Abstract
OF THE THESIS Design and Development of Performance Metrics for Elite Runners By Nikhil R Mittal Master of Science in Electrical Engineering University of California, Los Angeles, 2012 Professor William J. Kaiser, Chair Recent advancements in mobile health applications of human motion sensing systems have enabled the proliferation of low-cost solutions in health and athletics assessment. As sensor data analysis and systems become more sophisticated the next innovation in this field is to provide user guidance towards improved outcomes. To this end, physiology research in distance running provides the framework to compute physics-based metrics to characterize running performance. This thesis presents a novel running gait cycle characterization system to monitor the race performance of elite runners based on several monotonic, computationally light metrics and an overall efficiency metric. The presented real time race results demonstrate the ability of motion sensing algorithms and systems to monitor, guide and enhance runner performance.
iii The thesis of Nikhil R Mittal is approved. Robert N. Candler Gregory J. Pottie William J. Kaiser, Committee Chair University of California, Los Angeles 2012
iv TABLE OF CONTENTS 1 INTRODUCTION ........................................................................................................................................... 1 1.1 Overview ................................................................................................................................................. 1 1.2 Motivation and Background ............................................................................................................ 2 1.3 Thesis Contribution ............................................................................................................................ 4 2 SENSOR SYSTEMS ....................................................................................................................................... 7 2.1 Basic Sensor Principles ..................................................................................................................... 7 2.2 Accelerometers ..................................................................................................................................... 8 2.3 Data Aggregation & Analysis ........................................................................................................ 10 2.3.1 Sensor Node ............................................................................................................................... 12 2.3.2 Online Data Analysis Node ................................................................................................... 13 2.3.3 Offline Data Analysis Network ............................................................................................ 15 2.4 Prototype Running Performance System: MDAWN Datalogger ..................................... 16 2.4.1 Specification ............................................................................................................................... 16 2.4.2 Data Management .................................................................................................................... 19 2.4.3 Sensor Placement..................................................................................................................... 20 2.4.4 Sensing System Orientations ............................................................................................... 21 3 PERFORMANCE METRICS ..................................................................................................................... 22 3.1 Metric characteristics ..................................................................................................................... 22
v 3.2 Metric description ............................................................................................................................ 24 3.2.1 Cadence ........................................................................................................................................ 25 3.2.2 Ground Contact Time (GCT) ................................................................................................ 25 3.2.3 Distal Leg Lift (DLL) ................................................................................................................ 31 3.2.4 Over-stride Metric ................................................................................................................... 34 3.2.5 Contact De-acceleration Energy Loss (CDEL) Metric ................................................ 35 3.2.6 Pronation .................................................................................................................................... 37 4 ALGORITHM ARCHITECTURE AND STRIDE SEGMENTATION ............................................... 41 4.1 Pre-processing of raw data ........................................................................................................... 42 4.1.1 Stride segmentation/recognition ...................................................................................... 42 4.1.2 Activity Detection .................................................................................................................... 45 4.2 Real time metric computation ..................................................................................................... 46 4.3 Post processing of data ................................................................................................................... 46 5 REAL TIME DEPLOYMENT AND RACE RESULTS ......................................................................... 48 5.1 Real Time Deployment ................................................................................................................... 49 5.2 Race Results ........................................................................................................................................ 50 5.3 Los Angeles Rock & Roll Half Marathon .................................................................................. 51 5.3.1 Final Race Results .................................................................................................................... 52 5.3.2 Cadence ........................................................................................................................................ 52
