Data-driven digital entertainment: a computational perspective
Yueting Zhuang · Frontiers of Information Technology & Electronic Engineering · 2013
Today massive collections of data can be obtained across different sources (or domains), e.g., the depth data from Kinect, the geometrical data from scanning devices, the imagery/video data from cameras, and the motion data from mocap devices.Since heterogeneous data may have different discriminative powers and are intrinsically complementary for certain tasks, it is desirable to leverage all the information available in digital entertainment.For example, the acquired 3D geometry and texture are jointly exploited to construct the colored 3D environment models; high-resolution geometry and motioncaptured data are obtained to synthesize and re-target facial animations; both visual and acoustical features are contextually applied to classification.Therefore, it poses a significant challenge for the appropriate utilization of the different varieties of heterogeneous data in digital entertainment.The intuition behind this challenge can be considered from a data-driven computational perspective.The underlying formalism of the data-driven computation is to derive mathematical expressions directly from the database with a minimum of a priori hypotheses.The typical approaches of data-driven computation are generative models (e.g., Gaussian mixture model, hidden Markov model, Naive Bayes, and latent Dirichlet allocation) and discriminative models (e.g., support vector machines, boosting, conditional random fields, and neural networks).Generative models are desirable when they capture