Neonatal Pain Facial Expression Recognition Based On Sparse Representation And Matching Pursuit Algorithm

Recently，pain assessment in neonates has prevalently become a new research orientation inhealth science. Due to neonates are not capable of verbally describing the location, duration andintensity of their pain experiences，untreated pain in infants may result in central nervous systemchanges that slow development. Therefore，the method of properly assess the pain and how to takegood care of infants has been the focus of medical researchers. For now，facil expressions，inparticular，are considerd the gold standard in pain assessment because they are the most specific andfrequent indicators of pain.In this paper，we present a new method to assess the neonatal pain which based on the SAMPalgorithm to solve the sparse solution.By deeply investigating the facial expression recognition viacompressive sensing methods, this thesis proposes a new method of neonatal pain facial expressionrecognition by sparse representation and matching pursuit algorithm. The main contributions of thisthesis are summarized as follows:（1） This thesis provides a generalized framework of neonatal pain facial expression recognitionwhich based on the matching pursuit algorithms. We propose a new sparse representation basedon SAMP algorithm, and apply it to our facial expression recognition research.（2） This thesis introduce3popular extract features methods to reuduce the dimensionality,whichare down-sampled，traditional PCA and2DPCA.We put them into our neonatal pain facialexpression recognition framework,as the function of preprocess.The experiment results show thatcompared other2extract features methods, the proposed algorithm framework which combinedSAMP and2DPCA can achieve good recognition rate on neonatal pain facial expressionrecognition..(3) In the experiment, the SAMP has been compared with other matching pursuit algorithm asthe way to solve the problem of sparse representation.The results suggest that sparse representationbased on SAMP algorithm be superior to the other matching pursuit algorithm both on identifyefficiency and identify time, and with no need for value of sparsity. Through integrated into account,A novel method for neonatal pain facial expression recognition based on SAMP is better thanothers.