動作判斷方法及系統

專利類型

發明

專利國別 (專利申請國家)

中華民國

專利申請案號

109120196

專利證號

I 758749

專利獲證名稱

動作判斷方法及系統

專利所屬機關 (申請機關)

國立臺北科技大學

獲證日期

2022/03/21

技術說明

本專利針對棒壘球運動項目中，應用於上肢投擲姿態的學習、受傷後的姿態復健及投擲姿態的矯正，研發上肢投擲姿態辨識裝置，並能透過姿態顯示裝置及手機來顯示訊息。目前關於投擲姿態辨識的研究文獻中，大多數皆使用高速動態攝影機，其配合研究所需要的場地、設備裝置要苛刻的多，並且都以離線方式進行分析。有鑒於此，本專利以嵌入式系統設計一套穿戴式裝置來實現我們創新的想法，並利用連續投擲姿態辨識技術，即時辨識每個階段之上肢投擲姿態，再透過即時顯示訊息的方式，讓使用者能重複進行正確上肢投擲姿態學習、復健及矯正的用途。完整的投擲姿態階段分成準備期、跨步期、揮臂準備期、手臂加速期、手臂減速期及投擲後期六個階段。投擲姿態辨識相較於一般簡單動作的姿態辨識複雜很多，因投擲姿態不僅僅包含每個階段的姿態特徵，還包含了連續不斷的姿態動作變換。感測訊號經濾波去雜訊處理並轉成一維資料後，投擲姿態每個階段之連續動作得以序列表示，之後採用LCS演算法進行分類運算。 本專利設計了一套具有智慧姿態辨識系統，能分析連續姿態細節、姿態平衡判斷、自動啟動辨識機制、姿態優先權比對、投擲姿態辨識速度處理以及中途中斷處理功能的系統，此系統還能分析較不適當的投擲姿態，其系統準確率能到八成五以上，達到完整的上肢投擲姿態辨識。 4.本申請案之英文摘要(Description of Invention in English) In this patent, we focus on real-time upper limb throwing gesture recognition which can be utilized for improving the throwing action in baseball and softball. This gesture recognition is helpful for baseball player or softball player to study his action and make improvements or for correcting the throwing action after a player returns from an injury and so on. The implemented device recognizes the upper limb throwing gesture and the information is displayed on the gesture display device and mobile phone. Most of the literatures on throwing gesture recognition used high speed dynamic camera. But in all these recognition methods, the environment setup, device implementation and algorithm are more complicated. Also, all those methods uses offline process for analysis. Our study focus on the design of an embedded wearable device to realize our innovative ideas using the continuous throwing gesture recognition technology to instantly identify the upper limb throwing gesture at each phase. Furthermore, the device can send instant message to the user, who can repeat upper limb throwing gesture and this helps in the training session to improve the action and make proper correction wherever required. The phase of the complete throwing gesture is divided into six: Wind-Up, Stride, Arm Cocking, Arm Acceleration, Arm Deceleration and Follow-Through. General simple gesture recognition include only every phase of the gesture characteristic and there is no continuous phase change recognition. But in throwing gesture, a continuous phase change is occurring. So the throwing gesture recognition is more complex than general simple gesture recognition. After the signal is completely processed, filtered and converted into one-dimensional information, the action of each phase of the throwing gesture is represented by a sequence. After that, we use Longest Common Subsequence (LCS) to determine the throwing gesture. In addition, we added a set of intelligent gesture recognition functionalities to analyze the continuous gesture details, determine the balance of gestures, start identification of the mechanism automatically, compare gesture priority, process the speed of throwing gesture recognition and handle interrupt situation. The system can also analyze improper throwing gesture. The system precision is more than 85% and therefore it can achieve a complete upper limb throwing gesture recognition successfully.

備註

連絡單位 (專責單位/部門名稱)

專利技轉組

連絡電話

02-87720360

網址