Micro?expression recognition by two?stream difference network

Facial micro-expression is a superposition of features and identity information subject. For the problem interference in recognition, this study proposes new method for facial recognition by de-identity information, called two-stream difference network (TSDN). First, encoder-decoder trained convolutional neural network, where input stream image, image. The image apex onset sequence. are recorded intermediate layer stream, while contains only Then, removed but stored stream. Given sequence micro-expressions, TSDN model learns that stores expression Two public spontaneous data sets (SMIC CASME II) employed our experiments. experiment results show can achieve superior performance recognition.