A project to analyze lindy hop movement
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


DanceLab: general information

This project is designed to ease data collection for the analysis of Lindy Hop movement. On every run, the application produces 3 timestamped files:

  • Acceleration and gyroscope data
  • Audio data recorded from the microphone that corresponds exactly to the logged acceleration data
  • Metadata giving the start and stop timestamps, the beep timestamps, and the computed sampling rate

Compatibility: currently, this only runs on Gingerbread devices and higher due to the use of the gyroscope and AAC APIs.

Usage notes

  • Hit either “Start” or “Sync” to start recording. Long-press “Stop” to stop.
  • NTP Sync: to better align the data, the app performs an NTP query on startup; you can force the NTP query through a menu item. NTP data is saved for later use until the phone is restarted (this is useful for devices that do not have permanent network connectivity).
  • Sync functionality (non-NTP): when the sync button is activated (and turns red) the phone waits for an acceleration spike that exceeds the threshold of the dotted line on the screen (controllable in preferences). Once it registers the spike, the button turns green and the time is recorded in the metadata file; it is also set as the zero timepoint in the RUN_MS time data column. To reset the peak detector (e.g. after an erroneous registration), or to turn on the peak detector after the logger has been started, hold down the sync button. Holding down the sync button while it is red turns off the peak detector.

    One way to use the sync functionality is to activate sync on the two phones, hold them both in the same hand, and rotate the wrist abruptly (you can see that this produces a large acceleration spike). Hopefully, they will both register a clean acceleration spike that will serve as the sync signal.

    In practice, sync gives good acceleration trace alignment, but is not superior to NTP under optimal conditions.

  • The screen stays on throughout logging. Apparently, collecting acceleration data in standby mode is non-trivial or impossible (I have not seen definitive answers about whether partial wake locks solve this).
  • Data collection will stop if home or power buttons get pressed during logging, so try to avoid that.
  • Current data columns: RUN_MS, SENSOR_TYPE, EPOCH_MS, ACCEL{x,y,z}, ACCEL_MAGNITUDE, GYRO{x,y,z} Note that RUN_MS is not guaranteed to be monotonic; for instance, every time peak detector sync is completed, the RUN_MS time is re-zeroed for subsequent datapoints. EPOCH_MS stores the absolute time (epoch format), with the NTP correction. SENSOR_TYPE is 1 for accelerometer, 4 for the gyro.
  • audio can be converted into a form suitable for Matlab/SciPy/Mathematica analysis via ffmpeg -i test.3gp -vn -f wav -acodec pcm_u8 test.wav