TW202228596A - Hand-held motion analysis sysyem and method - Google Patents
Hand-held motion analysis sysyem and method Download PDFInfo
- Publication number
- TW202228596A TW202228596A TW110102787A TW110102787A TW202228596A TW 202228596 A TW202228596 A TW 202228596A TW 110102787 A TW110102787 A TW 110102787A TW 110102787 A TW110102787 A TW 110102787A TW 202228596 A TW202228596 A TW 202228596A
- Authority
- TW
- Taiwan
- Prior art keywords
- ball
- signal
- hitting
- action
- batting
- Prior art date
Links
Images
Landscapes
- User Interface Of Digital Computer (AREA)
Abstract
Description
本發明關於一種分析系統與方法,特別關於一種手持運動分析系統及分析方法。The present invention relates to an analysis system and method, in particular to a handheld motion analysis system and analysis method.
球類運動,例如羽毛球、桌球、或網球等,由於入門門檻低,場地取得也相對容易,相當受到一般人的喜愛。但是,要成為代表國家出賽的國手或成為職業運動員則相當不容易,除了需付出相當的努力之外,可能也需要擊球動作分析系統的輔助。Ball sports, such as badminton, billiards, or tennis, are very popular among ordinary people due to the low entry threshold and relatively easy access to venues. However, it is not easy to become a national player representing the country or become a professional athlete. In addition to considerable effort, it may also require the assistance of a batting action analysis system.
傳統用於研究擊球運動的動作分析系統中,研究者為了擷取擊球動作的過程,必須架設高速攝影機,並將攝影機取得的影像傳送到電腦,另外再接一傳輸線連接一保持在預期觸發狀態的運動感應器(sensor),以記錄擊球過程之資訊並將其傳送至電腦設備中。In the traditional action analysis system used to study the batting movement, in order to capture the process of the batting action, the researcher must set up a high-speed camera, transmit the image obtained by the camera to the computer, and then connect a transmission line to keep it at the expected trigger. The state of the motion sensor (sensor), to record the information of the hitting process and transmit it to the computer equipment.
然而,習知的動作分析系統必須考慮攝影機的架設角度以利拍攝,在實驗測試前必須將標誌點固定於受試者的關節上以利在進行身體角度分析時能有明確的點選目標;在拍攝影像後必須拍攝一段比例尺(例如長一米)畫面以利資料的轉換;且使用紅外線發射器及接收器的設計而形成一非常複雜笨重的分析系統,不但使用上麻煩不便、只能用於特定場所使用,而且價格昂貴,在實驗拍攝的過程、資料的處理和分析等程序中會因人為因素而產生極大的誤差。However, the conventional motion analysis system must consider the camera's erection angle to facilitate shooting, and the marker points must be fixed on the subject's joints before the experimental test, so that the body angle analysis can have a clear click target; After shooting the image, it is necessary to take a scale (for example, one meter long) picture to facilitate the conversion of data; and the design of the infrared transmitter and receiver is used to form a very complex and bulky analysis system, which is not only inconvenient to use, but can only be used It is used in a specific place and is expensive. In the process of experimental shooting, data processing and analysis, there will be great errors due to human factors.
因此,如何提供一種手持運動分析系統及分析方法,除了具有方便、專業且實用經濟的優點外,還可即時且客觀地提供相關的擊球指標,已是相當重要的課題之一。Therefore, how to provide a handheld sports analysis system and analysis method, which not only has the advantages of convenience, professionalism, practicality and economy, but also provides relevant hitting indicators in real time and objectively, has become one of the very important topics.
有鑑於上述課題,本發明的目的為提供一種手持運動分析系統與方法,相較於習知的擊球動作分析系統來說,本發明除了具有方便、專業且實用經濟的優點外,還可即時且客觀地提供相關的擊球指標給運動員或/及教練參考,進而改進運動員的擊球動作。In view of the above problems, the purpose of the present invention is to provide a handheld motion analysis system and method. Compared with the conventional batting action analysis system, the present invention has the advantages of convenience, professionalism, practicality and economy, and can also be used in real time. And objectively provide relevant batting indicators to the players or/and coaches for reference, so as to improve the batting action of the players.
為達上述目的,依據本發明之一種手持運動分析系統,包括一訊號感測模組、一伺服器以及一顯示模組。訊號感測模組設置於手持球具,訊號感測模組感測手持球具的擊球動作並輸出一感測訊號;伺服器與訊號感測模組耦接,伺服器包括一或多個處理單元及一記憶單元,該一或多個處理單元與記憶單元耦接,記憶單元儲存一或多個程式指令,當該一或多個程式指令被該一或多個處理單元執行時,該一或多個處理單元進行:一姿態估測步驟,係依據感測訊號執行擊球動作的手持球具姿態估測;一擊球軌跡重建步驟,係依據感測訊號和姿態估測步驟的結果執行擊球軌跡訊號的重建;一擊球分期偵測步驟,係依據感測訊號和姿態估測步驟的結果區分擊球過程的不同時期;一擊球球種辨識步驟,係依據感測訊號將擊球的球種類型進行分類;及一擊球動作一致性評估步驟,係依據感測訊號計算和評估擊球動作與樣板動作之間的一致性。顯示模組與伺服器耦接,顯示模組呈現分析結果。To achieve the above objective, a handheld motion analysis system according to the present invention includes a signal sensing module, a server and a display module. The signal sensing module is arranged on the hand-held ball, and the signal-sensing module senses the hitting action of the hand-held ball and outputs a sensing signal; the server is coupled to the signal-sensing module, and the server includes one or more A processing unit and a memory unit, the one or more processing units are coupled to the memory unit, the memory unit stores one or more program instructions, when the one or more program instructions are executed by the one or more processing units, the One or more processing units perform: an attitude estimation step, which is based on the sensing signal to perform the attitude estimation of the hand-held golf equipment for hitting the ball; a hitting trajectory reconstruction step, which is based on the sensing signal and the result of the attitude estimation step The reconstruction of the batting trajectory signal is performed; a batting stage detection step is based on the sensing signal and the result of the attitude estimation step to distinguish different periods of the batting process; a batting ball identification step is based on the sensing signal. classifying the type of the ball hit; and a step of evaluating the consistency of the hitting action, which calculates and evaluates the consistency between the hitting action and the sample action based on the sensing signal. The display module is coupled to the server, and the display module presents the analysis result.
為達上述目的,依據本發明之一種手持運動的分析方法,應用於一手持運動分析系統,手持運動分析系統包括一訊號感測模組,訊號感測模組設置於一手持球具,並感測手持球具的擊球動作且輸出一感測訊號,該分析方法包括:一姿態估測步驟:依據感測訊號執行擊球動作的手持球具姿態估測;一擊球軌跡重建步驟:依據感測訊號和姿態估測步驟的結果執行擊球軌跡訊號的重建;一擊球分期偵測步驟:依據感測訊號和姿態估測步驟的結果區分擊球過程的不同時期;一擊球球種辨識步驟:依據感測訊號將擊球的球種類型進行分類;以及一擊球動作一致性評估步驟:依據感測訊號計算和評估擊球動作與樣板動作之間的一致性。In order to achieve the above-mentioned purpose, a method for analyzing hand-held motion according to the present invention is applied to a hand-held motion analysis system. The hand-held motion analysis system includes a signal sensing module. Measuring the hitting action of the hand-held ball equipment and outputting a sensing signal, the analysis method includes: an attitude estimation step: performing the posture estimation of the hand-held ball equipment for the hitting action according to the sensing signal; a hitting trajectory reconstruction step: according to the The sensing signal and the result of the attitude estimation step are used to reconstruct the batting trajectory signal; the staged detection step of a stroke: distinguish different periods of the batting process according to the sensing signal and the result of the attitude estimation step; The identification step: classifying the type of the ball hit according to the sensing signal; and the step of evaluating the consistency of a hitting action: calculating and evaluating the consistency between the hitting action and the sample action according to the sensing signal.
在一實施例中,訊號感測模組包括三軸加速度計、三軸陀螺儀及三軸磁力計。In one embodiment, the signal sensing module includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer.
在一實施例中,在進行姿態估測步驟之前,該一或多個處理單元更進行:一訊號前處理步驟,係校正訊號感測模組輸出的感測訊號,並濾除感測訊號中的雜訊。In one embodiment, before performing the attitude estimation step, the one or more processing units further perform: a signal preprocessing step, which is to correct the sensing signal output by the signal sensing module, and filter out the sensing signal. 's noise.
在一實施例中,姿態估測步驟透過一擴展式卡爾曼濾波器演算法利用感測訊號進行擊球動作的手持球具姿態估測,進而獲得準確的手持球具速度及手持球具軌跡;其中,擴展式卡爾曼濾波器演算法包括一狀態預測步驟、一重力狀態更新步驟、及一磁北狀態更新步驟。In one embodiment, the attitude estimation step utilizes an extended Kalman filter algorithm to estimate the attitude of the hand-held golf equipment by utilizing the sensing signal, thereby obtaining accurate hand-held golf equipment speed and hand-held golf equipment trajectory; The extended Kalman filter algorithm includes a state prediction step, a gravity state update step, and a magnetic north state update step.
