Algorithm A2. ADL Conversion Algorithm—Pseudocode [26]

Input: 3D Accelerometer Raw Data Output: ADL (Sitting, Standing, Walking, Running, Jumping) and Unknown Initialization: CLOCK_FREQUENCY=100 ticks/second WINDOW_SIZE=100 Data Frame à df Time Parsing: For each entry in the ‘Time (elapsed)’ column Parse_time (h:m:s) = h x 3600 + m x 60 + s Remove NaN rows. Feature Extraction: For each sliding window of size n from df: mean_v = ${\displaystyle \frac{1}{n}}\sum _{i=1}^n{v}_{acceleration}\left[i\right]$ mean_ml = ${\displaystyle \frac{1}{n}}\sum _{i=1}^n{ml}_{acceleration}\left[i\right]$ mean_ap = ${\displaystyle \frac{1}{n}}\sum _{i=1}^n{ap}_{acceleration}\left[i\right]$ std_v = $\sqrt{{\displaystyle \frac{1}{n-1}}\sum _{i=1}^n(v_{acceleration}\left[i\right]{-mean\_v)}^2}$ std_ml = $\sqrt{{\displaystyle \frac{1}{n-1}}\sum _{i=1}^n({ml}_{acceleration}\left[i\right]{-mean\_ml)}^2}$ std_ap = $\sqrt{{\displaystyle \frac{1}{n-1}}\sum _{i=1}^n({ap}_{acceleration}\left[i\right]{-mean\_ap)}^2}$ Activity Classification: Sitting Standing Walking Running Jumping