Academic • 2025
Time Series Feature Extraction
A coursework analysis of statistical features and bootstrap intervals for the UCI AReM activity-recognition dataset.
My Role
Implemented feature extraction, resampling, comparison plots and notebook documentation.
Team Size: 1 person
Problem
Multichannel sensor sequences need a useful representation before a classifier can distinguish activity patterns. Feature variability also needs to be examined.
Approach
Extracted time-domain summaries from the sensor channels, used bootstrap resampling to examine variability and documented the prescribed training/test split in a notebook.
Impact
Created and compared feature representations for an academic dataset. The analysis does not establish a deployed activity-recognition system or benefits in healthcare and wearable products.
Technologies
PythonJupyterpandasNumPySciPy