在一實施例中,在狀態預測步驟中,係利用感測訊號中的角速度訊號進行擴展式卡爾曼濾波器的狀態預測;其中,在重力狀態更新步驟和磁北狀態更新步驟中,係利用感測訊號中的加速度訊號與磁力訊號進行擴展式卡爾曼濾波器的狀態更新,以得到最佳手持球具姿態估測狀態。In one embodiment, in the state prediction step, the angular velocity signal in the sensing signal is used to predict the state of the extended Kalman filter; wherein, in the gravity state update step and the magnetic north state update step, the sensor is used to predict the state. The acceleration signal and the magnetic force signal in the signal are used to update the state of the extended Kalman filter, so as to obtain the best estimation state of the posture of the hand-held ball.
在一實施例中,擊球軌跡重建步驟係透過一軌跡重建演算法取得執行擊球動作時所產生的手持球具姿態、手持球具速度及手持球具軌跡訊號;其中,軌跡重建演算法包括一動作訊號分割步驟、一座標轉換與重力補償步驟、一速度估測與零速度補償步驟、及一軌跡重建步驟。In one embodiment, in the step of reconstructing the ball trajectory, a trajectory reconstruction algorithm is used to obtain the posture of the golf equipment, the speed of the golf equipment and the trajectory signals of the golf equipment generated when the hitting action is performed; wherein, the trajectory reconstruction algorithm includes: A motion signal segmentation step, a coordinate conversion and gravity compensation step, a speed estimation and zero speed compensation step, and a trajectory reconstruction step.
在一實施例中,擊球分期偵測步驟係透過一擊球動作分期演算法取得擊球過程中的每一時期訊號;其中,擊球動作分期演算法包括一動作訊號分割步驟、一座標轉換與重力補償步驟、一動作訊號極點偵測步驟、及一動作訊號分期偵測步驟;其中,動作訊號極點偵測步驟找出擊球動作的預備期起始點、加速期起始點、擊球點及餘勢期結束點。In one embodiment, the step of batting stage detection is to obtain each period signal in the batting process through a batting action staging algorithm; wherein, the batting action staging algorithm includes a motion signal segmentation step, coordinate conversion and Gravity compensation step, a movement signal pole detection step, and a movement signal phase detection step; wherein, the movement signal pole detection step finds out the starting point of the preparatory period, the starting point of the acceleration period and the hitting point of the hitting action and the end of the residual period.
在一實施例中,在擊球分期偵測步驟中,擊球過程的不同時期包括一初始靜止期、一預備期、一加速期、一餘勢期、及一結束靜止期。In one embodiment, in the step of detecting ball hitting stages, different periods of the hitting process include an initial rest period, a preparatory period, an acceleration period, a residual period, and an end rest period.
在一實施例中,擊球球種辨識步驟係透過一擊球球種辨識演算法取得擊球球種類型;其中,擊球球種辨識演算法包括一動作訊號分割步驟、一訊號正規化步驟、一卷積神經網路分類步驟、及一球種辨識步驟。In one embodiment, the ball type identification step is to obtain the type of ball type through a ball type identification algorithm; wherein, the ball type identification algorithm includes a motion signal segmentation step and a signal normalization step , a convolutional neural network classification step, and a ball species identification step.
在一實施例中,經由擊球球種辨識步驟分類出的球種類型,包括正手發後場球、反手發前場球、正手挑後場高遠球、反手挑後場高遠球、正手推挑後場球、反手推挑後場球、前場正手短球、前場反手短球、中場正手平抽球、中場反手平抽球、中場正手接殺擋網前球、中場反手接殺擋網前球、後場正手切球、後場正手高遠球、後場正手殺球、及中場正手突襲球。In one embodiment, the types of balls classified through the batting type identification step include forehand hits, backhand hits, forehand hits, backhand hits, and forehand push hits. The ball, the backhand push and the backcourt ball, the forehand short ball in the frontcourt, the backhand short ball in the frontcourt, the midfield forehand draw, the midfield backhand draw, the midfield forehand catch and block the front ball, the midfield backhand catch and block Front of the net, backcourt forehand chip, backcourt forehand high ball, backcourt forehand smash, and midfield forehand raid.
在一實施例中,擊球動作一致性評估步驟係透過一一致性估測演算法對擊球動作進行一致性比對;其中,一致性估測演算法包括一動作訊號分割步驟、一樣板挑選步驟、一面積邊界動態時間扭曲估測步驟、及一一致性評估步驟;其中,樣板挑選步驟包括一樣板訊號的取得,樣板訊號係一使用者利用手持球具進行擊球動作所產生的感測訊號經由邊界面積計算所重新取樣後的訊號。In one embodiment, the step of estimating the consistency of the batting action is to compare the batting action with a consistency estimation algorithm; wherein, the consistency estimation algorithm includes a motion signal segmentation step, a sample A selection step, an area boundary dynamic time warp estimation step, and a consistency evaluation step; wherein, the template selection step includes obtaining a template signal, and the template signal is generated by a user using a hand-held ball to hit the ball. The sensing signal is the resampled signal calculated by the boundary area.
承上所述,在本發明的手持運動分析系統與方法中,包括:姿態估測步驟,其依據感測訊號執行擊球動作的手持球具運動姿態估測;擊球軌跡重建步驟,其依據感測訊號和姿態估測步驟的結果執行擊球軌跡訊號的重建;擊球分期偵測步驟,其依據感測訊號和姿態估測步驟的結果區分擊球過程的不同時期;擊球球種辨識步驟,其依據感測訊號將擊球的球種類型進行分類;以及擊球動作一致性評估步驟,其依據感測訊號計算和評估擊球動作與樣板動作之間的一致性。藉此,相較於習知的擊球動作分析系統來說,本發明的手持運動分析系統與分析方法除了具有方便、專業且實用經濟的優點外,還可即時且客觀地提供相關的擊球指標給運動員或/及教練參考,進而改進運動員的擊球動作。Continuing from the above, the hand-held motion analysis system and method of the present invention includes: an attitude estimation step, which performs the motion attitude estimation of the hand-held ball equipment for hitting the ball according to the sensing signal; and a hitting trajectory reconstruction step, which is based on the The sensing signal and the result of the attitude estimation step perform reconstruction of the batting trajectory signal; the batting stage detection step, which distinguishes different periods of the batting process according to the sensing signal and the result of the attitude estimation step; a step of classifying the types of balls hit according to the sensing signal; and a step of evaluating the consistency of the hitting action, which calculates and evaluates the consistency between the hitting action and the sample action according to the sensing signal. Therefore, compared with the conventional batting action analysis system, the hand-held motion analysis system and analysis method of the present invention not only have the advantages of convenience, professionalism, practicality and economy, but also provide relevant batting action instantly and objectively. Indicators can be used as a reference for players and/or coaches to improve athlete's hitting action.
以下將參照相關圖式,說明依本發明實施例之手持運動分析系統與分析方法,其中相同的元件將以相同的參照符號加以說明。The following will describe the handheld motion analysis system and analysis method according to the embodiments of the present invention with reference to the related drawings, wherein the same elements will be described with the same reference signs.
圖1A為本發明一實施例之一種手持運動分析系統的功能方塊示意圖,圖1B為圖1A的手持運動分析系統的伺服器的功能方塊圖,而圖2為本發明之手持運動分析方法的一流程步驟示意圖。1A is a functional block diagram of a handheld motion analysis system according to an embodiment of the present invention, FIG. 1B is a functional block diagram of a server of the handheld motion analysis system of FIG. 1A , and FIG. 2 is a schematic diagram of a handheld motion analysis method of the present invention. Schematic diagram of the process steps.
