. Sampling Theory Sampling Theory � Aliasing - example � Annoying audio aliasing applet � Example of aliasing in animation. Foley/VanDam Sampling Theory Sampling Theory � Anti-Aliasing � Fourier analysis � What to do in an aliasing situation � Given f(x) we can generate a function F(u) which indicates how much contribution � Increase your sampling rate (supersampling) each frequency u has on the function f. � Decrease the frequency range of your signal (Filtering) � F(u) is the Fourier Transform � Fourier Transform has an inverse � How do we determine the contribution of each frequency on our signal? 8

Sampling Theory Sampling Theory � Fourier Transforms � How do we calculate the Fourier Transform? f(x) � Use Mathematics Inverse � For discrete functions, use the Fast Fourier Fourier F(u) Fourier Transform Transform algorithm (FFT) Transform f(x) Sampling Theory Sampling Theory � Filtering -- Frequency domain � Anti-Aliasing � Place function into frequency domain F(u) � What to do in an aliasing situation � simple multiplication with box filter S(u) � Increase your sampling rate (supersampling) � Decrease the frequency range of your signal − ≤ ≤ ⎧ (Filtering) 1 , when k u k = ⎨ S ( u ) � Since we already have the data sampled, ⎩ 0 , elsewhere we can’t supersample motion capture data � Thus, we need to filter Sampling Theory Sampling Theory � Filtering - frequency domain � Filtering -- Spatial Domain � Convolution ∞ ∫ = ∗ = τ − τ τ h ( x ) f ( x ) g ( x ) f ( ) g ( x ) d − ∞ Taking a weighted average of the neighborhood around each point of f, weighted by g centered at that point. Foley/VanDam 9