基于偏序拓扑图的帕金森病语音障碍分析方法
|
摘要
从形式概念分析角度,提出将偏序拓扑图用于帕金森病语音障碍分析与诊断。首先,在属性拓扑的基础上,结合偏序结构表示,构造偏序拓扑图的形式背景表示方法,并利用偏序拓扑图进行概念本体计算,获得原始形式背景的层次化概念树结构。进而结合决策属性,对概念树进行着色与约简,获得约简概念树。根据约简概念树的偏序关系,可获得分析对象的概念分类结构。将该方法应用于帕金森病语音特征数据集进行概念提取,实验表明不但可以在概念层面分析帕金森病与语音特征的关系,同时可以作为诊断依据进行数据诊断。将该方法应用于多个帕金森病数据集(样本数分别为197、5 875、1 040、220)进行分类精度对比测试,表明基于偏序拓扑图的帕金森病语音障碍分析在不同的帕金森病语音数据集下的平均诊断精度达到76.64%,高于LDA(67.36%)、QDA(70.83%)、kNN(71.83%)、parzen窗(70.24%)、SVM(74.61%)等经典分类器的诊断精度,高出经典分类器SVM 2.72%,表明该方法能有效应用于帕金森病语音障碍分析。
In this paper, we proposed a novel dysphonic analysis method on Parkinson′s disease based on partially ordered topological graph from the view of formal concept analysis. Firstly, we constructed a representation(method) named partially ordered topological graph(POT graph) from attribute topology and attribute partially ordered graph, which gained the ability of concept searching and hierarchical concept tree structure representation. Coloring and briefing the concept tree could obtain the brief concept tree. The concept classification structure of the analysis object could be obtained according to the partial order relation of the brief concept tree. Applying the method to concept searching in Parkinson′s disease dataset, results showed that the POT graph could not only analyze the relationship between Parkinson′s disease and speech feature in the view of formal concept, but also be used as a diagnostic basis for data analysis. Results obtained from several Parkinson′s disease datasets(the numbers of sample are 197, 5875, 1040 and 220)showed that the average precision was 76.64% by POT graph. Compared with the classical classifier such as LDA(67.36%), QDA(70.83%), kNN(71.83%), Parzen window(70.24%), and SVM(74.61%), our result was higher than SVM 2.72%. In conclusion, the proposed method could be beneficial to the dysphonic analysis of Parkinson′s disease.
In this paper, we proposed a novel dysphonic analysis method on Parkinson′s disease based on partially ordered topological graph from the view of formal concept analysis. Firstly, we constructed a representation(method) named partially ordered topological graph(POT graph) from attribute topology and attribute partially ordered graph, which gained the ability of concept searching and hierarchical concept tree structure representation. Coloring and briefing the concept tree could obtain the brief concept tree. The concept classification structure of the analysis object could be obtained according to the partial order relation of the brief concept tree. Applying the method to concept searching in Parkinson′s disease dataset, results showed that the POT graph could not only analyze the relationship between Parkinson′s disease and speech feature in the view of formal concept, but also be used as a diagnostic basis for data analysis. Results obtained from several Parkinson′s disease datasets(the numbers of sample are 197, 5875, 1040 and 220)showed that the average precision was 76.64% by POT graph. Compared with the classical classifier such as LDA(67.36%), QDA(70.83%), kNN(71.83%), Parzen window(70.24%), and SVM(74.61%), our result was higher than SVM 2.72%. In conclusion, the proposed method could be beneficial to the dysphonic analysis of Parkinson′s disease.
引文
[1] Kehagia AA, Barker RA, Robbins TW. Cognitive impairment in Parkinson′s disease: The dual syndrome hypothesis[J]. Neuro-degenerative Diseases, 2013, 11(2):79-92.
[2] Barnett R. Parkinson′s disease[J]. Lancet, 2016, 387(10015):217-217.
[3] Proen?a J, Veiga A, Candeias S, et al. Characterizing Parkinson′s Disease Speech by Acoustic and Phonetic Features[M]// Computational Processing of the Portuguese Language. Berlin: Springer International Publishing, 2014:24-35.
[4] Little MA, Mcsharry PE, Roberts SJ, et al. Exploiting Nonlinear Recurrence and Fractal Scaling Properties for Voice Disorder Detection[J]. BioMedical Engineering OnLine, 2007, 6(1):23-42.
[5] Athanasios T, Little MA, Mcsharry PE, et al. Accurate telemonitoring of Parkinson′s disease progression by noninvasive speech tests.[J]. IEEE Transactions on Biomedical Engineering, 2010, 57(4):884-893.
