Generic data structure

In a large universe of acquisition systems, data are stored sparsely within internal structures specific to each constructor. Thus, users have to deal with several non-compatible structure to prepare some processing or analyses. What about the possibility to store data in a generic structure without any compromise in terms of contents or possibilities?
This project proposes the use of a graph-like data structure composed of nodes to store your data. Nodes can be linked between them to represent complex structure, like trials, hardware systems, or skeleton models.

File formats

To be able to access to acquired data we need file parsers. Indeed, file formats of each company are most of the time different and proprietary. Sometimes, companies offer us their file specification or we receive code from contributors and we can easily write the parser. Otherwise, it is necessary to pass by a reverse engineering process that can be long and painful. The contribution of everyone (e.g. researcher, engineer, laboratory, or company) is very welcome to increase the number of file formats read by the project.

Raw data processing

Depending on your setup and your needs, you might have to filter your data, resample them, or transform them to estimate your parameters of interest. In some case you also need to take into account occluded data. This dramatically complexifies the processing and the reading of your algorithm. OpenMA automatically manages the last point directly in its mathematical module.

Movement reconstruction

Another major feature is the possibility to create models based on predefined markers sets (e.g. Plug-in Gait, Helen Hayes) or custom ones. You can then compute joint kinematics and kinetics or pass specific information to other tools dedicated for example to computer-graphic animation or musculoskeletal analysis.

Joint kinematics & kinetics

Several descriptors are proposed to extract joint kinematics as Euler angles, quaternions, axis angles, and so on. You only have to specify which ones you want to use for any joint setup in your model. The same idea exists for joint kinetics estimation. You can express them in the global frame, the adjacent relative frames, or the joint frame.

Event detection

Each activity has its own events that distinguish in which part of the activity the subject is. While these events are well described in the literature, you might not want to detect them manually. Automatic methods were already published and should not need to reimplement them. Moreover, sometimes it is interesting to have benchmarks between methods. OpenMA wants to offer you that.

Cycles averaging

From events, you might want to extract cycles in one trial or several trials. This task is not the most difficult but can be cumbersome. This is without talking about the need to resample data to be able to average them at specific times or percentages. You may want also use median and quartiles instead of mean ad standard deviations.

And much more!

Previous sections are only examples of what type of tools you could use to prepare your analysis. Several tools will be implemented to help you to design and run your analysis.

User case: Accelerate CGA report generation

A gait lab had a working solution composed of several tools to generate a clinical gait analysis (CGA) report. After data acquisition, it was necessary to convert data into several files, launch a first software to compute joint kinematics and kinetics, save the data into a new file, launch a second software to do the analysis, save it in a new file, and finally launch the generation of the report. The use of OpenMA replaced several of these steps and removed the need to write temporary files. At the end, the user only had to do its acquisition and launch one supplementary software to do computation and generate the CGA report!

User case: Collaboration with different motion capture systems

Several laboratories wanted to analyze a rare pathologic population. However, each of them used different motion capture systems that propose different predefined markers sets. Thus, theoretically, it was not possible to aggregate their data together. Indeed, segment definitions are not the same between markers sets and this largely influences joint parameters estimation. The solution would be to use a single markers set across these systems. Thanks to OpenMA, this solution was possible and eventually they successfully realized the study!