Facial expression recognition based on radon and discrete wavelet transform using support vector machines

Extracting facial features remains a difficult task because of unpredictable of facial features largely due to variations in pixel intensities and subtle changes of facial features. The Radon transform inherits rotational and translational properties that are capable of preserving pixel intensities variations and also is used to derive the directional features. Thus, this paper presents a new pattern framework for facial expression recognition based on Radon and wavelet transform using Support Vector Machines classifier to recognize the seven facial emotions. Firstly, the pre-processed facial images are projected into Radon space via Radon transform at a specified angle. Then, the obtained Radon space or sinogram that represent the facial emotions is subjected to wavelet transform. In this framework, the Radon space is decomposed into four sub-band at a different level of decomposition. The approximate coefficients sub-band are independently extracted and used as intrinsic features to recognize the facial emotion. To reduce the data dimensionality, principal component analysis (PCA) is applied to the extracted features. Then, the Support Vector Machines (SVM) classifier is adopted as a classifier to classify seven (anger, disgust, fear, happiness, neutral, sadness and surprise) facial emotions. To evaluate the effectiveness of the proposed method, the JAFFE database has been employed. Experimental results show that the proposed method has achieved 93.89% accuracy