[6] Sakar BE, Isenkul ME, Sakar CO, et al. Collection and analysis of a Parkinson speech dataset with multiple types of sound recordings.[J]. IEEE Journal of Biomedical & Health Informatics, 2013, 17(4):828-834.
[7] Orozco-Arroyave JR, H?nig F, Arias-Londoňo JD, et al. Automatic detection of Parkinson′s disease in running speech spoken in three different languages[J]. Journal of the Acoustical Society of America, 2016, 139(1):481-500.
[8] 张涛, 洪文学, 常凤香,等. 基于元音分类度的帕金森病语音特征分析[J]. 中国生物医学工程学报, 2011, 30(3):476-480.
[9] Das R. Classification of Parkinson′s disease by using voice measurements[J]. International Journal of Reasoning-based Intelligent Systems, 2010,2(3-4):279-284.
[10] Frid A, Kantor A, Svechin D, et al. Diagnosis of Parkinson′s disease from continuous speech using deep convolutional networks without manual selection of features[C]// 2016 IEEE International Conference on the Science of Electrical Engineering (ICSEE). Eilat: IEEE, 2016:1-4.
[11] Meghraoui D, Boudraa B, Merazi-Meksen T, et al. Parkinson′s Disease Recognition by Speech Acoustic Parameters Classification[M]// Modelling and Implementation of Complex Systems: Lecture Notes in Networks and Systems. Cham: Springer, 2016:165-173.
[12] 李勇明, 杨刘洋, 刘玉川,等. 基于语音样本重复剪辑和随机森林的帕金森病诊断算法研究[J]. 生物医学工程学杂志, 2016(6):1053-1059.
[13] Al-Fatlawi AH, Jabardi MH, Ling SH. Efficient diagnosis system for Parkinson′s disease using deep belief network[C]// 2016 IEEE Congress on Evolutionary Computation (CEC). Vancouver: IEEE, 2016:1324-1330.
[14] 张涛, 洪文学, 任宏雷. 基于计算几何分类器的帕金森病语音障碍可视化诊断分析[J]. 中国生物医学工程学报, 2013, 32(1):119-123.
[15] Knecht S, Wille R. Congruence Lattices of Finite Lattices as Concept Lattices[M]. Boston: Springer, 1990:323-325.
[16] Xu Weihua, Li Wentao. Granular Computing Approach to Two-Way Learning Based on Formal Concept Analysis in Fuzzy Datasets[J]. IEEE Transactions on Cybernetics, 2016, 46(2):366-379.
[17] 张涛, 任宏雷, 洪文学,等. 基于属性拓扑的可视化形式概念计算[J]. 电子学报, 2014, 42(5):925-932.
[18] 张涛, 白冬辉, 李慧. 属性拓扑的并行概念计算算法[J]. 软件学报, 2017, 28(12):3129-3145.
[19] 张涛, 魏昕宇. 属性拓扑关联规则发现[J]. 小型微型计算机系统, 2017, 38(3):548-552.
[20] Czerniak J, Zarzycki H. Application of rough sets in the presumptive diagnosis of urinary system diseases[M]// Artificial Intelligence and Security in Computing Systems: The Springer International Series in Engineering and Computer Science. Boston:Springer, 2003:41-51.
[21] 张涛.属性拓扑理论及其应用[M].北京:科学出版社, 2017:134-156.
[22] 徐伟华, 李金海, 魏玲,等.形式概念分析理论与应用[M].北京:科学出版社,2016:98-121.
[23] 张涛, 宋佳霖, 刘旭龙,等. 基于色度学空间的多元图表示[J]. 燕山大学学报, 2010, 34(2):111-114.
[24] 张涛, 洪文学. 基于计算几何的非线性可视化分类器设计[J]. 电子学报, 2011, 39(1):53-58.
[25] 张涛. 基于语音特征的帕金森病可视化诊断方法研究[D]. 秦皇岛:燕山大学, 2012.
[26] 张涛, 师浩斌, 李林,等. 决策连续形式背景的可视化数据离散化方法[J]. 计算机应用研究, 2016, 33(2):388-391.
[2] Barnett R. Parkinson′s disease[J]. Lancet, 2016, 387(10015):217-217.
[3] Proen?a J, Veiga A, Candeias S, et al. Characterizing Parkinson′s Disease Speech by Acoustic and Phonetic Features[M]// Computational Processing of the Portuguese Language. Berlin: Springer International Publishing, 2014:24-35.
