| Pain in neonates has prevalently attracted the attention of health professionals during the latest decade. Pain assessment is considered as one of the new research topics in neonatology because neonates cannot verbalize their pain experiences, which has a wide range of applications and potential market use.Compressive sensing (CS) is a novel signal sampling theory under the condition that the signal is sparse or compressible. It has the ability of compressing a signal during the process of sampling. And theory of sparse representation is one of the most core topics in compressive sensing. With the decrease of its computational complexity, it is widely adopted in nearly every area of signal processing. As a special type of signal, image signal itself or its semantic meaning has sparsity, which makes it possible to introduce the sparse representation technique to the area of image processing and pattern recognition.A novel method for Infant pain facial expression recognition based on sparse representation which is one of the most core theories in compressive sensing was presented in this paper.The sparsest representation of the test infant facial expression image computed by L-1 minimization had distinct class information. Therefore, it is easy to classify the infant facial expression, and then decide whether the infant is pain in this moment.We conduct extensive experiments to verify this new approach using the infant facial expression image database constructed by us. In the research, we find that if sparsity in pattern recognition problem is properly used, the proposed algorithm framework can achieve good recognition rate on infant pain facial expression recognition. Furthermore, the experiment results show that if the number of attracted features and training samples is sufficient (reach a threshold) and the sparse representation is correctly found, using unconventional features such as down-sampled images will perform well in the recognition problem. |
PDF Full Text Request