Portfolio

Follow our weekly progress!

Week 9

July 29, 2019 - August 11, 2019

Used open source code from Github called 'OpenPose' for human pose estimation in efforts to train a teacher-student network. New Plan: Use human pose estimations based off mmWave data and use for human activity recognition.

Weekly Presentation

Week 8

July 22, 2019 - July 28, 2019

Extracted 5 features (mean, median, skewness, kurtosis and standard deviation) from data to trains SVM. Tested SVM and was able to get perfect prediction for activities, however, off a small pool of data that we collected.
Began looking into CNN's for higher classification accuracy.

Weekly Presentation

Week 7

July 15, 2019 - July 21, 2019

Processed the data for the new experiments: walking toward/away, skipping toward/away and walking in a circle. Implemented an algorithm to reduce noise in Doppler FFT plot and began training SVM classifier for 5 activities.

Weekly Presentation

Week 6

June 27, 2019 - July 3, 2019

Conducted more experiments on five different activities. The activities were: walking toward and away form the sensor, skipping toward and away from the sensor, and walking in a circle in front of the sensor. By capturing a lot of data, we were able to run the results through our algorithm and use a SVM classifier to train the program to differentiate between the different activities.

Weekly Presentation

Week 5

June 20, 2019 - June 26, 2019

Conducted two experiments. One where we walk from left to right and the other where we swipe our hand from left to right in front of the mmWave radar. The goal is to use to data extracted to differentiation between the two activities.

Weekly Presentation