v4.0.1

What's Changed

• Setup for packaging (PyPI, Conda) by @chanshing in #177
• Major flake8 refactoring + GitHub workflows by @chanshing in #179
• Fixes and improvements: d61d86f, aacd87f, 582c0b2

v3.2

What's Changed

• add --timeZoneOffset argument for cwa cwa.gz files; expose csv arguments to python script by @xiangnandang in #21
• fix minor issue in accUtils.loadTimeSeriesCSV for --timeSeriesDataCol… by @xiangnandang in #23
• Circadian analysis including PSD, Relative Amplitude and Fourier Frequency by @R-Walmsley in #26
• Gt3v2 support round 1 by @angerhang in #24
• Major fixes and improvements in the Java feature extraction code by @chanshing in #28
• Gt3v2 fix get device id method by @angerhang in #29
• Document by @angerhang in #30
• fix lowpass cutoff to be capped by Nyquist freq by @chanshing in #31
• fix sampleRate not passed for calibration by @chanshing in #32
• fix rounding errors in integer divisions by @chanshing in #33
• fix convert pandas to numpy before numpy call by @chanshing in #34
• fix FFT and entropy normalization by @chanshing in #36
• hotfix to allow sampleRate of 25Hz by @chanshing in #39
• propagate verbose argument to CsvReader.java by @xiangnandang in #41
• Arm angel by @angerhang in #42
• allow display of activity classification during imputed period of time by @xiangnandang in #47
• allow daylight savings crossover check from different timezones by @xiangnandang in #48
• timeZone for check_daylight_savings_crossover by @xiangnandang in #51
• fix small bug: missing field in summary.json by @chanshing in #54
• Native DST crossover correction for CWA files by @chanshing in #55
• fix minor issue: arm angle result was in integer, corrected to double by @xiangnandang in #57
• make use of cutpoint based definitions explicit by @R-Walmsley in #59
• fix bug in get_interrupts, wrong index slicing by @chanshing in #61
• change dir structure when process multiple files by @chanshing in #60
• autogenerate folder names to avoid default errors by @R-Walmsley in #63
• add junit framework by @angerhang in #67
• Fix startTime and endTime comparison in summariseEpoch by @angerhang in #74
• fix start/endTime flag time zone by @angerhang in #76
• add clipsBefore/AfterCalibr and total reads to the summary. by @R-Walmsley in #82
• Add regression test; Add data quality outputs in summary by @R-Walmsley in #85
• Fixed gt3x v1 and v2 parsing with ci tests by @angerhang in #70
• Update default activity classification model to Walmsley 2020 by @aidendoherty in #88
• Fix sklearn dependency problem and new tests for python and java builds on macos and linux by @aidendoherty in #89
• avoid error message when epoch folder manually changed by @R-Walmsley in #95
• fix bug in get_clips and get_total_reads by @chanshing in #91
• handle potential nan in correlation features by @chanshing in #56
• fix bug in interrupts; deprecation; refactor by @chanshing in #99
• fix hrLabels missing last tick; also use 24 format by @chanshing in #104
• Changes to use scikit-learn 0.24.0 by @R-Walmsley in #108
• Update reference epoch.csv to reflect changes introduced in 2fabe36 by @chanshing in #112
• Implement installation with requirements.txt by @R-Walmsley in #109
• fix 'imputed' column in timeSeries file by @chanshing in #105
• fix NA for avgArmAngelAbsDiff with epoch size 5sec by @angerhang in #113
• update installation instructions addressing #114 by @R-Walmsley in #123
• add feature set used in model Walmsley et al 2020 and remove old files by @R-Walmsley in #118
• add CONTRIBUTING.md by @chanshing in #125
• fix install doc by @angerhang in #129
• Suggestion in CONTRIBUTING.md to checkout PR locally for review by @chanshing in #141
• add support for nearest neighbor resampling by @chanshing in #140
• Bump urllib3 from 1.26.2 to 1.26.4 by @dependabot in #143
• Bump pillow from 8.1.0 to 8.1.1 by @angerhang in #138
• Bump jinja2 from 2.11.2 to 2.11.3 by @angerhang in #135
• fix warning pandas keep_tz by @chanshing in #142
• update github badges by @angerhang in #149
• add to installation instructions; update version to current tag by @R-Walmsley in #148
• fix small bug in calibration rmse calculation by @chanshing in #160
• Bump urllib3 from 1.26.4 to 1.26.5 by @dependabot in #163
• Bump pillow from 8.1.1 to 8.2.0 by @dependabot in #164
• Plot update by @aidendoherty in #166
• fix bug which before read the packetLength field rather than the device… by @aidendoherty in #171
• Fall back for downloadDataModel by @angerhang in #172
• new feature to flexibly read in raw .csv files by @aidendoherty in #165
• update citation by @R-Walmsley in #176
• Bump pillow from 8.2.0 to 8.3.2 by @dependabot in #175

New Contributors

• @chanshing made their first contribution in #28
• @aidendoherty made their first contribution in #88

Full Changelog: v2.3.0...v3.2

Major updates to machine learning models and java code

This stable release contains a number of major updates including:

• Complete rewrite of underlying java codebase
• Fixing a number of minor issues in extracting frequency domain features (thanks @chanshing)
• Retraining of machine learning activity models (consistent with fixed features and now using n=100 trees in Random Forest)
• Support for Actigraph gt3x files, both versions 1 and 2 (thanks @angerhang)
• Support for circadian rhythm analysis (thanks @Alaina121, @R-Walmsley, and @chanshing)

Other updates include:

• Resampling of signal now possible (thanks @chanshing)
• Timezone parameterisation (thanks @xiangnandang)
• Improved plotting capabilities (thanks @xiangnandang)
• New arm angle features (thanks @angerhang)

Refactor to improve command line options

Major refactor to improve checks, prints, and cleanup (thanks @chanshing)

• Rename variable 'rawFile' to 'inputFile' so that we dont confuse it with 'rawOutputFile', which is now named 'rawFile'. Similarly 'npyOutputFile' is now simply 'npyFile'. Method 'processRawFile' is now called
  'processInputFile'.
• Use Python 'assert' to perform checks. Also simplify some checks, e.g. ArgumentParser automatically checks for missing arguments.
• Add new argument 'accProcess.py --npyFolder' for consistency.
• Use 'os.path.join' to create paths. This properly handles both cases whether a directory name is provided with or without the trailing slash '/'.
• Use 'atexit' to schedule deletion of intermediate files. This has the benefit that it will still run if the code exits prematurely. Some exceptions apply.
• Slightly improve prints.

Other updates:

• Filtering now correctly activated (thanks @chanshing)

Preliminary support for circadian rhythm analysis

v2.1.0

v2.1.0

Support for more input/output file types and activity classification training

Multiple updates include:

• Support to write raw data to .npy files (thanks @chanshing)
• Support to read compressed .gz files (offers a 75% space saving)
• Support to train one's own activity classification model
• More consistent code styling (thanks @chanshing)
• Improved documentation

Activity classification

Major update that includes:

• Activity classification
• Better unix command line support (we no longer support windows and GUI interfaces)
• Many minor bug fixes

The first release

As used for the 2017 UK Biobank PLOS ONE paper.

Updates for this release included:

• GUI support (thanks @supersational)
• Command line parsing (thanks @supersational)
• Improved documentation (for that point in time)