智能牙刷可以藉由監測刷牙方式來維持牙齒的健康，正確的刷牙位置的監測非常重要。本論文提出了一個改善刷牙位置辨識的方法。首先運用中值濾波降低感測器雜訊，並且抑制不規則的高頻刷牙運動，減少偵測的角度突波，以便得到更精確的運動特徵，接著我們使用機率神經網路(Probabilistic Neural Network, PNN)分類方法，來進行刷牙區域辨識。依據實驗結果，相對於現有k-means的刷牙區域分類方法，PNN分類方法能獲得更佳的區域分類性能，讓使用者能夠更準確監測正確刷牙狀態。此一成果可以結合智能手機作為閘道器連結雲端，藉收集即時刷牙資料，發展遠端牙齒健康照護和疾病預防的應用。;Smart toothbrushes can be designed to monitor tooth brushing procedures to facilitate the maintenance of dental health. This necessitates a monitoring mechanism that identifies proper tooth brushing regions. In this study, an improved technique for identifying tooth brushing regions was proposed. First, a median filter was used to attenuate sensor noise and control irregular high-frequency tooth brushing motions, thereby reducing angular surges during detection and obtaining as precise of tooth brushing motion characteristics as possible. Then, a probabilistic neural network (PNN) was applied to identify tooth brushing regions. The experimental results showed that, compared with the conventional k-means algorithm, the PNN was more effective at classifying tooth brushing regions to enable users to monitor their tooth brushing more precisely. The proposed technique can be integrated through smartphone gateways to cloud services to collect real-time tooth brushing data for application in the development of remote dental healthcare and dental disease prevention services.