手持運動分析系統1可應用於分析一手持球具的擊球動作。在此,手持運動可例如但不限於羽毛球、網球、桌球、棒球、或高爾夫球等、或其他利用手持球具擊球的運動,因此,上述的手持球具可為羽毛球拍、網球拍、桌球拍、球棒、或高爾夫球桿、或其他擊球的球具。以下實施例的手持球具是以羽毛球拍為例。因此,本文出現的“擊球”動作就是“揮羽毛球拍”的動作,或稱“揮拍”動作。當然,如果應用在高爾夫球運動時,擊球動作就是揮擊高爾夫球桿的動作,以此類推。另外,本文中的“運動員”指的是,在教練的指導下進行擊球訓練的人員。The hand-held
請參照圖1A及圖1B所示,本實施例的手持運動分析系統1包括一訊號感測模組11、一伺服器12以及一顯示模組13。Referring to FIG. 1A and FIG. 1B , the handheld
訊號感測模組11設置於手持球具中。訊號感測模組11可感測運動員拿著手持球具(如羽手球拍)的擊球動作(如揮拍動作)並輸出一感測訊號SS。其中,訊號感測模組11例如但不限於設置於手持球具的握柄內。以羽毛球拍為例,訊號感測模組11例如可設置於羽毛球拍的握柄內,或握柄的後套內,然並不以此為限,在不同的實施例中,訊號感測模組11也可裝設於手持球具的其他部位,例如握柄的其他位置或中管內。The
以下實施例是將訊號感測模組11設置於勝利體育事業股份有限公司提供的羽毛球拍的握柄內為例。因此,當運動員拿著該羽毛球拍進行揮拍動作時,訊號感測模組11可感測運動員的擊球(揮拍)動作並輸出感測訊號SS。 關於勝利公司提供的羽毛球拍的具體結構可參照中華民國發明專利證書號:TW I673088,在此不再多作說明。The following embodiment is an example in which the
本實施例的訊號感測模組11包括慣性感測器,例如包括三軸加速度計、三軸陀螺儀及三軸磁力計,藉此得到更為精準的擊球(揮拍)動作。因此,感測訊號SS為慣性感測訊號,其可包括揮拍動作過程的加速度訊號、角速度訊號及磁力訊號。在一些實施例中,可使用包含加速度計及陀螺儀的例如六軸感測器(如ICM-20649)及三軸磁力計(如LIS2MDL)做為九軸的慣性感測器。其中,加速度計用以感測地球重力及運動動作所產生的運動加速度;陀螺儀用以感測運動動作所產生的角速度;而磁力計用以感測地球磁場向量,經由運算後可獲得方位角資訊。The
在一些實施例中,訊號感測模組11還可包括微控制單元和電源供應單元。電源供應單元可例如為鋰電池,其可提供訊號感測模組11所需電力;而微控制單元可擷取並收集慣性感測器(速度計、陀螺儀及磁力計)因擊球動作所產生的感測訊號SS且進行處理(例如暫存及編碼),經處理過後的感測訊號SS可透過例如批次的方式利用例如Wi-Fi模組或藍牙(bluetooth)模組無線傳輸至伺服器12,以進行擊球動作的分析。In some embodiments, the
伺服器12與訊號感測模組11耦接。在一些實施例中,伺服器12與訊號感測模組11的耦接可為無線方式耦接,例如透過Wi-Fi模組或藍牙模組無線耦接,藉此接收、儲存及處理訊號感測模組11輸出的感測訊號SS。伺服器12可為本地伺服器(local server)、遠端伺服器(remote server)、或雲端伺服器(cloud server)。本實施例的伺服器12是以雲端伺服器為例。The
伺服器12可包括一或多個處理單元121及一記憶單元122,一或多個處理單元121與記憶單元122耦接。圖1B是以一個處理單元121與一個記憶單元122為例。處理單元121可存取記憶單元122所儲存的資料,並可包含伺服器12的核心控制組件,例如可包含至少一中央處理器(CPU)及一記憶體,或包含其它控制硬體、軟體或韌體。另外,記憶單元122可為一非暫態電腦可讀取記錄媒體(non-transitory computer readable storage medium),例如可包含至少一記憶體、一記憶卡、一光碟片、一錄影帶、一電腦磁帶,或其任意組合。在一些實施例中,前述的記憶體可包含唯讀記憶體(ROM)、快閃(Flash)記憶體、可程式化邏輯閘陣列(Field-Programmable Gate Array, FPGA)、或固態硬碟(Solid State Disk, SSD)、或其他形式的記憶體,或其組合。The
由於本實施例的伺服器12是以雲端伺服器為例,因此,記憶單元122為一雲端記憶體,而處理單元121則為雲端處理器。當感測訊號SS傳送至伺服器12時(伺服器12具有對應的無線傳輸模組),可儲存於記憶單元122,以供處理單元121處理及分析。另外,記憶單元122還可儲存至少一應用軟體,該應用軟體可包含一或多個程式指令1221,當該應用軟體的該一或多個程式指令1221被該一或多個處理單元121執行時,如圖2所示,該一或多個處理單元121可至少進行以下步驟:一姿態估測步驟S2、一擊球軌跡重建步驟S3、一擊球分期偵測步驟S4、一擊球球種辨識步驟S5、以及一擊球動作一致性評估步驟S6。另外,除了步驟S2至步驟S6之外,在取得感測訊號SS後,本實施例的處理單元121還可進行一訊號前處理步驟S1。以下,請參照圖3以說明上述步驟S1至步驟S6的詳細技術內容。Since the
圖3為本發明之手持運動分析方法的另一流程步驟示意圖。在此,圖3除了顯示圖2的步驟S1至步驟S6內部的詳細流程步驟(或稱子步驟)外,還顯示一結果呈現步驟S7。先說明的是,本文中提到的步驟S1至步驟S7及其內部的功能方塊(步驟)可以軟體程式方式實現其功能,或者,也可應用硬體或韌體的方式實現其功能,本發明不限制。FIG. 3 is a schematic diagram of another process step of the handheld motion analysis method of the present invention. Here, in addition to showing the detailed flow steps (or sub-steps) inside steps S1 to S6 of FIG. 2 , FIG. 3 also shows a result presentation step S7 . First of all, it should be noted that the steps S1 to S7 mentioned in this article and their internal functional blocks (steps) can realize their functions in the form of software programs, or can also realize their functions in the form of hardware or firmware. The present invention not limited.
如圖3所示,在進行姿態估測步驟S2之前,為了使後續的分析過程及其產生的結果更為準確,需先進行訊號前處理步驟S1。As shown in FIG. 3 , before performing the attitude estimation step S2 , in order to make the subsequent analysis process and the generated results more accurate, the signal preprocessing step S1 needs to be performed first.
訊號前處理步驟 S1 :係校正訊號感測模組11輸出的感測訊號SS,並濾除感測訊號SS中的雜訊。其中,訊號前處理步驟S1可包括一訊號校正步驟S11及一訊號濾波步驟S12。訊號校正步驟S11可校正感測訊號SS,而訊號濾波步驟S12可濾除感測訊號SS中的雜訊。以下詳細介紹步驟S11及步驟S12的技術內容。
The signal preprocessing step S1 : correcting the sensing signal SS output by the
訊號校正步驟S11:由於慣性感測器本身特性及其他外界環境因素所影響,往往會造成加速度計、陀螺儀及磁力計所量測到的感測訊號SS會產生量測誤差或訊號漂移,在一些實施例中,可以使用比例因子(scale factor, SF)及偏移量(bias, B)來校正慣性感測器。其校正過程可如下;依次將加速度計及陀螺儀擺放於水平放置的旋轉運動平台上14個不同方位,並藉由加速度計在靜止擺放狀況下其三個軸向(X軸、Y軸、Z軸)感測值合力為重力加速度讀值(1g),以及在使用旋轉運動平台定速旋轉下陀螺儀之三軸感測值合力值為等角速度讀值(𝜔)的物理現象,來加以獲得三軸慣性感測器之比例因子(𝑆F 𝑥、𝑆F 𝑦、𝑆F 𝑧)及偏差值(𝐵 𝑥、𝐵 𝑦、𝐵 𝑧)。而在校正磁力計時,可以在一個沒有強力磁場干擾環境下,將磁力計在一固定時間下進行三維空間均勻旋轉,使磁力感測值可涵蓋到三維空間中的各個方位,並將其感測磁場合力等比正規化為1高斯(Gauss)的固定常數。最後,透過最小平方誤差法求解出各感測器之校正矩陣(𝐐)後,即可透過以下式(1)來加以校正慣性感測器。 (1) Signal calibration step S11: Due to the characteristics of the inertial sensor itself and other external environmental factors, the sensing signal SS measured by the accelerometer, gyroscope and magnetometer will often produce measurement errors or signal drift. In some embodiments, the inertial sensor may be calibrated using a scale factor (SF) and an offset (bias, B). The calibration process can be as follows; the accelerometer and gyroscope are placed in 14 different directions on the horizontal rotating motion platform in turn, and the three axes (X axis, Y axis) of the accelerometer are placed in a static state. , Z axis) The resultant force of the sensing value is the gravitational acceleration reading (1g), and the resultant force of the three-axis sensing value of the gyroscope under the constant speed rotation of the rotating motion platform is the physical phenomenon of the equal angular velocity reading (𝜔), to Add the scale factor (𝑆F 𝑥 , 𝑆F 𝑦 , 𝑆F 𝑧 ) and the bias value (𝐵 𝑥 , 𝐵 𝑦 , 𝐵 𝑧 ) of the three-axis inertial sensor. When calibrating the magnetometer, in an environment without strong magnetic field interference, the magnetometer can be rotated uniformly in three-dimensional space at a fixed time, so that the magnetic sensing value can cover all directions in the three-dimensional space, and it can be sensed The net force of the magnetic field is proportionally normalized to a fixed constant of 1 Gauss. Finally, after solving the calibration matrix (𝐐) of each sensor through the least square error method, the inertial sensor can be calibrated through the following equation (1). (1)
其中,𝐒 𝑖為未校正過的加速度計、陀螺儀或磁力計的感測值,而𝐒 𝑐為校正後的加速度計、陀螺儀或磁力計的感測值。 Among them, 𝐒 𝑖 is the sensing value of the uncalibrated accelerometer, gyroscope or magnetometer, and 𝐒 𝑐 is the sensing value of the corrected accelerometer, gyroscope or magnetometer.
