ppt2 - Soft Computing Lab.

advertisement
Dynamic Time Warping and Neural Network
J.-Y. Yang, J.-S. Wang and Y.-P. Chena,
Using acceleration measurements for
activity recognition: An effective learning
algorithm for constructing neural classifiers
Pattern Recognition Letters, vol. 29, no. 16, pp. 2213-2220, 2008.
Spring Semester, 2010
Outline
 Background
 Activity Recognition Strategy
 Experiments
 Summary
2
Background
 Accelerometers can be used as a human motion
detector and monitoring device
– Biomedical engineering, medical nursing, interactive
entertainment, …
– Exercise intensity / distance, sleep cycle, and calorie
consumption
3
Background
Proposed Method Overview
 One 3-D accelerometer on the dominant wrist
 NNs
– Pre-classifier  static classifier or dynamic classifier
 Eight domestic activities
– Standing, sitting, walking, running, vacuuming, scrubbing,
brushing teeth, and working at a computer
4
Background
Neural Classifier
 Neurons in the Brain
– A neuron receives input from other neurons (generally
thousands) from its synapses
– Inputs are approximately summed
– When the input exceeds a threshold the neuron sends an
electrical spike that travels from the body, down the axon, to
the next neuron(s)
5
Background
Neurons in the Brain (cont.)
 Amount of signal passing through a neuron
depends on:
– Intensity of signal from feeding neurons
– Their synaptic strengths
– Threshold of the receiving neuron
 Hebb rule (plays key part in learning)
– A synapse which repeatedly triggers the activation of a
postsynaptic neuron will grow in strength, others will
gradually weaken
– Learn by adjusting magnitudes of synapses’ strengths
6
Background
Artificial Neurons
y
f(A)
+1
0
A
g( )
-1
Step Function
∑w.x
f(A)
+1
w1
w2
0
x3
x1
x2
7
w3
A
-1
Sigmoid Function
Background
Neural Classifier (Perceptron)
 Structure
 Learning
– Weights are changed in proportion to the difference (error)
between target output and perceptron solution for each
example
– Back-propagation algorithm
• The gradient descent method, Slow convergence and local minima
– The resilient back-propagation (RPROP)
• Ignore the magnitude of the gradient
8
Activity Recognition Strategy
 Pre-Classifier
 Static/Dynamic Classifier
9
Activity Recognition Strategy
Pre-Classifier (1/2)
 Two components of the acceleration data
– Gravitational acceleration (GA)
– Body acceleration (BA): High-pass filtering to remove GA
 Segmentation with overlapping windows
– 512 samples per window
10
Activity Recognition Strategy
Pre-Classifier (2/2)
 SMA (Signal Magnitude Area)
– The sum of acceleration magnitude over three axes
 AE (Average Energy)
– Average of the energy over three axes
– Energy: The sum of the squared discrete FFT component
magnitudes of the signal in a window
11
Activity Recognition Strategy
Feature Extraction
 8 attributes × 3axis = 24 features
– Mean, correlation between axes, energy, interquartile range
(IQR), mean absolute deviation, root mean square, standard
deviation, variance
12
Activity Recognition Strategy
Feature Selection (1/2)
 Common principal
component analysis (CPCA)
 If features are highly
correlated,
the corresponding vectors
are similar
 clustering to group similar loadings
13
Activity Recognition Strategy
Feature Selection (2/2)
 Apply the PCA
 Select the first p PCs (cumulative sum>90%)
 Estimate CPC
 Support vector clustering
14
Activity Recognition Strategy
Verification
15
Experiments: Environment (1/2)
 MMA7260Q tri-axial accelerometer
– Sensitivity: -4.0g ~ +4.0g, 100Hz
– Mount on the dominant wrist
 Eight activities from seven subjects
– Standing, sitting, walking,
running, vacuuming,
scrubbing, brushing teeth,
and working at a computer
– 2min per activity
16
Experiments
Environment (2/2)
 Window size = 512 (with 256 overlapping)
– 22 windows in one min., 45 windows in two min.
 Leave-one-subject-out cross-validation
– Training: 1min per activity = 22 windows × 8 activities× 6
subjects
– Test: 2min per activity = 45 windows × 8 activities
17
Experiments
FSS Evaluation
 Use six static selected features
18
Experiments
Recognition Result
 NN
– Hidden node
• Pre-classifier: 3
• Static-classifier: 5
• Dynamic-classifier: 7
– Epochs: 500
 Computational load of FSS
– Training without FSS = 7.457s, training with FSS = 8.46s
19
Summary
 Proposed method yielded 95% accuracy
– Pre-classifier  static / dynamic classifiers
 Author’s other publication
–
–
20
Yen-Ping Chen, Jhun-Ying Yang, Shun-Nan Liou, Gwo-Yun Lee, Jeen-Shing Wang:
Online classifier construction algorithm for human activity
detection using a tri-axial accelerometer.
Applied Mathematics and Computation 205(2): 849-860 (2008)
Download