How Research Helps Animators Create More Realistic Characters
TOM SHANNON PHD,FOUNDER
Perception by humans can best be described as organizing, identifying and interpreting chemical and physical nervous stimuli from the senses to understand their environment that is also shaped by a recipient’s expectations, cognition, and attention to translate the information into high level cerebral functions such as recognition and situational awareness.
Researchers drawing on archaeological findings during the course of the Pleistocene age have started investigating social evolution and the perception of familiarity, similarity and attraction among groups. Could this work help to explain a phenomenon first reported in the 1970’s by Masahiro Mori of the Tokyo Institute of Technology when he proposed the concept of the ‘uncanny valley’ to describe his observation that as robots appear more and more human, they become more appealing until a person’s affinity or positive emotional responses are countered by stronger feelings of unease as they descend into his valley?
The concept has been found to be equally valid when applied to the creation of characters or avatars in the film, games and virtual reality market sectors.
THE SCIENCE OF EXPECTATION
Contemporary neuro-imaging studies are now employing naturalistic sources such as film, avatars and androids to provide stimuli to accelerate research into a better understanding of how the brain processes complex situations and perception whilst maintaining experimental control. An exaggerated caricature can be accepted as familiar if the viewer expects it to be less human-like but still retaining recognizable characteristics. An example is the high-quality characterizations achieved in the two Disney© Maleficent films, made even more complex by requiring a seamless transition between conventional performances by well-known actors and their anthropometrically different fairy avatars. Similarly, Rémi Brun has employed his extensive special effects and motion capture experience to create amazing and seminal light sculptures, using small numbers of carefully placed and synchronized lights, to create compelling illusions of the motion of bipeds and quadrupeds in a viewer’s imagination.
The goal of every animator remains to elicit a desired emotional response from audiences to keep them engaged with their story. Technologies now exist to create realistic human-like characters to a degree where the subtlety of gait, body language, and expressions of emotion can be perceived by most current observers to be aligned to that presented by physical actors, with similar behavioral, physiological and neuronal stimulations, in particular empathy. We have all identified weaknesses in the anatomical realization of humanoid characters that even recently would have been considered innovative and ground-breaking. Producers, creators and manufacturers can continue to expect ever increasing demands for greater reality within imagery from their viewing public in the years to come.
CROSSING DISCIPLINES
A recent example of the application of basic clinical research outcomes to directly benefit animators was the work undertaken by Vicon’s Dr Jean-Charles Bricola on modeling the human spine to further improve motion capture body tracking and solving. He chose to describe the spine as seven hardy-spicer joints from the vertebra prominens to an estimation of the location of the sacrum between the two palpable posterior superior iliac spines. He added slider joints between each bone to account for the actions of the spinal muscles, in particular the erector spinae, upon the vertebral column. Jean-Charles faced the challenges to correlate the poses of the underlying vertebral bodies to track spinal motions using a sparse set of surface markers and to then verify that his model responded as expected.
The solutions came from research into quantifying changes in back surface shape among patients diagnosed with Adolescent Idiopathic Scoliosis, a nasty disease where a spinal deformity can progress over time by simultaneously curving towards the arms and rotating. In some cases, it can force the ribs outwards causing a characteristic hump. The recent discovery of the skeleton of Richard III confirmed he had suffered from scoliosis, described unkindly by William Shakespeare that he was “rudely stamp’d”, “deformed, unfinish’d”. Suspicion continues to linger widely that the playwright, with a mastery of perception, intentionally depicted the last Plantagenet king as a “hunchbacked spider” to curry favour with his Tudor masters and audiences.
Jean-Charles first challenge was addressed by drawing on published research undertaken by Dr Alan Turner- Smith of the Oxford Orthopaedic Engineering Centre in 1988 where he measured a selection of cadaver spines to create a mathematical relationship to calculate the offset in the Sagittal plane between each spinous process and locus of the vertebral body acquired from typically developed subjects and later patients.
A small number of scoliosis patients require surgery to stabilize the progression of the disease with a growing emphasis within the clinical community to include an assessment of an individual’s quality of life, interests and personal goals when planning a treatment. The intervention can introduce some physical impairment as the procedures rigidly fix vertebral bodies that are normally capable of inter segmental motion and affecting spinal global ranges of motion. A number of researcher centers have presented papers where they compared pre and post-operative forward and lateral trunk flexion and transverse rotation, concluding that post-operative range of motion was reduced in both fused and unfused regions above and below the surgery when compared against that acquired from un-affected subjects. Jean-Charles was given access to anonymized motion capture trials acquired from thirty typically developed adults. All subjects followed a common exercise protocol which he was able to use to successfully compare and to verify the performance of his model against the published normative clinical results.
LATERAL INFLUENCE
The experience gained through this and other studies towards improved body tracking and solving is now incorporated into the latest versions of the Shōgun software. This is just one example of how multi-faceted and lateral reviews of fundamental research can have a positive impact on the development of computer-generated imagery as it continues to evolve.