Modeling Distributed Signal Processing Applications

Wireless sensor networks in general and body sensor networks in particular enable sophisticated applications in pervasive healthcare, sports training and other domains, where interconnected nodes work together. Their main goal is to derive context from raw sensor data with feature extraction and classification algorithms. Body sensor networks not only comprise a single sensor type or family but demand different hardware platforms, e.g., sensors to measure acceleration or blood-pressure, or tiny mobile devices to communicate with the user. The problem arises how to efficiently deal with these heterogeneous platforms and programming languages. This paper presents a distributed signal processing framework based on TinyOS and nesC. The framework forms the basis for a model-driven software development approach. By raising the level of abstraction formal models hide implementation specifics of the framework in a platform specific model. A platform independent model further lifts modeling to functional and non-functional requirements independent from platforms. Thereby we promote cooperation between domain experts and software engineers and facilitate reusability of applications across different platforms.