由於頸部姿勢不良進而產生肩頸疾病已是現代生活中相當常見的情況，隨著體感應用與穿戴式裝置的蓬勃發展，目前在市面上已出現將慣性感測器穿戴於後頸部來達成頸部姿勢提醒及頸部體感互動功能的頸部穿戴式產品，但屬於其關鍵技術的頸部動作慣性感測訊號辨識之相關研究至今仍然尚為缺乏。
本論文基於如同上述產品將慣性感測器穿戴於後頸部的條件下，提出一種針對慣性感測器之六軸感測訊號進行辨識處理的即時頸部動作辨識方法。首先對感測訊號進行位準與振幅範圍調整及濾除雜訊之資料前處理，然後偵測頸部動作的始末以擷取頸部動作訊號，再計算頸部動作訊號與頸部動作訊號模型之間的距離以度量其訊號相似性，並將此距離作為頸部動作訊號的分類依據。頸部動作訊號模型是由使用者在系統初始化階段將本論文所定義之18種頸部動作皆進行一次來建立。
為了實現上述方法，本論文整合嵌入式系統開發板、慣性感測器模組及藍牙模組等硬體單元，實作出頸部穿戴式裝置以進行感測訊號的讀取、資料前處理及無線傳輸，亦實作Android應用程式來進行頸部動作訊號的擷取與辨識運算。而在驗證的部分則是經由10位受測者針對本論文所定義之18種頸部動作進行實驗，實驗結果之整體正確辨識率為94%。

Poor neck posture causes neck pain symptoms is common situation in modern life, with the vigorous development of somatosensory applications and wearable devices, neck wearable device product that provides neck posture remind and somatosensory function by wearing IMU(Inertial Measurement Unit) on the back of neck has appeared on the market, but the related research about the key technology of neck motion inertial sensing signal recognition are lacking.
Base on the condition that wear the IMU on the back of neck as the same as the product described above, this thesis proposes a method for real time neck motion recognition by IMU six-axis signal processing. The first step of this method is data preprocessing that perform signal level adjustment, signal amplitude range adjustment and noise filtering. Secondly, this method detect the start and end of neck motion to capture neck motion signals, then computing the distance between neck motion signals and neck motion signal models to measure the signal similarity, and take it as the classification basis of neck motion signals. Users need to perform the 18 types of neck motion as defined in this thesis by once in the system initialization phase to build neck motion signal models.
In order to implement the method described above, this thesis integrates embedded system development board, IMU module and Bluetooth module to implement neck wearable device for reading sensing signals, data preprocessing and wireless transmission, this thesis also integrates several program code to implement Android application program for neck motion signals capturing and recognition computing. About verification, this thesis experimented with 10 subjects perform the 18 types of neck motion as defined in this thesis, the total recognition accuracy of experiment results is 94%.

[1] 財團法人全民健康基金會, “低頭族SOS 小心頸椎病變找上你！,” http://www.twhealth.org.tw/index.php?option=com_zoo&task=item&item_id=592&Itemid=20, Retrieved on April 2015.
[2] Parimita Saikia and Karen Das, “Head Gesture Recognition using Optical Flow based Classification with Reinforcement of GMM based Background Subtraction,” International Journal of Computer Applications (0975–8887), Vol.65, No.25, March 2013.
[3] Ligeng Dong , Yuxin Jin, Linmi Tao and Guangyou Xu, “Recognition of Multi-Pose Head Postures in Human Conversations,” Fourth International Conference on Image and Graphics(ICIG), pp.650–654, August 2007.
[4] George Galanakis, Pavlos Katsifarakis, Xenophon Zabulis and Ilia Adami, “Recognition of Simple Head Postures Based on Head Pose EstimationAnalysis,” AMBIENT 2014 : The Fourth International Conference on AmbientComputing, Applications, Services and Technologies, pp.88–96, August2014.
[5] Ke-Yu Chen, Aniket Handa, Chaoyu Yang, Shuowei Li and Shwetak Patel, “NecX: Enabling Posture Control Using the Neck,” http://www.keyuc.com/research/NecX/NecX_2014.pdf, Retrieved on April 2015.
[6] Marco Hirsch, Jingyuan Cheng, Attila Reiss, Mathias Sundholm, Paul Lukowicz and Oliver Amft, “Hands-Free Posture Control with a Capacitive Textile Neckband,” ISWC '14 Proceedings of the 2014 ACM International Symposium on Wearable Computers, pp.55–58, September 2014.
[7] VEARI, “Fineck,” http://www.fineck.com/index/chn.html, Retrieved on April 2015.
[8] ANALOG DEVICES, “ADXL345 Data Sheet(3-Axis, ±2g/±4g/±8g/±16g Digital Accelerometer),” http://www.analog.com/media/en/technical-documentation/data-sheets/ADXL345.pdf, Retrieved on April 2015.