訊號濾波步驟S12:當運動員使用例如羽毛球拍進行擊球運動時,其所量測到的感測訊號SS包含了運動動作訊號、高頻雜訊及動作雜訊(例如身體不自主的顫抖),因此,為了能準確地量測到運動員運動時所產生的感測訊號SS,經上述步驟S11校正過後的慣性感測訊號仍需要經由低通濾波器來降低高頻雜訊及動作雜訊,使得到的感測訊號SS可實際反應出擊球動作的真正訊號。以下,經由訊號前處理步驟S1的校正與濾波處理後的慣性感測訊號仍標示為感測訊號SS。Signal filtering step S12: When the player uses a badminton racket to hit the ball, the sensed signal SS measured by the player includes motion signal, high-frequency noise and motion noise (such as involuntary body tremors), Therefore, in order to accurately measure the sensing signal SS generated when the athlete is exercising, the inertial sensing signal corrected in the above step S11 still needs to pass through a low-pass filter to reduce high-frequency noise and motion noise, so that the The received sensing signal SS can actually reflect the real signal of the hitting action. In the following, the inertial sensing signal after the calibration and filtering of the signal preprocessing step S1 is still denoted as the sensing signal SS.
姿態估測步驟 S2 :係依據感測訊號SS執行擊球動作的手持球具姿態估測。其中,姿態估測步驟S2係透過一擴展式卡爾曼濾波器演算法利用感測訊號SS進行擊球動作的手持球具姿態估測,進而獲得準確的手持球具速度及手持球具軌跡。在此,擴展式卡爾曼濾波器演算法可包括一狀態預測步驟S21、一重力狀態更新步驟S22、及一磁北狀態更新步驟S23。在狀態預測步驟S21中,係利用感測訊號SS中的角速度訊號進行擴展式卡爾曼濾波器的狀態預測;而在重力狀態更新步驟S22和磁北狀態更新步驟S23中,係利用感測訊號SS中的加速度訊號與磁力訊號進行擴展式卡爾曼濾波器的狀態更新,以得到最佳手持球具姿態估測狀態。以下詳細介紹步驟S21至步驟S23的技術內容。 Attitude Estimation Step S2 : Estimating the posture of the hand-held golf equipment according to the sensing signal SS to perform the hitting action. The attitude estimation step S2 utilizes the sensing signal SS through an extended Kalman filter algorithm to estimate the attitude of the golf equipment in the hitting action, so as to obtain the accurate speed of the golf equipment and the trajectory of the golf equipment. Here, the extended Kalman filter algorithm may include a state prediction step S21 , a gravity state update step S22 , and a magnetic north state update step S23 . In the state prediction step S21, the angular velocity signal in the sensing signal SS is used to predict the state of the extended Kalman filter; and in the gravity state updating step S22 and the magnetic north state updating step S23, the sensing signal SS is used in the state prediction step S22. The state of the extended Kalman filter is updated by the acceleration signal and the magnetic signal, so as to obtain the best posture estimation state of the hand-held ball. The technical contents of steps S21 to S23 are described in detail below.
狀態預測步驟S21:可以四元數表示之姿態定義為狀態轉移方程式之狀態變數(
),利用陀螺儀現在時間點(𝑡)所感測之角速度(𝛚
𝑡)及上一個時間點(𝑡–1)更新後之姿態角(
)建立一狀態轉移方程式,如式(2)所示。
其中, 為上一個時間點更新後之狀態誤差共變異數矩陣, 為角速度雜訊共變異數矩陣。 in, is the state error covariance matrix after the update at the last time point, is the angular velocity noise covariance matrix.
重力狀態更新步驟S22及磁北狀態更新步驟S23:由於角速度進行狀態預測所產生之誤差,將會隨時間增加而累積,因此,必須透過觀測量( )對預測後的狀態進行更新,在此,將觀測量定義為加速度與磁力值。另外,再建立一重力或磁北觀測方程式對角速度所預測之姿態 )進行狀態更新,如式(4)所示。 (4) Gravity state update step S22 and magnetic north state update step S23: The error generated by the state prediction due to the angular velocity will accumulate over time, so it must be observed through the measurement ( ) updates the predicted state, where the observed quantities are defined as acceleration and magnetic force values. In addition, establish a gravity or magnetic north observation equation to predict the attitude of the angular velocity ) to update the state, as shown in formula (4). (4)
其中, 為現在時間點之重力觀測矩陣或磁北觀測矩陣, 為現在時間點所預測之狀態變數, 為現在時間點的加速度或磁力值之白雜訊, 為現在時間點的預測觀測量。接著,獲得重力或磁北觀測方程式之後,便可計算現在時間點的重力或磁北更新卡爾曼增益 ,如式(5)所示。最後,即可利用重力或磁北更新卡爾曼增益對所預測之狀態及其狀態誤差共變異數矩陣進行狀態更新,如式(6)及式(7)所示。 (5) (6) (7) in, is the gravity observation matrix or the magnetic north observation matrix at the present time point, is the state variable predicted for the current time point, is the white noise of the acceleration or magnetic value at the current time point, is the predicted observation amount for the current time point. Then, after obtaining the gravity or magnetic north observation equation, you can calculate the gravity or magnetic north at the current time point to update the Kalman gain , as shown in formula (5). Finally, the predicted state and its state error covariance matrix can be updated by using gravity or magnetic north to update the Kalman gain, as shown in equations (6) and (7). (5) (6) (7)
其中, 為觀測量雜訊共變異數矩陣,當觀測量為加速度時, 為加速度雜訊共變異數矩陣( );而若觀測量為磁力值時, 即為磁力雜訊共變異數矩陣( ), 為現在時間點的重力或磁北實際觀測量。 in, is the observed quantity noise covariance matrix, when the observed quantity is acceleration, is the acceleration noise covariance matrix ( ); and if the observed quantity is a magnetic value, is the magnetic noise covariance matrix ( ), It is the actual observed amount of gravity or magnetic north at the current time point.
擊球軌跡重建步驟 S3 :係依據感測訊號SS和姿態估測步驟S2的結果執行擊球軌跡訊號的重建。其中,擊球軌跡重建步驟S3係透過一軌跡重建演算法取得執行擊球動作時所產生的手持球具姿態、手持球具速度及手持球具軌跡訊號。在此,軌跡重建演算法可包括一動作訊號分割步驟S31、一座標轉換與重力補償步驟S32、一速度估測與零速度補償步驟S33、及一軌跡重建步驟S34。提醒的是,廣義來說,軌跡重建演算法也可包括上述的狀態預測步驟S21、重力狀態更新步驟S22、及磁北狀態更新步驟S23。以下詳細介紹步驟S31至步驟S34的技術內容。 Step S3 of Reconstructing the Shot Track : Reconstructing the shot track signal according to the sensing signal SS and the result of the attitude estimation step S2. Wherein, the batting trajectory reconstruction step S3 uses a trajectory reconstruction algorithm to obtain the hand-held golf equipment posture, the hand-held golf equipment speed and the hand-held golf equipment trajectory signals generated when the batting action is performed. Here, the trajectory reconstruction algorithm may include a motion signal segmentation step S31 , a coordinate conversion and gravity compensation step S32 , a speed estimation and zero-speed compensation step S33 , and a trajectory reconstruction step S34 . It is reminded that, broadly speaking, the trajectory reconstruction algorithm may also include the above-mentioned state prediction step S21 , gravity state update step S22 , and magnetic north state update step S23 . The technical contents of steps S31 to S34 are described in detail below.
