基于人脸多特征融合的精神分裂症自动识别

Automatic recognition of schizophrenia based on fusion of facial multi-feature

针对精神分裂症诊断主观性强、缺乏客观辅助诊断指标的问题,提出基于人脸多特征融合的精神分裂症自动识别算法。提取人脸关键点,综合分析文本朗读过程中精神分裂症患者眼动、嘴部运动、头部姿态等非言语行为特征,构建融合特征并结合支持向量机进行自动分类识别。实验采集20位精神分裂症患者与20位正常对照组的文本朗读视频进行测试,其结果表明,提出方法在十折交叉验证中获得92.5%的分类正确率,算法实现了对精神分裂症的自动识别,能够为精神分裂症提供客观的辅助诊断指标。

Aiming at the lack of objective diagnostic indicators in the clinical diagnosis of schizophrenia,an automatic schizophrenia recognition algorithm was proposed.Based on the extracted key points of the face,the nonverbal behaviors in the eye movement,mouth movement,and head posture of schizophrenic patients were analyzed.The above abnormal behavior characteristics were fused and combined with the support vector machine to realize the automatic detection of schizophrenia.Experimental data of this work were text reading videos recorded by 20 schizophrenic patients and 20 normal controls.Experimental results indicate that the classification accuracy of the proposed automatic schizophrenia recognition algorithm with 10-fold cross-validation is 92.5%.The proposed algorithm realizes the automatic detection of schizophrenia and it can be used as an objective auxiliary diagnostic method for schizophrenia.