Utilization of Sensor Data

Your sensor data is collected and you have reliable and accurate body kinematics at hand. Now let us finally take care about our measurement’s main purpose – motion analysis!

Various reasons and motivations exist for the development of a technology supported motion analysis system. Examples are personal performance improvement, performance surveillance, support of coaches and judges or additional information for spectators. Such diversity has its price and comes with a wide spectrum of possible computation methodologies and algorithms.
For this reason, it is first important to be well aware of your main intention: should your analysis provide information about motion technique, errors during motion performance and/or suggestions for improvements? And which type of user should your analysis serve: athletes, coaches, judges and officials or even spectators? This allows you to specify all necessary system application requirements in the second step and ensures that suitable data mining and information retrieval algorithms will get chosen and implemented in the end.

field motion analysis setting

Most important for the provision of relevant motion information is to transform the kinematic (and pretty complex) motion data into a suitable (and less complex) performance description. As a general principle, the sensor data should be utilized in such a way that their transformations convey highly explicit motion information. This means that ideally no additional motion knowledge would be required from the user to understand the display of the computed data.

To fulfill this demand, it is often necessary to find alternative sources of motion knowledge. Most commonly, they consist of artificially built neural networks simulating or imitating the complex knowledge of the human brain. The essential question that has to be answered  in this context is how to represent biological perception of sensorimotor processes. Numerous machine learning strategies exist, but were seldom used on human motion (and particularly sports) data so far. Thanks to the availability of cheap and ubiquitous mobile motion capture systems, this state of research is likely to change within the next years.

The learning of machine motion knowledge for motion analysis is hence likely to become a matter of growing interest soon. And this in other words can only mean that you should grab your sensors now, find inspiration for the design of your own application and then get started…

…to become a pioneer of ubiquitous motion analysis and training!