The method that the present invention proposes a kind of tumble detecting and classification,By being worn on the acceleration value of the acceleration sensor with human body,Through calculating learn advanced jerk value (Modified Jerk),Jerk value (Jerk),The information such as velocity amplitude (Speed),By the Hou of the screening of felicity condition,Reuse the ascribed value that the fuzzy deduction system (Fuzzy System) with rule base calculates tumble posture； Most Hou,Ascribed value can respectively via Bayesian analysis classification (Bayesian Classifier),Vector machine classification (Support Vector Machine) and vector machine classification (SVM with Gaussian Distribution) etc. the mode classifications for having Gauss branch judge tumble before human body,Hou tumble,Left tumble,The postures such as right tumble.本發明提出一種跌倒偵測與分類之方法，由穿戴在人體身上的加速度感測器之加速度值，透過計算得知進階躍度值(Modified Jerk)、躍度值(Jerk)、速度值(Speed)等資訊，經過適當條件的篩選之後，再使用具有規則庫的模糊推論系統(Fuzzy System)計算跌倒姿態的歸屬值；最後，歸屬值可分別經由貝氏分析分類(Bayesian Classifier)、向量機分類(Support Vector Machine)以及具高斯分部的向量機分類(SVM with Gaussian Distribution)等分類方式判斷人體前跌倒、後跌倒、左跌倒、右跌倒等姿態。
