Share Email Print
cover

Proceedings Paper

Bayesian of inductive cognition algorithm for adaptive classification
Author(s): Longcun Jin; Wanggen Wan; Bin Cui; Yongliang Wu
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

In this paper, we proposed a Bayesian of inductive cognition algorithm using in virtual reality multimedia classification. We present a Bayesian of inductive cognition algorithm framework model for adaptively classifying scenes in virtual reality multimedia data. The Multimedia can switch between different shots, the unknown objects can leave or enter the scene at multiple times, and the scenes can be adaptively classified. The proposed algorithm consists of Bayesian inductive cognition part and Dirichlet process part. This algorithm has several advantages over traditional distance-based agglomerative adaptively classifying algorithms. Bayesian of inductive cognition algorithm based on Dirichlet process hypothesis testing is used to decide which merges are advantageous and to output the recommended depth of the scenes. The algorithm can be interpreted as a novel fast bottom-up approximate inference method for a Dirichlet process mixture model. We describe procedures for learning the model hyperparameters, computing the predictive distribution, and extensions to the Bayesian of inductive cognition algorithm. Experimental results on virtual reality multimedia data sets demonstrate useful properties of the Bayesian of inductive cognition algorithm.

Paper Details

Date Published: 11 July 2009
PDF: 6 pages
Proc. SPIE 7490, PIAGENG 2009: Intelligent Information, Control, and Communication Technology for Agricultural Engineering, 74902Y (11 July 2009); doi: 10.1117/12.836806
Show Author Affiliations
Longcun Jin, Shanghai Univ. (China)
Wanggen Wan, Shanghai Univ. (China)
Bin Cui, Shanghai Univ. (China)
Yongliang Wu, Shanghai Univ. (China)


Published in SPIE Proceedings Vol. 7490:
PIAGENG 2009: Intelligent Information, Control, and Communication Technology for Agricultural Engineering
Honghua Tan; Qi Luo, Editor(s)

© SPIE. Terms of Use
Back to Top