A research team at MU is developing an algorithm to recognize facial expressions at low-resolution images. Facial expressions are important ways to express human emotions and have many important applications such as human-computer interaction, social robots, medical treatment, and driver assistance applications.
Despite the remarkable success of deep convolutional neural networks, recognizing facial expressions at low-resolution images is still a difficult task. with that in mind, a group of researchers from MU is designing a new deep convolutional neural network to address this problem. The network consists of four main units which are the feature extraction layer, the attention layer for spatial channels, the spatial hierarchical aggregation layer, and the classification layer. All proposed network layers consist of convolutional layers that do not depend on full communication layers which helps to reduce the size of the network transactions that can be learned.
The results showed that the performance of the proposed method is superior to the latest other methods and can be run on low-resource devices such as mobile phone.
This research is conducted by the Engineering and Applied Research Center and funded by the Deanship of Scientific Research.