Development of diagnostic method of velopharyngeal closing function by speech using Chaos analysis

为了建立一种腭裂语音诊断方法，我们研究了唇动分析方法和语音混沌分析方法。嘴唇运动分析方法3台红外摄像机、1台彩色摄像机、1台红外投影仪、4台数码摄像机和1台同步器拍摄人脸的运动。我们成功地应用多基线立体算法生成了人脸的深度图像。利用高精度的致动器和已知目标对距离像的精度进行了研究。因此，根据距离图像计算的距离与已知值之间的差异在纵向为0.53~0.73 mm，横向为0.14~0.44 mm。认为在临床应用中，距离像的准确性是足够的。随后，我们开发了以下程序来分析嘴唇运动：1)将嘴唇划分为几个区域，2)采用UV坐标系，3)计算速度和特征值。现在对唇动在发音中的作用进行分析。言语的混沌分析--使用“ki-tsu-tsu-ki-tsu-tsu-ku”和“u-e-o-u”两个句子，计算Lyapunov指数。同时记录鼻音。受试者为16例健康成人和9例腭裂患者，包括2例明显的腭咽闭合不全(VPI)患者。结果发现：1)健康组鼻音评分非常稀疏，2)鼻音评分与Lyapunov指数无相关性，3)VPI患者鼻音评分和Lyapunov指数均较高。提示提高腭咽闭合功能诊断的准确性是可能的。

To establish a diagnostic method of cleft palate speech, we study on analyzing method of lip movement and chaos analyzing of speech.1. Analyzing method of lip movementA motion of the face was captured using 3 IR cameras, one color camera, a projector of infrared pattern, 4 digital video decks and a synchronizer. We succeeded in production of range images for a face by applying the algorithm of ~a multiple-baseline stereo. Accuracy of the range image was investigated using an actuator with high accuracy and a known object. Consequently, the differences between the distances calculated from the range images and the known values were ranging from 0.53-0.73mm and 0.14-0.44mm in longitudinal and lateral directions, respectively. It was considered that the accuracy of the range images was enough in clinical usage. Subsequently, we developed the following programs to analyze a lip motion : 1) the lip was divided into several areas, 2) uv coordinate system was applied, and 3) the speed and eigenvalue were calculated. Now lip motion in articulation is analyzed.2. Chaos analysis of speechUsing sentences of "ki-tsu-tsu-ki-tsu-tsu-ku" and "u-e-o-o-o-u", Lyapunov exponents were calculated. Simultaneously, nasalance was recorded. The subjects were 16 healthy adults and 9 cleft palate patients, including the two patients who displayed obvious velopharyngeal incompetence (VPI). As a result, the following findings were indicated : 1) nasalance scores in the healthy group were very sparse, 2) there was no correlation between nasalance scores and Lyapunov exponents, 3) both nasalance scores and Lyapunov exponents were high in the patients with VPI. This indicated the possibility to raise the precision of diagnosis for velopharyngeal closing function.