Action Recognition Based on Non-parametric Probability Density Function Estimation

  • Authors:

  • Yuta Mimura;Kazuhiro Hotta;Haruhisa Takahashi

  • Affiliations:

  • The University of Electro-Communications, Tokyo, Japan 182-8585;The University of Electro-Communications, Tokyo, Japan 182-8585;The University of Electro-Communications, Tokyo, Japan 182-8585

  • Venue:
  • ISVC '09 Proceedings of the 5th International Symposium on Advances in Visual Computing: Part II
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In recent years, it is desired that many surveillance cameras are set up for security purpose. If the automatic detection and recognition system of crimes and accidents, we can prevent them. Researchers work actively to realize the automatic system. Cubic Higher-order Local Auto-Correlation (CHLAC) feature has the shift-invariant property which is effective for surveillance. Thus, we use it to realize action recognition without detecting the target. The recognition of a sequence x=(x 1 ,...,x T ) can be defined as the estimation problem of posterior probability of it. If we assume that the feature of certain time is independent of other features in the sequence, the posterior probability can be estimated by the simple production of conditional probability of each time. However, the estimation of conditional probability is not easy task. Thus, we estimate the conditional probability by non-parametric model. This approach is simple and does not require the training of model. We evaluate our method using the KTH dataset and confirm that the proposed method outperforms conventional methods.