動作訊號分割步驟S31:由於揮拍動作進行時,在揮拍動作開始前及結束後均會存在一段動作靜態區間,此時,因為訊號感測模組11處於靜止,所以加速度計及陀螺儀的三軸合力值皆為0,因此,可以透過設定一動態門檻值來加以偵測揮拍動作區間,例如依據前200個取樣點之感測訊號,計算其標準分數(z-score),並以標準分數作為動態門檻值。Action signal segmentation step S31 : Since the swing action is in progress, there will be a static period of action before and after the swing action. At this time, because the
座標轉換與重力補償步驟S32:藉由姿態估測步驟S2可獲得運動員在運動期間的動態手持球具姿態後,即可獲得感測器座標系( s)與參考座標系( r)之間的座標轉換矩陣( ),如式(8)所示,並將感測器座標系( s)上濾波過後的加速度訊號( )轉換成參考座標系( r)上的加速度訊號( ),如式(9)所示。此外,由於加速度計所量測的加速度值會同時包含了運動所產生的運動加速度及重力加速度,因此,需將重力加速度( )移除,進而得到真實的運動加速度( )。 (8) . (9) Step S32 of coordinate conversion and gravity compensation: After obtaining the dynamic posture of the player during the exercise, the posture between the sensor coordinate system ( s ) and the reference coordinate system ( r ) can be obtained through the posture estimation step S2 . Coordinate transformation matrix ( ), as shown in equation (8), and the filtered acceleration signal ( s ) on the sensor coordinate system ( s ) ) into the acceleration signal ( r ) on the reference coordinate system ( r ) ), as shown in formula (9). In addition, since the acceleration value measured by the accelerometer includes both the motion acceleration and the gravitational acceleration generated by the movement, the gravitational acceleration ( ) is removed, and then the real motion acceleration ( ). (8) . (9)
速度估測與零速度補償步驟S33:當獲得運動加速度後,即可將運動加速度進行積分以估測其速度訊號,如式(10)所示。由於加速度訊號容易受到人體無意識顫抖之干擾而產生雜訊,而在進行速度估測之積分運算時,其加速度訊號之雜訊因積分運算而被放大,導致速度訊號的失真,故透過式(11)進行零速度更新,以補償失真的速度訊號。 (10) (11) Speed Estimation and Zero Speed Compensation Step S33: After the motion acceleration is obtained, the motion acceleration can be integrated to estimate the speed signal, as shown in equation (10). Since the acceleration signal is easily disturbed by the unconscious tremor of the human body, noise is generated, and when the integral operation of the velocity estimation is performed, the noise of the acceleration signal is amplified by the integral operation, resulting in the distortion of the velocity signal. Therefore, through the formula (11 ) for a zero speed update to compensate for a distorted speed signal. (10) (11)
其中, 為現在時間點之速度訊號; 為上個時間之速度訊號; 為經過零速度更新之速度訊號; 為訊號起始點之速度值; 為訊號結束點之速度值; 為時間間距; 為取樣週期。 in, is the speed signal at the current time point; is the speed signal of the previous time; is the speed signal after zero speed update; is the speed value of the starting point of the signal; is the speed value of the signal end point; is the time interval; is the sampling period.
軌跡重建步驟S34:將零速度補償後之速度訊號進行積分運算,即可重建運動員進行揮拍或移動動作時之軌跡,如式(12)所示。 (12) Trajectory reconstruction step S34: Integrate the velocity signal after zero velocity compensation to reconstruct the trajectory of the athlete when swinging or moving, as shown in equation (12). (12)
其中, 為現在時間點之運動軌跡; 為上個時間之運動軌跡; 為取樣週期。 in, is the trajectory of the current time point; is the trajectory of the last time; is the sampling period.
擊球分期偵測步驟 S4 :係依據感測訊號SS和姿態估測步驟S2的結果區分擊球過程的不同時期。其中,擊球分期偵測步驟S4係透過一擊球動作分期演算法取得擊球過程中的每一時期訊號。在此,擊球動作分期演算法可包括一動作訊號分割步驟S41、一座標轉換與重力補償步驟S42、一動作訊號極點偵測步驟S43、及一動作訊號分期偵測步驟S44。以下詳細介紹步驟S41至步驟S44的技術內容。 Step S4 of batting stage detection : Different stages of the batting process are distinguished according to the sensing signal SS and the result of the attitude estimation step S2. Wherein, the step S4 of the batting stage detection is to obtain each period signal in the batting process through a batting action staged algorithm. Here, the batting action staging algorithm may include a motion signal segmentation step S41 , a coordinate conversion and gravity compensation step S42 , a motion signal pole detection step S43 , and a motion signal period detection step S44 . The technical contents of steps S41 to S44 are described in detail below.
動作訊號分割步驟S41及座標轉換與重力補償步驟S42與上述軌跡重建演算法中的動作訊號分割步驟S31及座標轉換與重力補償步驟S32相同(亦即動作訊號分割步驟S31及座標轉換與重力補償步驟S32的結果可以應用於擊球分期偵測步驟S4中),在此不再多作說明。The motion signal segmentation step S41 and the coordinate conversion and gravity compensation step S42 are the same as the motion signal segmentation step S31 and the coordinate conversion and gravity compensation step S32 in the above trajectory reconstruction algorithm (that is, the motion signal segmentation step S31 and the coordinate conversion and gravity compensation step S31). The result of S32 can be applied to step S4 of batting stage detection), which will not be further described here.
動作訊號極點偵測步驟S43:可透過本步驟找出擊球動作的預備期起始點(start point)、加速期起始點(start point of acceleration)、擊球點(impact)及餘勢期結束點(end point)。在此,預備期起始點可定義為:將動作動態區間起始點定義為預備期起始點,並藉此區分初始靜止期與預備期。擊球點可定義為:在整個羽球揮拍過程中,球拍與球接觸的瞬間通常發生在球拍角速度最大值的瞬間,因此,可以透過此特性將整個羽球揮拍訊號中角速度合力訊號產生最大值時的時間點定義為擊球點,並藉此區分加速期及餘勢期。加速期起始點可定義為:當找到擊球時間點後,利用該時間點上的角速度值往回尋找第一個角速度訊號之波谷極點,即為加速期起始點,並藉此區分預備期與加速期。餘勢期結束點可定義為:將動作動態區間終止點定義為餘勢期結束點,並藉此區分餘勢期與結束靜止期。如圖4所示,其為擊球動作的訊號分期偵測示意圖。在此,圖4顯示感測訊號中的加速度和角速度訊號的分期。Action signal pole detection step S43: Through this step, the starting point of the preparatory period (start point), the starting point of the acceleration period (start point of acceleration), the impact point (impact) and the after-potential period of the hitting action can be found through this step end point. Here, the starting point of the preparatory period can be defined as: defining the starting point of the action dynamic interval as the starting point of the preparatory period, and thereby distinguishing the initial stationary period and the preparatory period. The hitting point can be defined as: in the whole badminton swing process, the moment when the racket contacts the ball usually occurs at the moment when the angular velocity of the racket reaches the maximum value. Therefore, this characteristic can be used to generate the maximum value of the angular velocity resultant force signal in the entire badminton swing signal. The time point at 2000 is defined as the hitting point, and thereby distinguishes the acceleration period and the remaining momentum period. The starting point of the acceleration period can be defined as: after finding the time point of hitting the ball, use the angular velocity value at this time point to search back for the trough of the first angular velocity signal, which is the starting point of the acceleration period, and use this to distinguish the preparation period and accelerated period. The end point of the residual period can be defined as: the end point of the action dynamic interval is defined as the end point of the residual period, and thus the residual period and the end stationary period are distinguished. As shown in FIG. 4 , it is a schematic diagram of signal detection by stages of a batting action. Here, FIG. 4 shows the staging of the acceleration and angular velocity signals in the sensed signal.
動作訊號分期偵測步驟S44:在完成動作訊號極點偵測步驟S43後,即可將擊球過程的不同時期定義出以下五個分期:1)在預備期起始點之前的時間區間可定義為初始靜止期(initial rest);2)在預備期起始點與加速期起始點之間的時間區間可定義為預備期(preparation);3)在加速期起始點與擊球點之間的時間區間可定義為加速期(acceleration);4)在擊球點與餘勢期結束點之間的時間區間可定義為餘勢期(follow through);5)在餘勢期結束點之後的時間區間可定義為結束靜止期(ending rest)。Action signal stage detection step S44: After completing the action signal pole detection step S43, the following five stages can be defined for different periods of the hitting process: 1) The time interval before the starting point of the preparatory period can be defined as The initial rest period (initial rest); 2) the time interval between the start point of the preparation period and the start point of the acceleration period can be defined as the preparation period (preparation); 3) between the start point of the acceleration period and the hitting point The time interval can be defined as the acceleration period (acceleration); 4) the time interval between the hitting point and the end of the after-potential period can be defined as the after-potential period (follow through); 5) after the end of the after-potential period A time interval can be defined as an ending rest.
擊球球種辨識步驟 S5 :係依據感測訊號SS將擊球的球種類型進行分類。其中,擊球球種辨識步驟S5係透過一擊球球種辨識演算法取得擊球球種類型。在此,擊球球種辨識演算法可包括一動作訊號分割步驟S51、一訊號正規化步驟S52、一卷積神經網路分類步驟S53、及一球種辨識步驟S54。以下詳細介紹步驟S51至步驟S54的技術內容。 The step S5 of identifying the ball type of the shot is to classify the type of the shot ball according to the sensing signal SS. Wherein, the ball type identification step S5 is to obtain the ball type through a ball type identification algorithm. Here, the batting ball identification algorithm may include a motion signal segmentation step S51 , a signal normalization step S52 , a convolutional neural network classification step S53 , and a ball identification step S54 . The technical contents of steps S51 to S54 are described in detail below.
