Neonatal Pain Recognition Based On Local Binary Patterns And Sparse Representation

Neonates cannot articulate their pain experiences, the assessment of pain is difficult. At present,most methods depend on the observations of health professionals that cannot respond to the pain ofneonates effectively and immediately, how to develop an automatic neonatal pain recognitionsystem based on the facial expression is very significant.A new algorithm to neonates’ pain expression analysis based on LBP and sparse representation-based classification method is proposed. The research work of this paper is as follows:(1) LBP algorithm is used to extract the feature of neonatal pain expression and the effects ofimage block size on recognition performance are also taken into account. The expression images aredivided into3×3blocks,4×4blocks,5×3blocks,6×5blocks, uniform LBP is adopted under fourdifferent block modes.(2) PCA is introduced to reduce the dimensions of feature.The dimension makes a difference tothe recognition performance.(3) SRC based on compressed sense theory is adopted as a classifier. Two different algorithms--gradient projection method (GPSR) and truncated Newton interior point method (TNIPM)--areapplied to solve the sparse solution. Experiments on the neonatal pain expression recognition showthat the proposed algorithm has a satisfied performance, the average correct rate can be achieved to87%.