vi 5.3.3 Ground Contact Time (GCT) ................................................................................................ 54 5.3.4 Distal Leg Lift (DLL) ................................................................................................................ 57 5.3.5 Over-stride Metric ................................................................................................................... 59 5.3.6 Contact De-acceleration Energy Loss (CDEL) ............................................................... 62 5.3.7 Overall Efficiency Metric ....................................................................................................... 64 5.4 Honolulu Marathon .......................................................................................................................... 65 5.4.1 Final Race Results .................................................................................................................... 66 5.4.2 Cadence ........................................................................................................................................ 68 5.4.3 Ground Contact Time (GCT) ................................................................................................ 69 5.4.4 Distal Leg Lift (DLL) ................................................................................................................ 72 5.4.5 Over-stride Metric ................................................................................................................... 74 5.4.6 Contact De-acceleration Energy Loss (CDEL) ............................................................... 76 5.4.7 Overall Efficiency Metric ....................................................................................................... 78 5.5 Other Important Results ................................................................................................................ 79 5.5.1 Angela Bizzari ............................................................................................................................ 80 5.5.2 Morgan Uceny ........................................................................................................................... 81 5.5.3 Anna Pierce ................................................................................................................................ 83 5.5.4 Key Observations ..................................................................................................................... 84 6 CONLUSIONS AND FUTURE WORK ................................................................................................... 86
vii Bibliography ......................................................................................................................................................... 89
viii LIST OF FIGURES Figure 2.1: High-level sensor diagram 7 Figure 2.2: Example Distributed wireless sensing system architecture 11 Figure 2.3: High-level hardware architecture of a sensor node 12 Figure 2.4: MDAWN Data-logger prototype 17 Figure 2.5: Sparkfun Electronics Logomatic v2 17 Figure 2.6: MDAWN Data uploader software interface 19 Figure 2.7: MDAWN sensor placement 21 Figure 2.8: Sensor system orientation 21 Figure 3.1: Typical human running gate cycle 24 Figure 3.2: Ground Contact phase of a rear foot striker (RFS) 27 Figure 3.3: Typical z-axis data 28 Figure 3.4: Graph highlighting the monotonic nature of average GCT_F values for the five cases mentioned in Table 3.1 30 Figure 3.5: DLL metric vs. the increasing degree of leg lift 33 Figure 3.6: Over-stride metric vs. increasing degree of stride length 35 Figure 3.7: System Acceleration in Abduction Direction for Left Shoe in Tempo Run 39 Figure 3.8: Combined Left and Right Foot Pronation Rotation at different stages of the run 40 Figure 4.1: Work flow for algorithm design 41 Figure 4.2: Raw acceleration data from the tri-axial accelerometer 42
ix Figure 4.3: Data vector generated for stride segmentation purposes highlighting the margin of error 43 Figure 5.1: A screenshot of the real time display 50 Figure 5.2: Course Elevation chart for the LA R&R Half Marathon 51 Figure 5.3: Final race timings and comparison of the three runners 52 Figure 5.4: Distance vs. Cadence time series for the three runners 53 Figure 5.5: GCT_S of right feet vs. Distance for the three runners 54 Figure 5.6: GCT_F of right feet vs. Distance for the three runners 55 Figure 5.7: GCT_S of left feet vs. Distance for the three runners 55 Figure 5.8: GCT_F of left feet vs. Distance for the three runners 56 Figure 1.9: Distal Leg Lift metric with distance for Josh for both feet 57 Figure 5.10: Distal Leg Lift Metric with distance for John for both feet 58 Figure 5.11: Distal Leg Lift metric with distance for Brian for both feet 58 Figure 5.12: Over-stride metric vs. Distance for Josh 60 Figure 5.13: Over-stride metric vs. Distance for John 60 Figure 5.14: Over-stride metric vs. Distance for Brian 61 Figure 5.15: CDEL vs. Distance for Josh 62 Figure 5.16: CDEL vs. Distance for John 63 Figure 5.17: CDEL vs. Distance for Brian 63 Figure 5.18: Overall Efficiency metric for the three runners 65 Figure 5.19: Course Elevation chart for the Honolulu Marathon 66 Figure 5.20: Final race timings and comparison of the three runners 67 Figure 5.21: Split timing comparison for the three runners 68
x Figure 5.22: Distance vs. Cadence time series for the three runners 68 Figure 5.23: GCT_S of right feet vs. Distance for the three runners 70 Figure 5.24: GCT_F of right feet vs. Distance for the three runners 70 Figure 5.25: GCT_S of left feet vs. Distance for the three runners 71 Figure 5.26: GCT_F of left feet vs. Distance for the three runners 71 Figure 5.27: Distal Leg Lift metric with distance for Josphat for both feet 73 Figure 5.28: Distal Leg Lift metric with distance for Jimmy for both feet 73 Figure 5.29: Distal Leg Lift metric with distance for Patrick for both feet 74 Figure 5.30: Over-stride metric vs. Distance for Josphat 75 Figure 5.31: Over-stride metric vs. Distance for Jimmy 75 Figure 5.32: Over-stride metric vs. Distance for Patrick 76 Figure 5.33: CDEL metric vs. Distance for Josphat 77 Figure 5.34: CDEL metric vs. Distance for Jimmy 77 Figure 5.35: CDEL metric vs. Distance for Patrick 78 Figure 5.36: Overall Efficiency metric with Distance for the three runners 79 Figure 5.37: Progression of GCT_S for the six 1K runs by Angela 80 Figure 5.38: Progression of DLL metric for the six 1K runs by Angela 81 Figure 5.39: Progression of GCT_S for the different runs by Morgan 82 Figure 5.40: Progression of DLL metric for the different runs by Morgan 82 Figure 5.41: Progression of GCT_S for the different runs by Anna 83 Figure 5.42: Progression of DLL metric for the different runs by Anna 84
xi LIST OF TABLES Table 2.1: Additional sensor figures of merit 8 Table 2.2: Figures of merit for Analog Devices ADXL326 accelerometer [17] 10 Table 2.3: Typical sensor node system components 13 Table 2.4: Motorola droid bionic specifications [19] 15 Table 2.5: lPC2148 main specifications [20] 17 Table 3.1: GCT_F values for runs with different technical flaws 29 Table 3.2: DLL metric values for different types of runs for the 3 runners 33 Table 3.3: Average CDEL metric values for 5 normal and 5 heavy footed runs 37
Description
A thesis on performance metrics for elite runners using motion sensing technology.