動作訊號分割步驟S51與上述軌跡重建演算法的動作訊號分割步驟S31相同(亦即動作訊號分割步驟S31的結果可以應用於擊球球種辨識步驟S5中),在此不再多作說明。The motion signal segmentation step S51 is the same as the motion signal segmentation step S31 of the above-mentioned trajectory reconstruction algorithm (that is, the result of the motion signal segmentation step S31 can be applied to the hitting ball type identification step S5 ), and will not be described here.
訊號正規化步驟S52:將慣性感測訊號經過上述訊號校正步驟S11、訊號濾波步驟S12及動作訊號分割步驟S51後,即可進行訊號正規化步驟S52,以正規化感測訊號SS。Signal normalization step S52 : After the inertial sensing signal is subjected to the signal calibration step S11 , the signal filtering step S12 and the motion signal segmentation step S51 , the signal normalization step S52 can be performed to normalize the sensing signal SS.
卷積神經網路分類步驟S53及球種辨識步驟S54:經訊號正規化步驟S52之後的感測訊號SS中,每個擊球揮拍動作訊號中的三軸角速度訊號可作為卷積神經網路(Convolution Neural Network, CNN)分類器的輸入,進而分類出正手發後場球、反手發前場球、正手挑後場高遠球、反手挑後場高遠球、正手推挑後場球、反手推挑後場球、前場正手短球、前場反手短球、中場正手平抽球、中場反手平抽球、中場正手接殺擋網前球、中場反手接殺擋網前球、後場正手切球、後場正手高遠球、後場正手殺球、及中場正手突襲球等十六種擊球球種類型。在此,卷積神經網路分類器的架構可包含兩卷積層、兩池化層、一全連接層與一輸出層,詳述如下:Convolutional Neural Network Classification Step S53 and Ball Type Identification Step S54: In the sensing signal SS after the signal normalization step S52, the triaxial angular velocity signal in each batting swing motion signal can be used as a convolutional neural network (Convolution Neural Network, CNN) the input of the classifier, and then classify the forehand to the backcourt, the backhand to the frontcourt, the forehand to pick the backcourt high ball, the backhand to pick the backcourt high ball, the forehand to push the backcourt ball, the backhand to push the backcourt Ball, Front forehand short ball, Front backhand short ball, Midfield forehand draw, Midfield backhand draw, Midfield forehand catch and block front ball, Midfield backhand catch and block front ball, Back field There are 16 types of shots, such as hand-cut, backcourt forehand high ball, backcourt forehand smash, and midfield forehand raid ball. Here, the architecture of the convolutional neural network classifier may include two convolutional layers, two pooling layers, a fully connected layer and an output layer, as detailed below:
卷積層:每個卷積層中含有多個卷積核,透過設定之卷積核大小並利用卷積原理進行視窗之逐步滑動並加權計算每個區域內數值,再經由活化函數計算獲得卷積層之輸出,藉此提取輸入訊號中重要的資訊。其中,在每層卷積層皆設置了128個大小為1×5的卷積核(convolutional kernel/filter)來進行圖像特徵的擷取。 Convolutional layer: Each convolutional layer contains multiple convolution kernels. By setting the size of the convolutional kernel and using the convolution principle, the window gradually slides and the values in each area are weighted to calculate the value of the convolutional layer. output, thereby extracting important information from the input signal. Among them, 128 convolutional kernels/filters with a size of 1×5 are set in each convolutional layer to extract image features.
其中, 為三軸角速度訊號所組成之輸入向量; 為每個步伐視窗中的資料點索引; 為每個步伐視窗中的資料點數; 為層的索引; 為卷積核大小(kernel/filter size); 為第 層的第 個特徵映射(feature map)之偏權值; 為輸入 與第 層第 個特徵映射之連結權重; 為線性整流活化函數。 in, is the input vector composed of three-axis angular velocity signals; index for the data point in each step window; is the number of data points in each step window; is the index of the layer; is the convolution kernel size (kernel/filter size); for the first the first The partial weights of a feature map; for input with the first layer The link weight of each feature map; is the linear rectification activation function.
池化層:其主要是將卷積層輸出作為其輸入並進行下採樣,在此採用最大池化(max pooling)運算,藉此降低特徵映射維度(網路訓練參數),而僅保留輸入圖像中的重要特徵。池化大小為1×2,跨度為2。其中, 為池化大小; 為池化的跨度 Pooling layer: It mainly uses the output of the convolutional layer as its input and performs downsampling. Here, the max pooling operation is used to reduce the dimension of the feature map (network training parameters), while only retaining the input image important features in . The pooling size is 1×2 and the stride is 2. in, is the pooling size; is the pooling span
全連接層:將經過多層卷積層與池化層運算後所得之特徵攤平成一特徵向量 作為該層的輸入,其中 為最後一層池化層的神經元個數,並進行以下運算: ) Fully connected layer: Flatten the features obtained after multi-layer convolutional layers and pooling layers into a feature vector as the input to this layer, where is the number of neurons in the last pooling layer, and performs the following operations: )
其中, 為全連結層中第 層的第 個神經元與第 層的第 個神經元連結權重值; 為全連結層中第 層的第 個神經元的偏權值; 為線性整流活化函數。最後, 即為經過卷積神經網路運算所得之深度特徵。 in, is the first in the fully connected layer the first neuron and the first neuron connection weight value; is the first in the fully connected layer the first The partial weights of each neuron; is the linear rectification activation function. at last, It is the depth feature obtained by the operation of the convolutional neural network.
輸出層:通常是以分類器進行,在此使用的是Softmax分類器,Softmax分類器為以log-sigmoid函數為基礎,X軸範圍為正無窮大至負無窮大,Y軸範圍為0至1,透過將全連接層之輸出映射至[0,1]區間內,將所得值轉換成相對應的機率,並取最大值為分類結果。Output layer: It is usually performed by a classifier. Here, the Softmax classifier is used. The Softmax classifier is based on the log-sigmoid function. The X-axis ranges from positive infinity to negative infinity, and the Y-axis ranges from 0 to 1. Map the output of the fully connected layer to the [0,1] interval, convert the obtained value into the corresponding probability, and take the maximum value as the classification result.
擊球動作一致性評估步驟 S6 :係依據感測訊號SS計算和評估擊球動作與樣板動作之間的一致性。其中,擊球動作一致性評估步驟S6係透過一一致性估測演算法對擊球動作進行一致性比對。在此,一致性估測演算法包括一動作訊號分割步驟S61、一樣板挑選步驟S62、一面積邊界動態時間扭曲估測步驟S63、及一一致性評估步驟S64。前述的樣板挑選步驟S62包括一樣板訊號的取得,該樣板訊號係一使用者利用手持球具進行擊球動作所產生的感測訊號SS經由邊界面積計算所重新取樣後的動作訊號。以下詳細介紹步驟S61至步驟S64的技術內容。 Step S6 of evaluating the consistency of the hitting action : calculating and evaluating the consistency between the hitting action and the template action according to the sensing signal SS. Wherein, in the step S6 of evaluating the consistency of the batting action, a consistency estimating algorithm is used to compare the consistency of the batting action. Here, the consistency estimation algorithm includes a motion signal segmentation step S61 , a sample selection step S62 , an area boundary dynamic time warp estimation step S63 , and a consistency evaluation step S64 . The aforesaid template selection step S62 includes obtaining a template signal, which is a motion signal resampled by the boundary area calculation of the sensing signal SS generated by the user using the hand-held ball to hit the ball. The technical contents of steps S61 to S64 are described in detail below.
動作訊號分割步驟S61與上述軌跡重建演算法的動作訊號分割步驟S31相同(亦即動作訊號分割步驟S31的結果可以應用於擊球動作一致性評估步驟S6中),在此不再多作說明。The motion signal segmentation step S61 is the same as the motion signal segmentation step S31 of the above-mentioned trajectory reconstruction algorithm (that is, the result of the motion signal segmentation step S31 can be applied to the batting action consistency evaluation step S6), and will not be further described here.
樣板挑選步驟S62:將該使用者與運動員相同的擊球動作所產生的角速度訊號經由以下介紹的邊界面積計算所重新取樣後的動作訊號視為樣板訊號( ),以供後續的比對。在此,該使用者可為具有較佳球技的人員,例如但不限於教練、或職業運動員。 Sample selection step S62: The angular velocity signal generated by the same hitting action of the user and the player is regarded as a sample signal ( ) for subsequent comparisons. Here, the user may be a person with better golf skills, such as, but not limited to, a coach, or a professional athlete.
