Research On Facial Expression Recognition Based On LBP Algorithm

In recent years,facial expression recognition has become a very popular research field.Facial expression recognition develops slower than other biometric technologies(iris recognition,fingerprint recognition,etc.),and is not widely used.However,in the field of human-computer interaction,expression recognition plays a very important role.Therefore,facial expression recognition has become the focus of attention of many scholars and experts at home and abroad,and some achievements have been achieved.Psychological research indicates that the dynamic characteristics of eyes and surrounding areas are highly correlated with emotional fluctuations,and both eyes and surrounding areas have specific changes with the fluctuations of expressions.Using eye region as an input signal is suitable for low-cost,non-invasive facial expression recognition systems.This work proposes two frameworks towards facial expression recognition.The first framework uses Local Binary Patterns(LBP) as the feature extractor on grayscale eye region images.The proposed system consists of three main modules: Pre-Processing,Feature Extraction,and Classification.In the pre-processing stage,any present facial region in the image is first localized using a selected luminance-based skin-tone detector,and then the eye region within the localized facial region is normalized,aligned,and extracted.Next,the system was evaluated with different LBP parameterized two different classifiers: nearest neighbor with chi-square distance metric,and SVMs with a linear kernel.The results of the study confirmed that this improved the applicability of the LBP operator from facial expression in the image of the eye region.It is successful to recognize facial expressions in the eye region by using LBP algorithm.The system is also able to generalize well across different environment conditions.In the second proposed framework,a color-based approach to facial expression recognition from the color eye region is explored using Local Color Vector Binary Patterns(LCVBP).The LCVBP feature-based system consists of five main modules: Pre-Processing,Feature Extraction,Feature Selection,Training,and Classification.LCVBP,a recent and an effective feature descriptor,extends the idea of grayscale LBP features to include color information extracting two sets of features from color images: color norm patterns(CNP),and color angular patterns(CAP).Through our preliminary evaluation on a dataset,we have exhibited its potential and backward compatibility(with grayscale LBP)as a suitable feature descriptor for expression recognition on eye or facial images.LCVBP extend the traditional LBP by incorporating color information extracting a rich and a highly discriminative feature set,thereby providing promising results in facial expression recognition.