本研究從實作評量之觀點，探討使用視覺監控分析人物目標之姿勢動作並結合樣式辨識，應用於診斷受試者感覺統合障礙之有效性。本研究之研究對象是239位4至6歲幼稚園兒童，主要是從不同角度蒐集受試者感覺統合臨床觀察之姿勢動作，搭配不同的分類器，診斷受試者感覺統合臨床觀察之姿勢動作是否有障礙。研究結論彙整如下：1.以視覺監控為基礎之感覺統合實作評量，搭配支撐向量機和貝氏網路能得到較好的診斷效果。2.以視覺監控為基礎之感覺統合實作評量，Cronbach's係數是.80，顯示本研究發展的評量工具具有良好之信度。3.以視覺監控為基礎之感覺統合實作評量，在診斷「趴姿伸直」姿勢動作之診斷正確率達86.2%，診斷「臥姿彎曲」姿勢動作之診斷正確率達85.0%，顯示具有良好之效度。

The purpose of this study is to develop a visual surveillance based posture movements assessment for sensory integration clinical observation. Based on the view of performance assessments, the pattern recognition is combined with the visual surveillance procedures to analyze the posture movements (prone extension and supine flexion) and to diagnose the sensory integration dysfunction of subjects in this study.A total of 239 participants with age from four to six years old were selected from kindergartens. The major findings of this study are summarized as follow: (1) for employing the support vector machine classifier and Baysian network classifier are best in diagnosing disorder of sensory integration functions in terms of important features and overall performance of prone extension and supine flexion, respectively. (2) With good prediction rates, prone extension: 86.2% and supine flexion: 85%, the visual surveillance based performance assessment for sensory integration has good validity. (3) With good internal index, Cronbach's α: .80, the visual surveillance based posture movements assessment for sensory integration has good reliability.