面積邊界動態時間扭曲估測步驟S63:包括正負波峰偵測、過零點偵測、邊界面積計算及一致性分數計算等。其中,正負波峰偵測為:設定一門檻值來加以偵測經由動作訊號分割處理後的動作時序訊號 ( )中找出區域最大值及最小值,即為正負波峰。過零點偵測為:偵測動作訊號通過零值之取樣點(zero-crossing points, ZC points),藉此獲得過零點之取樣訊號( )。另外,邊界面積計算為:依據過零點的數量( ),將原始長度為 的動作時序訊號分割為( )個片段,並針對每一個片段進行訊號積分計算出每個片段的面積,如下所示;則積分後的面積即可代表重新取樣後的時間序列( )。 Area boundary dynamic time warp estimation step S63 : including positive and negative wave peak detection, zero-crossing point detection, boundary area calculation, and consistency score calculation, etc. Among them, the detection of positive and negative peaks is: setting a threshold value to detect the action timing signal ( ) to find the maximum and minimum values in the region, which are positive and negative peaks. The zero-crossing point detection is: to detect that the motion signal passes through the zero-crossing points (zero-crossing points, ZC points), thereby obtaining the zero-crossing point sampling signal ( ). In addition, the boundary area is calculated as: according to the number of zero crossings ( ), converting the original length to The action timing signal of is divided into ( ) segments, and perform signal integration for each segment to calculate the area of each segment, as shown below; then the integrated area can represent the resampled time series ( ).
此外,一致性分數計算為:將運動員執行運動訓練動作時所產生的角速度訊號經由邊界面積計算後視為運動動作訊號( ),並與樣板訊號( )進行訊號時序扭曲比對,藉由以下式子計算出運動動作訊號與樣板訊號之間的一致性分數( ),藉此達成運動訓練動作比對的目的。 In addition, the consistency score is calculated as: the angular velocity signal generated when the athlete performs the sports training action is regarded as the sports action signal after the calculation of the boundary area ( ), combined with the template signal ( ) to compare the signal timing distortion, and calculate the consistency score between the motion signal and the template signal by the following formula ( ), so as to achieve the purpose of sports training action comparison.
另外,動態時間扭曲是由二時序訊號起始時序累加歐基里德距離至終止時序。ABDTW分數計算可如下所示: In addition, the dynamic time warping is the accumulation of the Euclidean distance from the starting timing of the two timing signals to the ending timing. The ABDTW score can be calculated as follows:
其中, 。 in, .
一致性評估步驟S64:揮拍動作一致性為運動員每次進行揮拍動作時,動作訊號之間的相似程度。而揮拍動作一致性越高,代表其揮拍技術較為穩定,在整體表現上較佳;反之,揮拍動作一致性越低,代表其揮拍技術較不穩定。在此,係藉由面積邊界動態時間扭曲演算法(AB-DTW)所計算的一致性分數來評估運動員與例如教練動作的一致性比對。其中,分數越低,代表兩相比之運動訊號的一致性越高;反之,分數越高,代表兩相比之運動訊號的一致性越低。Consistency evaluation step S64: The swing consistency is the degree of similarity between motion signals when the athlete performs a swing each time. The higher the consistency of the swing action, the more stable the swing technique, and the better the overall performance; on the contrary, the lower the consistency of the swing action, the less stable the swing technique. Here, the athlete's alignment with, for example, a coach's movements is assessed by the conformance score calculated by the Area Boundary Dynamic Time Warping Algorithm (AB-DTW). Among them, the lower the score, the higher the consistency of the motion signals between the two comparisons; conversely, the higher the score, the lower the consistency of the motion signals between the two comparisons.
透過上述步驟進行運動動作分析後,可以得到運動員相關的擊球指標,例如包括揮拍軌跡、擊球球種辨識、揮拍分期、擊球次數、擊球速度、平均擊球速度、最大擊球速度、平均殺球速度、最大殺球速度、擊球力度、揮拍弧度、揮拍動作一致性等。After analyzing the movement movements through the above steps, the athlete-related batting indicators can be obtained, for example, including swing trajectory, bat type identification, swing stage, number of batting, batting speed, average batting speed, and maximum batting speed. Speed, average smashing speed, maximum smashing speed, hitting force, swing arc, swing consistency, etc.
結果呈現步驟 S7 :係透過與伺服器12耦接的顯示模組13呈現分析結果,讓運動員或/及教練參考,進而改進運動員的擊球動作。在一些實施例中,顯示模組13可為固定式顯示裝置(例如電腦)、或為行動裝置(例如筆記型電腦、手機、平板電腦),或其他型式的顯示裝置。在一些實施例中,顯示模組13可顯示例如即時訊號呈現、擊球速度、力度數據呈現、個人綜合表現評估雷達圖,或/及呈現擊球球種辨識結果及擊球動作分析(如揮拍分期、擊球次數、擊球速度、最大擊球速度、平均殺球速度、最大殺球速度、擊球力度、揮拍弧度、揮拍軌跡及揮拍動作一致性)等相關羽球專項指標,運動員或/及教練可在顯示模組13自行選擇觀看哪些訊號或指標。
Result presenting step S7 : presenting the analysis result through the
本發明還提出一種手持運動的分析方法,可應用於上述的手持運動分析系統1。其中,手持運動分析系統1的元件組成及其功能已於上述中詳述,在此不再多作說明。The present invention also provides an analysis method for hand-held motion, which can be applied to the above-mentioned hand-held
如圖2或圖3所示,手持運動的分析方法可包括姿態估測步驟S2、擊球軌跡重建步驟S3、擊球分期偵測步驟S4、擊球球種辨識步驟S5、以及擊球動作一致性評估步驟S6。另外,在姿態估測步驟S2之前,該分析方法更可包括訊號前處理步驟S1。此外,在上述步驟S3、步驟S4、步驟S5、步驟S6之後,該分析方法更可包括結果呈現步驟S7,以呈現分析結果。As shown in FIG. 2 or FIG. 3 , the method for analyzing the hand-held motion may include an attitude estimation step S2 , a hitting trajectory reconstruction step S3 , a hitting stage detection step S4 , a hitting ball type identification step S5 , and a consistent hitting action Sexuality evaluation step S6. In addition, before the attitude estimation step S2, the analysis method may further include a signal preprocessing step S1. In addition, after the above step S3, step S4, step S5, and step S6, the analysis method may further include a result presentation step S7 to present the analysis result.
手持運動分析方法的各步驟(包括步驟S1至步驟S7)及其內部(子)步驟的詳細技術內容已於上述中詳述,在此不再贅述。要提醒的是,在上述步驟S1至步驟S7中,姿態估測步驟S2、擊球球種辨識步驟S5及擊球動作一致性評估步驟S6可依序或同時進行,但擊球軌跡重建步驟S3與擊球分期偵測步驟S4需在姿態估測步驟S2之後進行,且擊球軌跡重建步驟S3與擊球分期偵測步驟S4可依序或同時進行。The detailed technical contents of each step (including step S1 to step S7 ) and its internal (sub) steps of the hand-held motion analysis method have been described in detail above, and will not be repeated here. It should be reminded that, in the above steps S1 to S7, the attitude estimation step S2, the hitting ball type identification step S5 and the hitting action consistency evaluation step S6 can be performed sequentially or simultaneously, but the hitting trajectory reconstruction step S3 The batting stage detection step S4 needs to be performed after the attitude estimation step S2 , and the batting trajectory reconstruction step S3 and the batting stage detection step S4 can be performed sequentially or simultaneously.
將本實施例的手持運動分析系統1及其分析方法實際應用於羽球球場進行運動員的擊球動作分析。其中,係將訊號感測模組11裝設於例如球拍握柄的後套內,用以感測運動員在揮拍擊球時的動作。如下表所示,其為9位羽球運動員進行長球揮拍時的揮拍分期、擊球速度、擊球弧度及一致性等指標的統計數值。
另外,圖5A及圖5B分別為應用本發明的手持運動系統進行分析時,兩位運動員的揮拍分期動作訊號示意圖,而圖6A及圖6B分別為對應於圖5A及圖5B之動作訊號的揮拍軌跡示意圖。在此,將揮拍訊號分期為初始靜止期、預備期、加速期、餘勢期及結束靜止期。In addition, FIGS. 5A and 5B are schematic diagrams of the swing motion signals of two athletes in stages when the handheld exercise system of the present invention is applied for analysis, respectively, and FIGS. 6A and 6B are respectively corresponding to the motion signals of FIGS. 5A and 5B . Schematic diagram of swing trajectory. Here, the swing signal is divided into an initial stationary period, a preparatory period, an acceleration period, an after-potential period and an end stationary period.
