Abstract : Analysis of facial expressions in real time is a difficult and interesting topic that has wide-reaching effects in many fields, such as human-computer interaction and data-driven animation. For facial expression-based emotion recognition, it is very important to be able to get a reliable representation of the face from photos of high-quality source material. In this study, we do a thorough empirical study of Local Binary Patterns, a way of representing faces that uses statistical local features to recognise facial expressions no matter who is looking at them. Different data sets are used to compare a number of different machine learning strategies in depth. Extensive research shows that the properties of LBP can be used to figure out how people are feeling by looking at their faces. We suggest Hybrid-LBP as a way to get more of the LBP features that are the best at telling them apart. When Support Vector Machine classifiers and Hybrid-LBP features are used together, the best recognition performance is achieved. We also look into how LBP features can be used to recognise low-resolution facial expressions, which is an important problem that has only been looked at a few times in the relevant research. During our research, we came to the conclusion that LBP features work reliably and consistently with a wide range of low-resolution face photos. We also found that these features work well in compressed video sequences with low resolution that were shot in natural settings.

Keyword: facial expression, Hybrid-LBP , Machine classifiers