Minor feature update release (v1.1.0)

Introduces the smoothing function smooth_timeseries() which applies a Kolmogorov-Zurbenko filter. This functionality is based upon a code example from the Envision Bootcamp by @WimPouw and James Trujillo.

How-to:

When placed directly in the demo folder, the following code demonstrates how the new smoothing function works:

# Install lastest OpenPoseR package from Github and load package
devtools::install_github("trettenbrein/OpenPoseR")
require(OpenPoseR)

# Load data from file in "demo" folder and plot
data <- read.csv("data_openposer/psychologie_body25_cleaned_en_velocity.csv", sep="")
plot_timeseries(data)

# Apply filter and plot result
filtered_data <- smooth_timeseries(data$Euclidean_norm_velocity, span = 2, order = 2)
plot_timeseries(data.frame(filtered_data))

Alternatively, there also is a wrapper function file_smooth_timeseries() that makes it possible to simply pass an output file to the smoothing function and have the result saved to a new file psychologie_body25_cleaned_en_velocity_smoothed.csv:

file_smooth_timeseries("data_openposer/psychologie_body25_cleaned_en_velocity.csv", span = 2, order = 2)

Bugfix release (v1.0.5)

Fixes a bug in clean_data() that resulted in NAs in the data frame in cases where a point was not tracked for several frames towards the end of a file (#12). Otherwise identical to v1.0.4.

Minor feature update release (v1.0.4)

This release slightly alters the default behaviour of en_velocity() and en_acceleration() (and by extension also file_en_velocity() and file_en_acceleration()). The changes are as follows: By default, the output of these functions is now prefixed with a data point that contains a 0 value. This is done by a new argument start_from_zero for all these four functions which is set to TRUE by default.

The reason for this change in behaviour is that both velocity and acceleration are actually computed from frame to frame whereas the first output data point is computed using the values of frame 1 and frame 2 and represents the change from frame 1 to frame 2. Hence, the value for frame 1 in the output of all above-mentioned functions should actually be displayed as 0 given that no value can be computed for it (given that we don't have any information about frame 0).

To prevent this new default behaviour the argument start_from_zero can be set to FALSE.

Otherwise this release is identical to v1.0.3.

Bugfix release (v1.0.3)

Fixes a bug in the clean_data() function which failed when a model other than BODY25 was used. Otherwise identical to v1.0.2.

Bugfix release (v1.0.2)

Fixes a bug in all file_ functions which failed when the model was not specified. Otherwise identical to v1.0.1.

Bugfix release (v1.0.1)

Fixes a bug in clean_data(). Otherwise identical to v1.0.

First stable version (v1.0)

This release marks the first "complete" version of the OpenPoseR package with all its originally intended functionality, a complete documentation, as well as a short demo of the method. A publication describing this version of the software has undergone peer-review and was published in Frontiers in Psychology: https://doi.org/10.3389/fpsyg.2021.628728