The Research Of Facial Expression Recognition Based On The Fractional Cosine Transform

In recent years, automatic facial expression recognition to become an active area of research can be widely used in pattern recognition, multimedia information detection and analysis, human-computer interaction, security, intelligent monitoring systems, etc., with excellent prospects for development. Facial features extraction is the key issues for this problem are analyzed and Fractional Cosine transform which is a new time-frequency analysis tools applied to the feature extraction process. As a time-frequency analysis method, the Fractional Cosine transform (FRCT) with the Fractional Fourier transform (FRFT) has some kind of transformation relationship, it is an extension of traditional Cosine transform. It makes the signal processing from a single frequency domain transform to time-frequency domain because it increases the free parameter of rotation angle. So it is more suitable for facial expression feature extraction. FRCT not only can meet the performances as the traditional theories but also with good flexibility. As a result, the FRCT is likely to get better results in the area of facial expression recognition than the conventional extraction methods. The main content and contributions are:1. It provides a brief introduction the significance of FRCT and studies the one-dimensional discrete Fractional Cosine transform algorithm. Based on the theoretical basis and physical meaning, it can expand the one-dimensional transform into the form of two-dimensions (2D-FRCT), and analyze the discrete algorithm of2D-FRCT. Furthermore, it makes the relevant description on the natures of2D-FRCT.2. It introduces the databases of facial expressions in the simulation experiment and pre-processing of the facial expressions. Besides, we analyze the properties of2D-FRCT in the area of image processing and get the regulation on the extraction of the features from the facial images.3. It analyzes the relationship between the rotation angle and the recognizing results and compares the results in this paper with the methods of Gabor, which shows the effectiveness of the algorithm in this paper. In additional, it describes the Fisher classification algorithm and builds the model of recognizing human facial expression based on the method of2D-FRCT and gets a quantitative analysis.