1 of 33

WEEK 11

2 of 33

What we are going to cover this week

  • MyRuns5
  • Sensor Manager
  • Activity Recognition and machine learning

3 of 33

Sensors

4 of 33

5 of 33

6 of 33

iPhone X

7 of 33

8 of 33

let’s look at some signals again

9 of 33

10 of 33

11 of 33

12 of 33

13 of 33

14 of 33

15 of 33

16 of 33

17 of 33

18 of 33

activity recognition/ classification pipeline

Interpret

Observe

Features

Classifier

Actions

19 of 33

we have to “train” the classifier

20 of 33

we will use FFTs

21 of 33

1

1

1

1

1

1

1

1

1

1

1

1

22 of 33

what “classifier” should we use?

23 of 33

decision trees are simple and effective

24 of 33

workflow

  • collect training data for all classes (sitting, standing, etc) using the “collector app”
  • the collector uses FFT (fast fourier transforms) of the x, y, z accelerometer data and puts those features into a file called “features.arrf” with labels

25 of 33

workflow

​

  • use WEKA to input the training data file and train a decision tree
  • use that java class (it’s a decision tree classifier— if then statements) to classify behaviors in real-time by adding the class to your assignment

26 of 33

The pileline

27 of 33

28 of 33

features.arrf snippet

29 of 33

30 of 33

31 of 33

32 of 33

drop the class into

you android app

“pipeline:

33 of 33

~~~Break ~~~