Data quality is important for any type of research. In case of the full-body motion capture data I have collected during my real-world study, this means checking the avatar for “weird motions”.
This is a very important aspect: As the sensors are prone to electromagnetic disturbance, the body movements may no longer be accurately recorded. Together with one of my RAs at UCSB, I check each of the recordings for disturbances (e.g. a head being turned by 180 degrees) etc. Luckily enough, it is possible to mark the affected frames in the recording and, then, reprocess the data. During this process the whole file is being algorithmically corrected, drastically reducing the number of “weird motions”.
The results look really impressive! So far, I think I will need to exclude only a very small percentage of the collected data 🙂