加速度感測器辨識運動型態之研究

Movement-type Classification by Using Acceleration

柳君翰(Jyun-Han Liou)

指導教授 ： 李朱慧

朝陽科技大學/資訊學院/資訊管理系/碩士(2014年)

摘要

台灣的路跑與慢跑活動近年來越來越盛行，許多民眾透過社交媒體接觸這個運動，儘管跑步本來就是生活上的一種移動方式，但是剛接觸跑步類型的民眾在活動過程中不容易讓自己保持在活動的狀態，容易在身體即將面臨負荷的狀態下開始交錯運動，萬一速度漸慢甚至停止會感到運動過程更加吃力，在適應與訓練的過程中除了要調適心情，漸進式的訓練也是重要的。
然而要透過訓練增強運動效果必須先能讓過程被量化呈現，可是，先前的研究較著重發覺單一運動的特徵，而且鮮少有研究透過自由空間的運動過程蒐集到的動能資料進行觀察，因此，本研究希望能在受測者自由移動的狀態下，提出可以分割蒐集到的資料的分類方法，本論文採用加速度感應器記錄運動過程中產生的動能資料。除了走路運動可能會參雜在初學者的跑步運動過程，本研究也針對民眾平時可能會移動的狀態(上、下樓梯)進行運動類型的分類，提供一個可以分割複雜運動過程的部分貢獻。蒐集資料後的處理過程我們將藉由局部最大值、傅立葉分析作為擷取特徵的依據，再配合適當的分類方法完成整篇研究。

關鍵字

快速傅立葉轉換 ； 支持向量機 ； 局部最大值 ； 運動分類 ； 加速度感應器

並列摘要

Road running is getting popular in Taiwan, most of people contact this activity through their social circle. In moving procedure, rookies and trainee are hard to keep their limbs in a stable and regular oscillation when they feel weary. It is a big challenge for them even walking and running are the human being’s innate ability. In the case of they are easily interaction their feet-steps and the movement will stop or getting slower, that make them feel more strenuous. Therefore, there are not only keep a suitable mood is training task but also do the progressive movement training.
However, the moving procedure should be quantifiable and presentable that can improve any particular in the motion. In previous study, there are few research aim at observe the feature in different movement-type, but we believe to extract the feature which can distinguish different movement-type is a key to achieve the objective. Therefore, this research aims at: 1) to find any practicable features, 2) to suit the features to the classification by the accelerometer. This research propose two way to classify movement-type, one is just to judge walking and running, the other can also distinguish go up and down stairs movement-type. This research raises a way to be a basement of classify a complex movement procedure. And the feature is extract by counting local maximum and Fourier analysis.

並列關鍵字

accelerometer ； motion classification ； fast fourier transform ； support vector machine ； local maximum

參考文獻

[3] J. Cooley, and J. Tukey(1965), “An algorithm for the machine calculation of complex Fourier series,” mathematics of computation, Vol. 19, No. 90, pp. 297-301.

[4] C. Crisci, B. Chattas, and G. Perera(2010), “A review of supervised machine learning algorithms and their applications to ecological data,” Ecological Modelling, Vol. 240, pp. 113-122.

[6] P. Duhamel, and M. Vetterli(1990), “Fast Fourier Transforms: A Tutorial Review and a State of the Art,” Signal Processing, Vol. 19, Issue 4, pp. 259-299.

[7] M. Gazendam, and A. Hof(2007), “Averaged EMG profiles in jogging and running at different speeds,” Gait & Posture, Vol. 25, Issue4, pp. 604-614.

[8] L. Guidettil, G. Rivellini, and F. Figura(1996),“EMG Patterns During Running: Intra- and Inter-individualVariability,” Journal of Electromyography and Kinesiology,Vol. 1, No. 6, pp. 37-48.