Facial expression recognition based on ISOMAP
Zhiyong Liu · 2011
In artificial intelligence, pattern recognition, machine learning and many other fields of study, people often have to face a problem is the growth of the pattern space dimension, this is the so-called "curse of dimensionality". Dimension reduction can effectively avoid the "curse of dimensionality"; improve the performance of subsequent classifiers and computing efficiency, noise suppression, saving computing and storage resources. Manifold learning algorithms is a new dimension reduction method, the objective is to discover the low-dimensional manifold structure which embedded in high-dimensional data space, and gives an effective low-dimensional expression. ISOMAP algorithm with a large number of good properties, it can ensures that the data's structure and mutual relations in high-dimensional space can be well retained in the low-dimensional space, based on the feature, the ISOMAP can be applied to the facial expression recognition, it achieved a good recognition rate.