Human Activity Recognition Using Smartphone Sensor Data ECE

  • Slides: 7
Download presentation
Human Activity Recognition Using Smartphone Sensor Data ECE 539 Final Project Marc Petit 12/13/2017

Human Activity Recognition Using Smartphone Sensor Data ECE 539 Final Project Marc Petit 12/13/2017 ECE 539 Final Project 1

Objective • Activity tracker for cellphone applications (X, Y, Z) – Acceleration (Accelerometer) (X,

Objective • Activity tracker for cellphone applications (X, Y, Z) – Acceleration (Accelerometer) (X, Y, Z) – Angular Velocity (Gyroscope) Walking, Walking Upstairs, Walking Downstairs, Sitting, Standing, Laying + Transitions • Dataset from UCI Machine Learning Repository • 30 volunteers (19 – 48 years old) • Samsung Galaxy SII (accelerometer and gyroscope, 50 Hz sampling rate) • 61 experiments (~20 activities each) 1214 samples • Activities where extracted using video data • Using 128 reading window (2. 56 sec), a 561 feature vector was created (for each activity) 12/13/2017 ECE 539 Final Project 2

Approach • Use given feature vector 561 and various classifier Features (X, Y, Z)

Approach • Use given feature vector 561 and various classifier Features (X, Y, Z) Mean (X, Y, Z) Std (X, Y, Z) Max (X, Y, Z) Energy (X, Y, Z) max Freq (X, Y, Z) AR Coef. …. Classifier Linear Discr. KNN SVM • Use CNN to get features from raw data • Three Way Cross-Valid. (Prob. Of Misclass. /False Alarm) • Software: MATLAB (NN Toolbox, Stat. and ML Toolbox) 3/2/2021 ECE 539 Final Project 3

Sensor Data • Example Raw Data • Feature Space (using t-Dist. Stochastic Neighbor Embedding)

Sensor Data • Example Raw Data • Feature Space (using t-Dist. Stochastic Neighbor Embedding) Walking Upstairs Walking Downstairs Sitting Standing Laying 12/13/2017 ECE 539 Final Project Stand to Sit to Stand Sit to Lie to Sit Stand to Lie to Stand 4

Classification *) incl. transitions Prob. of Misclassification 12/13/2017 Prob. of False Alarm ECE 539

Classification *) incl. transitions Prob. of Misclassification 12/13/2017 Prob. of False Alarm ECE 539 Final Project 5

CNN for Raw Data Classification • CNN Structure [Ming Zeng, 2014] Image Input Layer

CNN for Raw Data Classification • CNN Structure [Ming Zeng, 2014] Image Input Layer Window Size x 2 x 3 Convolution 2 D Layer Conv. Size #Filter Max Pooling 2 D Layer Pool Size Fully Connected NN Softmax + Classification Layer • Optimization using MATLAB Genetic Algorithm Function Window Size (200), Conv. Size(23), #Filter (97), Pool Size (46) • Evaluation and Comparison Prob. of Misclassification 12/13/2017 Prob. of False Alarm ECE 539 Final Project 6

Discussion • Classifiers are more reliable than the CNN • (Linear/ 2 nd Order)

Discussion • Classifiers are more reliable than the CNN • (Linear/ 2 nd Order) SVM seem to work best for this problem (consistent with literature) • CNN does feature extraction automatically but needs more training data. • Classifiers and CNN were particularly good at identifying walking or laying, while sitting and standing was hard to distinguish. 12/13/2017 ECE 539 Final Project 7