[9] ANALOG DEVICES, “AN-1077 Application Note(ADXL345 Quick Start Guideby Tomoaki Tusuzki),” http://www.analog.com/media/en/technical-documentation/application-notes/AN-1077.pdf, Retrieved on April 2015.
[10] STMicroelectronics, “L3G4200D(MEMS motion sensor:ultra-stable three-axis digital output gyroscope),” http://www.st.com/web/en/resource/technical/document/datasheet/CD00265057.pdf, Retrieved on April 2015.
[11] STMicroelectronics, “AN4504 Application Note(L3G4200D: 3-axis digital output gyroscope),” http://www.st.com/web/en/resource/technical/document/application_note/DM00118758.pdf, Retrieved on June 2015.
[12] Arduino, “Arduino Fio,” https://www.arduino.cc/en/Main/ArduinoBoardFio, Retrieved on April 2015.
[13] DFROBOT, “DFRduino FIO (Arduino Compatible),” http://www.dfrobot.com/index.php?route=product/product&product_id=111#.Vb-1yfmqqko, Retrieved on July 2015.
[14] Jens Chr Brynildsen (jenschr on Github), jenschr/Arduino-libraries/ADXL345/, https://github.com/jenschr/Arduino-libraries/tree/master/ADXL345, Retrieved on April 2015.
[15] Kevin(Pololu)(kevin-pololu on Github), pololu/l3g4200d-arduino/L3G4200D/, https://github.com/pololu/l3g4200d-arduino/tree/master/L3G4200D, Retrieved on April 2015.
[16] Emartee, “CuteDigi BMX Bluetooth to UART/I2C/USB Module (GEN II),” http://www.dfrobot.com/image/data/TEL0026/TEL0026_Datasheet.pdf, Retrieved on April 2015.
[17] Colin Mercer, “Data Smoothing: RC Filtering And Exponential Averaging,” http://blog.prosig.com/2003/04/28/data-smoothing-rc-filtering-and-exponential-averaging/, Retrieved on June 2015.
[18] Wikipedia, “Low-pass filter,” https://en.wikipedia.org/wiki/Low-pass_filter, Retrieved on May 2015.
[19] Gavin Kistner, “Frame Rate-Independent Low-Pass Filter,” http://phrogz.net/js/framerate-independent-low-pass-filter.html, Retrieved on May 2015.
[20] “Low-pass filter,” http://helpful.knobs-dials.com/index.php/Low-pass_filter, Retrieved on June 2015.
[21] “一次指数平滑法,” http://wiki.mbalib.com/wiki/一次指数平滑法, Retrieved on June 2015.
[22] XantaCross (on Wikimedia Commons), “File:Euclidean vs DTW.jpg,” https://commons.wikimedia.org/wiki/File:Euclidean_vs_DTW.jpg, Retrieved on May 2015.
[23] T. K. Vintsyuk, “Speech Discrimination by Dynamic Programming,” Kibernetika, Vol.4, No.1, pp.81–88, 1968.
[24] Hiroaki Sakoe and Seibi Chiba, “Dynamic Programming Algorithm Optimization for Spoken Word Recognition,” IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol.ASSP-26, No.1, February 1978.
[25] David Clifford and Glenn Stone, “Variable Penalty Dynamic Time Warping Code for Aligning Mass Spectrometry Chromatograms in R,” Journal of Statistical Software, Vol.47, Issue. 8, April 2012.
[26] Jyh-Shing Roger Jang(張智星), “8-4 Dynamic Time Warping,” http://mirlab.org/jang/books/dcpr/dpDtw.asp?title=8-4+Dynamic+Time+Warping&language=english, Retrieved on April 2015.
[27] Wikipedia, “Dynamic time warping,” https://en.wikipedia.org/wiki/Dynamic_time_warping, Retrieved on May 2015.
[28] Cheol-Woo Jung, “source:GART/weka/edu/gatech/gart/ml/weka/DTW.java@3,” http://trac.research.cc.gatech.edu/GART/browser/GART/weka/edu/gatech/gart/ml/weka/DTW.java?rev=9, Retrieved on May 2015.
[29] 智慧生活科技專業社群, “Android Bluetooth 應用之HelloBTUart(RS-232),” http://cheng-min-i-taiwan.blogspot.tw/2012/01/android-bluetooth-hellobtuartrs-232.html, Retrieved on June 2015.
[30] Hung-Yu Chang, Hung-Ta Pei and Li-Shi Ju, “APK Bluetooth Service User Manual,” National Cheng Kung University Department of Engineering Science Multimedia Network Laboratory, Unpublished Manual, pp.12.
[31] T. M. Cover and P. E. Hart, “Nearest Neighbor Pattern Classification,” IEEE Transactions on Information Theory, Vol.IT-13, No.1, pp.21–27, January 1967.
[32] 陳士杰, 國立聯合大學資訊管理學系機器學習課程, “最近鄰居分類(Nearest-Neighbor Classification),” http://sjchen.im.nuu.edu.tw/MachineLearning/final/CLS_Nearest-neighbor.pdf, Retrieved on April 2015.