Using Fourier Analysis To Generate Believable Gait Patterns For Virtual Quadrupeds

Achieving a believable gait pattern for a virtual quadrupedal character requires a significant time investment from an animator. This thesis presents a prototype system for creating a foundational layer of natural-looking animation to serve as a starting point for an animator. Starting with video of an actual horse walking, joints are animated over the footage to create a rotoscoped animation. This animation represents the animal’s natural motion. Joint angle values for the legs are sampled per frame of the animation and conditioned for Fourier analysis. The Fast Fourier Transform provides frequency information that is used to create mathematical descriptions of each joint’s movement. A model representing the horse’s overall gait pattern is created once each of the leg joints has been analyzed and defined. Lastly, a new rig for a virtual quadruped is created and its leg joints are animated using the gait pattern model derived through the analysis