• plot_all_decel.py : plots all calibrated accelerations found in a folder

• plot_trips_selection.py allows to compare a selection of trips, both for calibrated/uncalibrated and raw/filtered data

• plot_FFT_single_trip.py: plots the FFT of calibrated accel of a single trip

• find_constant_accel.py was the first attempt to identify time windows where accel is constant. Leads to false positives (see ep.2 : https://www.linkedin.com/pulse/data-analysis-paris-metro-ep2-eric-cabrol-tq7ye/)

• find_jolts.py : second attempt, using "jolt-back" peaks => seems better (see ep.3 : https://www.linkedin.com/pulse/data-analysis-paris-metro-ep3-eric-cabrol-lynie) A file timestamps.txt is generated in each folder. It still requires a manual validation, then the file must be saved as timestamps_valid.txt.

• find_jolts_all.py : detect jolts in all recordings of a given directory => in fact there are as many difficulties to identify the stops, so I switched back to the constant accel solution

• find_constant_accel_all.py : asks for confirmation after each trip, which allows to modify the timestamps file before validating

• previously done in find_stops_from_trip_name.py, now in metro.py module

• cut_trips.py : could be merged with check_timestamps.py

• metro.py : contains a class named Trip to retrieve all useful information from the trip name. Shall I also manage quantitative data with this class (eg sampling frequency) ?

NB : zip files exported from SensorLogger must be manually downloaded from Google Drive into zip folder

1. unzip_recordings.py allows to unzip recordings in data
2. remove_total_accel.py removes TotalAcceleration.csv files, because they contain the same data as AccelerometerUncalibrated.csv