如圖5A所示,第一位運動員在長球揮拍的情況下的預備期、加速期及餘勢期時間分別為1.28秒、0.10秒及0.83秒;如圖5B所示,第二位運動員在長球揮拍的情況下的預備期、加速期及餘勢期時間分別為1.28秒、0.09秒及0.96秒。此外,第一位運動員長球揮拍的擊球速度為65.48kph、擊球力度為1.03N、揮拍弧度為253.24°,且揮拍動作一致性為93.10%;而第二位運動員長球揮拍的擊球速度為70.26kph、擊球力度為1.03N、揮拍弧度為244.76°,且揮拍動作一致性為94.33%,如下表所示。
此外,上述兩位運動員的長球揮拍軌跡可對應如圖6A及圖6B所示。需說明的是,由於本實施例之訊號感測模組11係設置於羽毛球拍之握柄的後套內,因此,圖6B中各時間點顯示的每一直線軌跡是代表羽毛球拍本身(到羽毛球拍之拍框的頂端)。In addition, the long-ball swing trajectories of the above two players may correspond to those shown in FIG. 6A and FIG. 6B . It should be noted that, since the
承上,由上述揭示內容可知,本發明的手持運動分析系統與分析方法,可以透過訊號感測模組自動擷取運動員進行擊球動作時的動作軌跡訊號(即感測訊號),經過伺服器進行訊號分析後可以得到多種相關的擊球指標,省去攝影機拍攝過程的麻煩、增加動作軌跡研究的便利性和實用性,同時可減低研究成本,並且不會受到不同人或不同地形、地物的影響而可將人為所產生的誤差降至最低。再者,運動員或/及教練可隨時藉由觀看分析系統的顯示模組所產生的相關擊球指標,以改進擊球動作。Continuing from the above, it can be seen from the above disclosure that the hand-held motion analysis system and analysis method of the present invention can automatically capture the motion trajectory signal (ie, the sensing signal) of the athlete's hitting action through the signal sensing module, and pass the server through the server. After signal analysis, a variety of relevant hitting indicators can be obtained, which saves the trouble of the camera shooting process, increases the convenience and practicability of movement trajectory research, and reduces research costs. The influence of human error can be minimized. Furthermore, the player or/and the coach can improve the batting action by viewing the relevant batting indicators generated by the display module of the analysis system at any time.
綜上所述,在本發明的手持運動分析系統與方法中,包括:姿態估測步驟,其依據感測訊號執行擊球動作的手持球具姿態估測;擊球軌跡重建步驟,其依據感測訊號和姿態估測步驟的結果執行擊球軌跡訊號的重建;擊球分期偵測步驟,其依據感測訊號和姿態估測步驟的結果區分擊球過程的不同時期;擊球球種辨識步驟,其依據感測訊號將擊球的球種類型進行分類;以及擊球動作一致性評估步驟,其依據感測訊號計算和評估擊球動作與樣板動作之間的一致性。藉此,相較於習知的擊球動作分析系統來說,本發明的手持運動分析系統與分析方法除了具有方便、專業且實用經濟的優點外,還可即時且客觀地提供相關的擊球指標給運動員或/及教練參考,進而改進運動員的擊球動作。To sum up, the hand-held motion analysis system and method of the present invention includes: an attitude estimation step, which performs attitude estimation of the hand-held golf equipment for hitting the ball according to the sensing signal; and a hitting trajectory reconstruction step, which is based on the sensing signal. The measurement signal and the result of the attitude estimation step perform reconstruction of the batting trajectory signal; the batting stage detection step, which distinguishes different periods of the batting process according to the sensing signal and the result of the attitude estimation step; the batting type identification step , which classifies the types of balls hit according to the sensing signal; and a batting action consistency evaluation step, which calculates and evaluates the consistency between the batting action and the sample action according to the sensing signal. Therefore, compared with the conventional batting action analysis system, the hand-held motion analysis system and analysis method of the present invention not only have the advantages of convenience, professionalism, practicality and economy, but also provide relevant batting action instantly and objectively. Indicators can be used as a reference for players and/or coaches to improve athlete's hitting action.
以上所述僅為舉例性,而非為限制性者。任何未脫離本發明之精神與範疇,而對其進行之等效修改或變更,均應包含於後附之申請專利範圍中。The above description is exemplary only, not limiting. Any equivalent modifications or changes that do not depart from the spirit and scope of the present invention shall be included in the appended patent application scope.
1:手持運動分析系統 11:訊號感測模組 12:伺服器 121:處理單元 122:記憶單元 1221:程式指令 13:顯示模組 S1,S11,S12,S2,S21,S22,S23,S3,S31,S32,S33,S34,S4,S41,S42,S43,S44,S5,S51,S52,S53,S54,S6,S61,S62,S63,S64,S7:步驟 SS:感測訊號 1: Handheld motion analysis system 11: Signal sensing module 12: Server 121: Processing unit 122: Memory Unit 1221: Program command 13: Display module S1, S11, S12, S2, S21, S22, S23, S3, S31, S32, S33, S34, S4, S41, S42, S43, S44, S5, S51, S52, S53, S54, S6, S61, S62, S63, S64, S7: Steps SS: sensing signal
圖1A為本發明一實施例之一種手持運動分析系統的功能方塊示意圖。 圖1B為圖1A之手持運動分析系統的伺服器的功能方塊圖。 圖2為本發明之手持運動分析方法的一流程步驟示意圖。 圖3為本發明之手持運動分析方法的另一流程步驟示意圖。 圖4為擊球動作的訊號分期偵測示意圖。 圖5A及圖5B分別為應用本發明的手持運動系統進行分析時,兩位運動員的揮拍分期動作訊號示意圖。 圖6A及圖6B分別為對應於圖5A及圖5B之動作訊號的揮拍軌跡示意圖。 FIG. 1A is a functional block diagram of a handheld motion analysis system according to an embodiment of the present invention. FIG. 1B is a functional block diagram of a server of the handheld motion analysis system of FIG. 1A . FIG. 2 is a schematic diagram of a process step of the handheld motion analysis method of the present invention. FIG. 3 is a schematic diagram of another process step of the handheld motion analysis method of the present invention. FIG. 4 is a schematic diagram of signal stage detection of a batting action. FIG. 5A and FIG. 5B are schematic diagrams of the swing motion signals of two athletes in stages when the handheld exercise system of the present invention is applied for analysis, respectively. 6A and 6B are schematic diagrams of swing trajectories corresponding to the motion signals of FIGS. 5A and 5B, respectively.
S1,S2,S3,S4,S5,S6:步驟 S1, S2, S3, S4, S5, S6: Steps
SS:感測訊號 SS: sensing signal
Claims (12)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110102787A TWI770787B (en) | 2021-01-26 | 2021-01-26 | Hand-held motion analysis sysyem and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110102787A TWI770787B (en) | 2021-01-26 | 2021-01-26 | Hand-held motion analysis sysyem and method |
Publications (2)
Publication Number | Publication Date |
---|---|
TWI770787B TWI770787B (en) | 2022-07-11 |
TW202228596A true TW202228596A (en) | 2022-08-01 |
Family
ID=83439280
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW110102787A TWI770787B (en) | 2021-01-26 | 2021-01-26 | Hand-held motion analysis sysyem and method |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI770787B (en) |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI541050B (en) * | 2014-06-17 | 2016-07-11 | 南臺科技大學 | Device and method for improving hitting posture |
CN111111121B (en) * | 2020-01-16 | 2023-10-03 | 合肥工业大学 | Racket and batting identification method |
-
2021
- 2021-01-26 TW TW110102787A patent/TWI770787B/en active
Also Published As
Publication number | Publication date |
---|---|
TWI770787B (en) | 2022-07-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10791968B2 (en) | Information processing device, sensor device, information processing system, and storage medium | |
US20150018111A1 (en) | Interpretation of characteristics of a golf swing using motion analysis | |
US7887440B2 (en) | Method for matching a golfer with a particular club style | |
US20130018494A1 (en) | System and method for motion analysis and feedback with ongoing dynamic training orientation determination | |
US7041014B2 (en) | Method for matching a golfer with a particular golf club style | |
CN110102035B (en) | Dynamic sampling in a sports device | |
US8506425B2 (en) | Method for matching a golfer with a particular golf club style | |
US11769424B2 (en) | Method, apparatus, and computer program product for measuring and interpreting metrics of an athletic action and an object associated therewith | |
CN105797319B (en) | A kind of badminton data processing method and device | |
TW201242645A (en) | Sensing device and method used for virtual golf simulation apparatus | |
US11850498B2 (en) | Kinematic analysis of user form | |
CN114788951B (en) | Handheld motion analysis system and method | |
KR100907704B1 (en) | Golfer's posture correction system using artificial caddy and golfer's posture correction method using it | |
Torigoe et al. | Strike Activity Detection and Recognition Using Inertial Measurement Unit towards Kendo Skill Improvement Support System. | |
Kos et al. | Tennis stroke consistency analysis using miniature wearable IMU | |
Ghosh et al. | Stancescorer: A data driven approach to score badminton player | |
KR101972526B1 (en) | golf swing correcting system | |
Peploe et al. | A curve fitting methodology to determine impact location, timing, and instantaneous post-impact ball velocity in cricket batting | |
TWI770787B (en) | Hand-held motion analysis sysyem and method | |
US10918920B2 (en) | Apparatus and methods to track movement of sports implements | |
CN107433030B (en) | Ball game training system, ball and intelligent motion tracking device | |
Kos et al. | Smart wearables for tennis game performance analysis | |
Mohd Hisam et al. | Automated Classification of Woodball Swinging Phases from Inertial Measurement Unit Using Least Square Method | |
Toshniwal et al. | Ai coach for badminton | |
CN110075494A (en) | A kind of sensor-based tennis ball speed estimating system |