top of page

Methodology & Results

Method

  1. mmWave sensor triggers 150 frames over 10 seconds and captures data

  2. Process mmWave data and perform 2-D Fast Fourier Transform (FFT)

  3. Camera takes a picture in sync with mmWave sensor

  4. Images are further processed using open-source project OpenPose [1] to be used as labels

  5. Classification model is a teacher-student network similar to [2] that is composed of a Convolutional Neural Network (see Figure 2), using built-in Python packages

  6. Model is further tested using dynamic data

  7. Provides a human pose estimated figure performing activity and  classification of activity

block diagram.PNG

Results

We trained our classification model for 150 epochs with an Adam optimizer and a total of 1050 data samples.  Our current model can classify amongst three different activities: stretching, raising dumbbells, and sitting down. The experiments for these activities have 450, 300, and 300 samples respectively.

diag2.PNG
bottom of page