Using Taguchi Methods with Absolute Voltage Time Histogram to Construct Diagnostic Model of Tennis Elbow

• Main Authors: Ting-Jen Chang, 張庭禎
• Other Authors: Tsair-Fwu Lee
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/6beswr

碩士 === 國立高雄應用科技大學 === 電子工程系碩士班 === 102 === Purpose: Using Taguchi method to select the variables from the absolute voltage time histogram and equivalent uniform voltage to build tennis elbow predictive models. Materials and methods: In this study, seventy-eight subjects were tested tennis elbow EMG signal data, and if the visual analog scale over than grade 3 means the subject will be defined Suffering by tennis elbow, we use the absolute voltage time histogram to quantify the EMG data and calculated equivalent uniform voltage, which can able to build tennis elbow predictive models. The parameter n which can affect the model result was start from-10 to 10. Using Taguchi method to select the parameter n by observation Signal-to-noise ratio and pick out the biggest n value of Signal-to-noise ratio. We used logistic regression models, Probit prediction model and fuzzy theory to build prediction models and finally, three methods will be compared. Results: This study the selected the parameter n from -10 to 10 for chose the best n to build tennis elbow predictive models. In the first experiment, the Taguchi method is found when n value is 3, and the second experiment of Taguchi method was selected the n value is 2.3. We using when n is 2.3 to calculated equivalent uniform voltage to build prediction models. Logistic regression models, Probit prediction model and fuzzy theory all passed the test. Tennis elbow prediction models of logistic regression model has the AUC value of 0.86, Probit tennis elbow prediction model has the AUC value of 0.87, the fuzzy theory has the AUC is 0.95. Conclusions: The parameter n of equivalent uniform voltage is a very important factor of tennis elbow prediction model. At three prediction models, fuzzy theory has the best predictive effect.