Conclusion

In summary, the experimental finding shows a smaller synergy index during tracking of familiar compared to unfamiliar trajectories. In contrast, motor control theory predicts that minimizing the tracking error in the target space implies that the synergy index decreases with impaired prior knowledge about the target trajectory. This prediction is independent of whether the synergy index is explained by control mechanisms for the compensation of planning noise or peripheral motor noise. Consequently, the opposite experimental finding suggests that the movement goal (formalized by the cost function, and achieved by the control strategy) differs between tracking of familiar and unfamiliar

trajectories. The difference can be characterized by a modification of the task error being minimized. For visually driven tracking of unfamiliar trajectories, the task error seems to be defined in target coordinates, whereas for familiar trajectories it seems to be defined in motor coordinates. This strategic shift between visually driven and automated tracking movements explains the observed decrease of the synergy index on familiar trajectories.

Chapter 3

A Development Framework for Arm Movement Measurements

Subsections 3.1and 3.2 in the current chapter are based on the author’s article entitled

“A measurement system for wrist movements in biomedical applications”. [C1]

3.1 Background

The area of health assisting technology have been more and more active in the last few years in the field of medical instrumentation, movement rehabilitation, prosthetic devices and fitness accessories, just to name a few. This process resulted in a wider spread of devices addressing different areas on the border of life sciences and engineering, from simple commercial products (e.g. small pulse monitors) to very complex research projects like the Modular Prosthetic Limb¹ and its control interfaces.

Human movement recording – a complementary area to these fields – however, did not show this level of activity. The presumable reason for this is that the measurement meth-ods in motion tracking applications are standardized, tested and validated since decades and manufacturers keep providing high level laboratory systems to meet these require-ments. The most widely used measurement systems apply line-of-sight (LoS) methods (optical or ultrasound-based) that require a fixedmarker-sensor structure. In these cases

1http://www.jhuapl.edu/prosthetics/scientists/mpl.asp

28

passive (optical) or active (ultrasound) markers are placed on anatomically relevant lo-cations of the studied subject. Having the markers in place, measurements have to be performed in a specialized laboratory environment where locations and orientations of the sensor elements (i.e. cameras or microphones) are known and invariant (at least across trials). This means that even though the spatial positions of the markers can be determined with good accuracy – especially with optical systems – the possible range of motion will always be constrained by the actual measurement volume covered by the sensors of the system. Although this property is not an issue for many movement anal-ysis scenarios, there are cases when a measurement method allowing unconstrained free space movement would be more beneficial (e.g. various outdoor activities or ergonomic assessment of work environments, to name a few).

Advancements in the field of inertial sensor technology have given rise to new devel-opment directions in laboratory-free movement analysis methods. The main difference between LoS and inertial systems is the recorded modality: while LoS methods determine thespatial locations of markers based on planar position (optical) or timing (ultrasound) information, inertial sensors give theirorientationin space by measuring physical quanti-ties acting on them directly. These quantiquanti-ties are linear acceleration and angular velocity in most cases while they are supplemented with magnetic field measurements in more complete setups. To obtain orientation from raw inertial measurements, various sensor fusion algorithms have been developed utilizing Kalman-filters [39,40], gradient descent methods [41], complementary filters [42,43] and other techniques [44], most of them being capable for real-time operation in embedded systems. In addition, the recent evolution of chip-scale inertial sensors based on MEMS technology further widened the possibilities of wearable measurement device development by making the core sensing elements avail-able for better integration. While there is no gold standard among fusion algorithms and sensor chips as compromises have to be made in aspects of accuracy, system complexity and computational demand of the fusion algorithm, it can be stated that inertial sensor technology is taking a more and more growing part in human movement measurements (a good example for this progress is Xsens’ product portfolio).

A further aspect with regard to biomedical applications is the possibility of recording other modalities, most importantly bioelectric signals like muscle activities (electromyo-gram, EMG). Measuring EMG is a key aspect to get deeper insight into the dynamics of movements because it gives closer information about muscle activation patterns. While

there are many approaches in the literature for individual EMG recording [45, 46, 47, 48], research efforts towards integrated systems utilizing kinematic measurements and muscle activity recording in the same package has only been started recently [49, 50]

2. As to the best of the author’s knowledge, there is only one commercially available EMG+IMU system on the market (TRIGNO^TMIM, provided by Delsys from 2016Q3³), that incorporates many small sensor units, each containing a single-channel EMG elec-trode and an inertial sensor. This system, while seeming to be a good overall solution for whole-body performance assessment in various situations, has some major differences from the system presented in the current chapter, e.g. it may use a proprietary and closed source kinematic model for movement reconstruction that needs to deal with the unique placement of the sensors. In particular, sensor placement seems to be optimized for EMG measurements (i.e. targeting muscles) and not for IMU operation.

The work described in this chapter is an effort towards the development of a research oriented, fully customizable, integrated and wearable measurement system specifically targeting arm movements in order to record and analyze movement patterns for biomed-ical applications. The system aims to provide a framework for development and testing of inertial sensor based movement measurement and analysis techniques including custom hardware setup, calibration routines, sensor fusion and device control from various plat-forms. While the complete system design includes EMG measurement capability, this chapter focuses mainly on the practical considerations and implementation of kinematic movement recording.