基于SVM的新生儿疼痛表情识别

近十年来新生儿疼痛引起医护人员的广泛关注。由于新生儿不能自述疼痛的感受，疼痛评估成为新生儿科学中最具挑战性的一个难题。新生儿“疼痛面容”（蹙眉、挤眼、鼻唇沟加深、张口）被认为是最可靠的疼痛指标，且持续时间最长，因而被国际上常用的新生儿疼痛评估工具作为评估指标。然而，这些疼痛评估工具往往受到临床医护人员主观因素的影响。文中旨在解决上述问题，提出利用支持向量机（SVM）技术对新生儿疼痛与非疼痛面部表情进行分类识别。对210幅照片的表情图像进行了研究，比较了线性核函数SVM、多项式核函数SVM（d=2，3，4）以及径向基函数SVM等5种不同分类器的性能。实验结果表明，阶数d=3的多项式核函数SVM分类器的性能最佳，对疼痛和非疼痛表情分类的识别率达到93．33％，对疼痛与安静表情的分类识别率为94．17％，对疼痛与哭表情的分类识别率为83．13％，初步具备了在新生儿疼痛评估中的潜在应用价值。

Recognition for Expression of Pain in Neonate Using Support Vector Machine

Pain in neonates has prevalently attracted the attention of health professionals during the latest decade. Pain assessment is considered as one of the most challenging tasks in neonatology because neonates cannot verbalize their pain experiences. Facial expressions including brow bulge, eye squeeze, nasal-labial furrowing and mouth open are considered critical factors in pain assessment because they are the most specific and frequent indicators of pain. As such, most pain assessment tools developed for neonates and infants rely in whole or in part on facial ex- pressions. However,these pain assessment tools are highly observer-dependent, and studies demonstrate that clini- cians arc not entirely impartial in their judgments. This study investigates these problems by applying the Support Vector Machine （SVM） technique in distinguishing expression of pain from facial expressions that are similar but not triggered by pain. In our experiments,210 facial images are investigated by using SVM with linear kernel,SVM with a polynomial kernel （degree = 2,3,4）, SVM with Radial Basis Function （RBF）. The best recognition rates of pain versus n0n-pain （93.33%） ,pain versus rest （94.17%）,pain versus cry （83.13%） are obtained from an SVM with a polynomial kernel of degree 3. The results of this study indicate that the application of SVM technique in pain assessment is a promising area of investigation.