Design und Evaluation von Hardware-Architekturen zur mobilen Sonifikation von Bewegungen in der Schlaganfallrehabilitation

Today's challenges for the existing healthcare systems worldwide are the aging society and rising healthcare costs creating demand for computer-aided support systems in the medical environment. Exemplarily, this can be illustrated by electronic patient records, robot-assisted surgery and assistive devices for rehabilitation. An upcoming technique in motor rehabilitation is interactive auditory movement feedback. In contrast to visual feedback the auditory perception is used for feedback.

Main objectives of the so-called movement sonification are mobile motion capturing and real-time sound synthesis with a maximum latency of 30 ms to ensure simultaneous perception for visual, auditory and proprioceptive movement information for humans.

The purpose of this thesis is to provide an objective evaluation of the orientation estimation accuracy of IMU-based orientation estimation algorithms. Therefore, the raw data of different IMU-systems and a camera-based optical motion capturing system serving as golden reference, are used for the evaluation. Furthermore, a new error measure is developed and applied to enhance objective assessment. In addition, the performed design space exploration of hardware architectures allows the identification of Pareto-optimal hardware architectures. Besides commercial GPP and RISC processors, a customized ASIP core and a dedicated hardware accelerator for Kalman filter based orientation estimation illustrate the potential of inter- and intra-architecture optimization on software and hardware level.

A comprehensive latency and execution time analysis based on various sonification models enables effective optimization of execution time critical tasks. Application-specific real-time criteria allow platform independent performance assessment. Furthermore, the presented design space exploration of processing platforms for interactive movement sonification enables the identification of Pareto-optimal architectures regarding power dissipation and overall system latency. Comparable to the previous exploration regarding orientation estimation as application a customized ASIP and an optimized SoC architecture complete the evaluation and demonstrate the capabilities of inter- and intra-architecture optimization in software and hardware level.

The overall goal of this thesis is to characterize algorithms and processing platforms regarding the applications of IMU-based orientation estimation and interactive movement sonification.