本篇論文提出一個基於慣性測量單元的穿戴式系統，結合一系列方法用於估算步距。系統從安裝在行人的兩腿及鞋子上的三軸陀螺儀和三軸加速度計搜集角速度和加速度。藉由一個基於長短期記憶(LSTM)的神經網路，我們偵測腳步事件發生的時間點，接著使用一個由腿長及關節角度組成的步態模型，以及一個同樣基於長短期記憶神經網路的回歸方法計算步距。

實驗使用六個受試者以及三種不同走路速度的走路資料，用來檢驗提出的方法對於不同使用者及不同走路速度的兼容性。在偵測鞋跟踏地及鞋尖離地事件的實驗中，我們得到-0.015秒的平均誤差及0.046秒的標準差，證明長短期記憶可以有效的偵測腳步事件發生的時間點。在估算步距的實驗中，使用步態模型及回歸方法同樣得到0.22 - 0.3公分的平均誤差及3.8公分的標準差。實驗結果顯示提出的步態模型改進了步距估算的準確度，以及在回歸方法中，從步態模型取出特徵有助於長短期記憶神經網路學習到更準確的步距。

This thesis proposes a wearable system and a chain of methods for estimating stride lengths from inertial measurement units (IMU). The system collects inertial sensor data from several IMUs mounted on the legs and shoes of the walker, where each IMU provides data in terms of angular velocity from a triaxial gyroscope and acceleration from a triaxial accelerometer. The data are first processed by a Long Short-Term Memory (LSTM)-based method to determine the timing of step events. The raw IMU data and extracted features are also fed to LSTM to construct a regression model for learning stride lengths. A mechanical model that calculate stride lengths by the angles at joints and leg lengths is also proposed.

The experiments consist of a user-dependency test and a walking-speed dependency test. The results show that the proposed step event detector can detect heel-strike and toe-off events with -0.0008s to 0.015s mean errors and 0.015s to 0.046s precisions. The proposed stride-length estimator, whose performance is measured in terms of mean error ± precision, achieves -0.3 ± 3.8cm for the mechanical model and -0.22 ± 3.8cm for the LSTM model with extracted features. The results also show that using the features extracted from our mechanical model makes the LSTM model learn better compared to the LSTM model using raw IMU data.

