Spontaneous Micro-Expression Recognition Using 3DCNN on Long Videos for Emotion Analysis

Research related to the recognition of micro-expression continues to develop yearly. New ways are always found, from the preprocessing stage and feature extraction to the classification stage, with learning models to recognize emotions and measure their performance. Based on previous studies, many problems still need to be encountered in recognizing micro-expression in long video sequences. This study proposes a new framework for the introduction of micro-expression, which consists of several stages of the process, namely the stages of preprocessing tiered novels and the Three Dimensional Convolutional Neural Network (3DCNN) learning model and ends with measuring its accuracy. By implementing this preprocessing stage, the micro-expression features found on the face are protected from damage and loss. In addition, some experiments on the 3DCNN learning model have increased the accuracy of micro-expression recognition. We used the SMIC dataset to evaluate the capabilities of the proposed micro-expression recognition framework. The accuracy results obtained in this dataset are 81.70%. These results demonstrate the effectiveness of the proposed new micro-expression recognition framework.