[4] Little MA, Mcsharry PE, Roberts SJ, et al. Exploiting Nonlinear Recurrence and Fractal Scaling Properties for Voice Disorder Detection[J]. BioMedical Engineering OnLine, 2007, 6(1):23-42.
[5] Athanasios T, Little MA, Mcsharry PE, et al. Accurate telemonitoring of Parkinson′s disease progression by noninvasive speech tests.[J]. IEEE Transactions on Biomedical Engineering, 2010, 57(4):884-893.
[6] Sakar BE, Isenkul ME, Sakar CO, et al. Collection and analysis of a Parkinson speech dataset with multiple types of sound recordings.[J]. IEEE Journal of Biomedical & Health Informatics, 2013, 17(4):828-834.
[7] Orozco-Arroyave JR, H?nig F, Arias-Londoňo JD, et al. Automatic detection of Parkinson′s disease in running speech spoken in three different languages[J]. Journal of the Acoustical Society of America, 2016, 139(1):481-500.
[8] 张涛, 洪文学, 常凤香,等. 基于元音分类度的帕金森病语音特征分析[J]. 中国生物医学工程学报, 2011, 30(3):476-480.
[9] Das R. Classification of Parkinson′s disease by using voice measurements[J]. International Journal of Reasoning-based Intelligent Systems, 2010,2(3-4):279-284.
[10] Frid A, Kantor A, Svechin D, et al. Diagnosis of Parkinson′s disease from continuous speech using deep convolutional networks without manual selection of features[C]// 2016 IEEE International Conference on the Science of Electrical Engineering (ICSEE). Eilat: IEEE, 2016:1-4.
[11] Meghraoui D, Boudraa B, Merazi-Meksen T, et al. Parkinson′s Disease Recognition by Speech Acoustic Parameters Classification[M]// Modelling and Implementation of Complex Systems: Lecture Notes in Networks and Systems. Cham: Springer, 2016:165-173.
[12] 李勇明, 杨刘洋, 刘玉川,等. 基于语音样本重复剪辑和随机森林的帕金森病诊断算法研究[J]. 生物医学工程学杂志, 2016(6):1053-1059.
[13] Al-Fatlawi AH, Jabardi MH, Ling SH. Efficient diagnosis system for Parkinson′s disease using deep belief network[C]// 2016 IEEE Congress on Evolutionary Computation (CEC). Vancouver: IEEE, 2016:1324-1330.
[14] 张涛, 洪文学, 任宏雷. 基于计算几何分类器的帕金森病语音障碍可视化诊断分析[J]. 中国生物医学工程学报, 2013, 32(1):119-123.
[15] Knecht S, Wille R. Congruence Lattices of Finite Lattices as Concept Lattices[M]. Boston: Springer, 1990:323-325.
[16] Xu Weihua, Li Wentao. Granular Computing Approach to Two-Way Learning Based on Formal Concept Analysis in Fuzzy Datasets[J]. IEEE Transactions on Cybernetics, 2016, 46(2):366-379.
[17] 张涛, 任宏雷, 洪文学,等. 基于属性拓扑的可视化形式概念计算[J]. 电子学报, 2014, 42(5):925-932.
[18] 张涛, 白冬辉, 李慧. 属性拓扑的并行概念计算算法[J]. 软件学报, 2017, 28(12):3129-3145.
[19] 张涛, 魏昕宇. 属性拓扑关联规则发现[J]. 小型微型计算机系统, 2017, 38(3):548-552.
[20] Czerniak J, Zarzycki H. Application of rough sets in the presumptive diagnosis of urinary system diseases[M]// Artificial Intelligence and Security in Computing Systems: The Springer International Series in Engineering and Computer Science. Boston:Springer, 2003:41-51.
[21] 张涛.属性拓扑理论及其应用[M].北京:科学出版社, 2017:134-156.
[22] 徐伟华, 李金海, 魏玲,等.形式概念分析理论与应用[M].北京:科学出版社,2016:98-121.
[23] 张涛, 宋佳霖, 刘旭龙,等. 基于色度学空间的多元图表示[J]. 燕山大学学报, 2010, 34(2):111-114.
[24] 张涛, 洪文学. 基于计算几何的非线性可视化分类器设计[J]. 电子学报, 2011, 39(1):53-58.
[25] 张涛. 基于语音特征的帕金森病可视化诊断方法研究[D]. 秦皇岛:燕山大学, 2012.
[26] 张涛, 师浩斌, 李林,等. 决策连续形式背景的可视化数据离散化方法[J]. 计算机应用研究, 2016, 33(2):388-391.