Features representation by multiple local binary patterns for facial expression recognition
Li Wang, Ruifeng Li, Ke Wang · 2014
To recognize expressions conveniently and effectively, an enhanced feature representation method is proposed for facial expression recognition. Local binary pattern histogram Fourier (HF-LBP) features is used to represent facial expression features. Multiple HF-LBP features are extracted to form recognition vectors for facial expression recognition in the approach, which include sign and magnitude LBP in the completed LBP scheme with multiple radii and different size neighborhoods to achieve enough features. It represents images from different scales and directions in the local neighborhood by overall considerations from the aspect. K-nearest neighborhoods classifier is applied for expression recognition after representing facial features using HF-MLBP. Comparisons are made with other extension LBP operators to evaluate the approach. The experimental results show that our method has good performance in facial expression recognition.