390 likes | 739 Views
ADSP - Oral presentation 3D Accelerometer. Presenter : Chen Yu R0094049. Outline. Introduction 3D Accelerometer Applications about 3D accelerometers A Real-Time Human Movement Classifier Analysis of Acceleration Signals using Wavelet Transform Activity Recognition Conclusion
E N D
ADSP - Oral presentation3D Accelerometer Presenter : Chen Yu R0094049
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Introduction • Accelerometer is a device which can detect and measure acceleration.
Introduction • By measuring the vertical value of gravity, we can acquire the tilt angle of the accelerometer. the G value derived from the angle.
Introduction • There are a lot of types of accelerometers • Capacitive • Piezoelectric • Piezoresistive • Hall Effect • Magnetoresistive • Heat Transfer
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
3D Accelerometer • Basic Principle of Acceleration • Velocity is speed and direction so any time there is a change in either speed or direction there is acceleration. • Earth’s gravity: 1g • Bumps in road: 2g • Space shuttle: 10g • Death or serious injury: 50g
3D Accelerometer • Basic Accelerometer • Newton’s law • Hooke’s law • F = kΔx = ma
3D Accelerometer • Piezoelectric Systems
3D Accelerometer • Electromechanical Systems
3D Accelerometer • Tilt angle
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Applications about 3D accelerometers • Calculate the user’s walking state • Analyze the lameness of cattle • Detect walking activity in cardiac rehabilitation • Examine the gesture for cell phone or remote controller for video games
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
A Real-Time Human Movement Classifier • Human body’s movements are within frequency below 20 Hz (99% of the energy is contained below 15 Hz) • Median filter • remove any abnormal noise spikes • Low pass filter • Gravity • bodily motion
A Real-Time Human Movement Classifier Walk Upstair Downstair
A Real-Time Human Movement Classifier • Activity and Rest • Appropriate threshold value • Above the threshold -> active • Below the threshold -> rest
A Real-Time Human Movement Classifier • We define the Φ, which is the tilt angle between the positive z-axis and the gravitational vector g. • we can determine that a tilt angle between 20 and 60is sitting, and angles of 0 to 20 standing, and the angle between 60 and 90 is lying.
A Real-Time Human Movement Classifier • When the patient is lying down, their orientation is divided into the categories of right side (right), left side (left), lying face down (front), or lying on their back (back)
A Real-Time Human Movement Classifier • Feature Generation • Average: Average acceleration (for each axis) • Standard Deviation: Standard deviation (for each axis) • Average Absolute Difference: Average absolute difference between the value of each of the data within the ED and the mean value over those values (for each axis) • Average Resultant Acceleration: Average of the square roots of the sum of the values of each axis squared over the ED
A Real-Time Human Movement Classifier • Time Between Peaks: Time in milliseconds between peaks in the sinusoidal waves associated with most activities (for each axis) • Binned Distribution: We determine the range of values for each axis (maximum – minimum), divide this range into 10 equal sized bins, and then record what fraction of the 200 values fell within each of the bins.
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Analysis of Acceleration Signals using Wavelet Transform • Wavelet Transform g[n] 2 xLL[n] g[n] 2 xL[n] h[n] 2 xLH[n] x[n] xHL[n] g[n] 2 h[n] 2 xH[n] 2 xHH[n] h[n]
Analysis of Acceleration Signals using Wavelet Transform • the original signal x[n] can also be expanded by the mother wavelet function and the scaling function.
Analysis of Acceleration Signals using Wavelet Transform • Preprocessing : • Windowing • The acceleration signals are accessed in real time in the system. Therefore, the system must cut a sequence of data into consecutive windows before data analysis. • Feature Selection • The advantage of the WT is that the wavelet coefficients imply the details in different bands.
Analysis of Acceleration Signals using Wavelet Transform • Power of maximum signal: • Mean: • Variance: • Energy: • The energy of neighbor difference:
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Activity Recognition • There are several machine learning algorithms that can be used for classification, • Gaussian mixture model (GMM) • decision tree (J48) • logistic regression
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Conclusion Time analysis use decision tree Time analysis use logistic regression
Conclusion The Wavelet transform use decision tree The Wavelet transform use logistic regression
Outline • Introduction • 3D Accelerometer • Applications about 3D accelerometers • A Real-Time Human Movement Classifier • Analysis of Acceleration Signals using Wavelet Transform • Activity Recognition • Conclusion • Reference
Reference • P. Barralon, N. Vuillerme and N. Noury, “Walk Detection With a Kinematic Sensor: Frequency and Wavelet Comparison,” IEEE EMBS Annual International Conference New York City, USA, Aug 30-Sept 3, 2006 • M. Sekine, T. Tamura, M. Akay, T. Togawa, Y. Fukui, “Analysis of Acceleration Signals using Wavelet Transform,” Methods of Information in Medicine, F. K. Schattauer Vrlagsgesellschaft mbH (2000) • Elsa Garcia, Hang Ding and Antti Sarela, “Can a mobile phone be used as a pedometer in an outpatient cardiac rehabilitation program?,” IEEE/ICME International Conference on Complex Medical Engineering July 13-15,2010, Gold Coast, Australia
Reference • NiranjanBidargaddi, AnttiSarela, LasseKlingbeil and MohanrajKarunanithi, “Detecting walking activity in cardiac rehabilitation by using accelerometer,” • Masaki Sekine, Toshiyo Tamura, MetinAkay, Toshiro Fujimoto, Tatsuo Togawa, and Yasuhiro Fukui, “Discrimination of Walking Patterns Using Wavelet-Based Fractal Analysis,” IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 10, NO. 3, SEPTEMBER 2002 • “ Accelerometers and How they Work ” • “ Basic Principles of Operation and Applications of the Accelerometer ” Paschal Meehan and Keith Moloney - Limerick Institute of Technology.
Reference • From the lecture slide of “ Time Frequency Analysis and Wavelet Transform” by Jian-Jiun Ding • Jennifer R. Kwapisz, Gary M. Weiss, Samuel A. Moore “Activity Recognition using Cell Phone Accelerometers” • Jian-Hua Wang, Jian-Jiun Ding, Yu Chen “Automatic Gait recognition based on wavelet transform by using mobile phone accelerometer”