Contents
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Related Work and Background Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Step Events Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Accelerometer-based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 Gyroscope-based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.3 Comparison Between Gyroscope and Accelerometer . . . . . . . . . . . . . . . . . . . . . 4
2.1.4 Machine-learning Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Stride Length Estimation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 IMU-based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Camera-based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.3 Other Kinds of Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.4 Machine-Learning Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Background Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Long Short-Term Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3 Technical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1 Step Event Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.1 Data Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.2 LSTM Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.3 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Stride Length Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.1 Mechanical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2 LSTM Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 System Architecture and Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1 Node Subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1.1 Sensor Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Host Subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2.1 Handling BLE Packet Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.1.1 Ultrasonic sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.1.2 Force-Sensitive Resistor (FSR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.2 Step Event Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.2.1 User Dependency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.2.2 Walking Speed Dependency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.2.3 Sensor Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2.4 Comparing with related works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 Stride Length Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3.1 Cross-subjects Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3.2 Walking Speed Dependency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.3.3 User-dependent Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3.4 Performance Comparison Between Mechanical Model and LSTM Model . . . . . . . . . . . . 33
5.3.5 Performance Comparison with Related Works . . . . . . . . . . . . . . . . . . . . . . . 34
6 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Bibliography
[1] Antoon Th. M. Willemsen, Fedde Bloemhof, and Herman B.K. Boom. Automatic stance-swing
phase detection from accelerometer data for peroneal nerve stimulation. IEEE Transactions on
Biomedical Engineering, 37(12):1201–1208, 1990.
[2] Avril Mansfield and Gerard M. Lyons. The use of accelerometry to detect heel contact events
for use as a sensor in FES assisted walking. Medical Engineering and Physics, 25(10):879–885,
2003.
[3] R.R. Torrealba, J. Cappelletto, L. Fermın-León, J.C. Grieco, and G. Fernandez-Lopez. Statisticsbased
technique for automated detection of gait events from accelerometer signals. Electronics
Letters, 46(22):1483–1485, 2010.
[4] Eric Watelain, Jérôme Froger, Franck Barbier, Ghislaine Lensel, Marc Rousseaux, François-
Xavier Lepoutre, and André Thevenon. Comparison of clinical gait analysis strategies by French
neurologists, physiatrists and physiotherapists. Journal of Rehabilitation Medicine, 35(1):8–14,
2003.
[5] Cristian F. Pasluosta, Heiko Gassner, JuergenWinkler, Jochen Klucken, and Bjoern M. Eskofier.
An emerging era in the management of Parkinson’s disease: wearable technologies and the
Internet of Things. IEEE Journal of Biomedical and Health Informatics, 19(6):1873–1881,
2015.
[6] Siddhartha Khandelwal and Nicholas Wickström. Identification of gait events using expert
knowledge and continuous wavelet transform analysis. In 7th International Conference on Bioinspired
Systems and Signal Processing (BIOSIGNALS 2014), Angers, France, March 3-6, 2014,
pages 197–204. SciTePress, 2014.
[7] Ruud W. Selles, Margriet A.G. Formanoy, Johannes B.J. Bussmann, Peter J. Janssens, and
Henk J. Stam. Automated estimation of initial and terminal contact timing using accelerometers;
development and validation in transtibial amputees and controls. IEEE Transactions on
Neural Systems and Rehabilitation Engineering, 13(1):81–88, 2005.
[8] Hafiz Farhan Maqbool, Muhammad Afif Bin Husman, Mohammed I. Awad, Alireza Abouhossein,
Nadeem Iqbal, and Abbas A. Dehghani-Sanij. A real-time gait event detection for lower
limb prosthesis control and evaluation. IEEE Transactions on Neural Systems and Rehabilitation
Engineering, 25(9):1500–1509, 2017.
[9] Darwin Gouwanda et al. A robust real-time gait event detection using wireless gyroscope and
its application on normal and altered gaits. Medical Engineering and Physics, 37(2):219–225,
2015.
[10] Jung Keun Lee and Edward J. Park. Quasi real-time gait event detection using shank-attached
gyroscopes. Medical & Biological Engineering & Computing, 49(6):707–712, 2011.
[11] Paola Catalfamo, Salim Ghoussayni, and David Ewins. Gait event detection on level ground and
incline walking using a rate gyroscope. Sensors, 10(6):5683–5702, 2010.
[12] Kaiyu Tong and Malcolm H. Granat. A practical gait analysis system using gyroscopes. Medical
Engineering and Physics, 21(2):87–94, 1999.
[13] Kamiar Aminian, B. Najafi, C. Büla, P.-F. Leyvraz, and Ph. Robert. Spatio-temporal parameters
of gait measured by an ambulatory system using miniature gyroscopes. Journal of Biomechanics,
35(5):689–699, 2002.
[14] XunSheng Ji, Shourong Wang, Yishen Xu, Qin Shi, and Dunzhu Xia. Application of the digital
signal procession in the MEMS gyroscope de-drift. In Proc. 1st IEEE International Conference
on Nano/Micro Engineered and Molecular Systems, 2006. NEMS’06, pages 218–221. IEEE,
2006.
[15] Li Qiang, Teng Jianfu, Wang Xin, Zhang Yaqi, and Quo JiChang. Research of gyro signal denoising
with stationary wavelets transform. In Proc. Canadian Conference on Electrical and
Computer Engineering, 2003. IEEE CCECE 2003., volume 3, pages 1989–1992. IEEE, 2003.
[16] Marcus Edel and Enrico Köppe. An advanced method for pedestrian dead reckoning using
BLSTM-RNNs. In Proc. 2015 International Conference on Indoor Positioning and Indoor
Navigation (IPIN), pages 1–6. IEEE, 2015.
[17] Arash Salarian, Pierre R. Burkhard, François J.G. Vingerhoets, Brigitte M. Jolles, and Kamiar
Aminian. A novel approach to reducing number of sensing units for wearable gait analysis
systems. IEEE Transactions on Biomedical Engineering, 60(1):72–77, 2013.
[18] Alberto Ferrari, Pieter Ginis, Michael Hardegger, Filippo Casamassima, Laura Rocchi, and
Lorenzo Chiari. A mobile Kalman-filter based solution for the real-time estimation of spatiotemporal
gait parameters. IEEE Transactions on Neural Systems and Rehabilitation Engineering,
24(7):764–773, 2016.
[19] John R. Rebula, Lauro V. Ojeda, Peter G. Adamczyk, and Arthur D. Kuo. Measurement of foot
placement and its variability with inertial sensors. Gait & posture, 38(4):974–980, 2013.
[20] Diana Trojaniello, Andrea Cereatti, Elisa Pelosin, Laura Avanzino, Anat Mirelman, Jeffrey M.
Hausdorff, and Ugo Della Croce. Estimation of step-by-step spatio-temporal parameters of
normal and impaired gait using shank-mounted magneto-inertial sensors: Application to elderly,
hemiparetic, Parkinsonian and choreic gait. Journal of Neuroengineering and Rehabilitation, 11
(1):152, 2014.
[21] Nicolas Saunier, Ali El Husseini, Karim Ismail, Catherine Morency, Jean-Michel Auberlet, and
Tarek Sayed. Estimation of frequency and length of pedestrian stride in urban environments
with video sensors. Transportation Research Record: Journal of the Transportation Research
Board, (2264):138–147, 2011.
[22] Chiraz BenAbdelkader, Ross Cutler, and Larry Davis. Stride and cadence as a biometric in
automatic person identification and verification. In Proc. Fifth IEEE International Conference
on Automatic Face and Gesture Recognition, pages 372–377. IEEE, 2002.
[23] Erik E. Stone and Marjorie Skubic. Passive in-home measurement of stride-to-stride gait variability
comparing vision and Kinect sensing. In Proc. 2011 Annual International Conference
of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 6491–6494. IEEE,
2011.
[24] Wei Chen, Ruizhi Chen, Xiang Chen, Xu Zhang, Yuwei Chen, JianyuWang, and Zhongqian Fu.
Comparison of EMG-based and accelerometer-based speed estimation methods in pedestrian
dead reckoning. The Journal of Navigation, 64(2):265–280, 2011.
[25] Julius Hannink, Thomas Kautz, Cristian F. Pasluosta, Jens Barth, Samuel Schülein, Karl-Günter
Gaßmann, Jochen Klucken, and Bjoern M. Eskofier. Stride length estimation with deep learning.
arXiv preprint arXiv:1609.03321, 2016.
[26] Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory. Neural Computation, 9
(8):1735–1780, 1997.
[27] Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint
arXiv:1412.6980, 2014.
[28] Sebastian Madgwick. An efficient orientation filter for inertial and inertial/magnetic sensor
arrays. Report x-io and University of Bristol (UK), 25:113–118, 2010.
[29] Robert Mahony, Tarek Hamel, and Jean-Michel Pflimlin. Nonlinear complementary filters on
the special orthogonal group. IEEE Transactions on automatic control, 53(5):1203–1218, 2008.