Wireless Sensor Networks (WSN) have been well researched as a generic concept. In this project, part of a DFG Forschergruppe (DFG Research Group), we are concerned with the specifics of wireless acoustic sensor networks.
On one hand, fulfilling a data acquisition task does not only depend on the acoustic environment but also on the available sensor nodes capacities (microphones, energy). On the other hand, data processing depends primarily on available resources for computing and communication which corresponds to the required quality of the data acquisition task. Another important aspect is the algorithm selection which depends mainly on the scene at hand (e.g. single or multiple sources).
Knowing that acoustic applications require a high level of resolution and given an acoustic signal processing task and a sensor network, key questions are which sensors should record, process or store acoustic data and which algorithm should be run on which node?
Given that there is a strong interdependence between data acquisition, processing and algorithm selection, we approach this interdependence as a joint question of:
- Role assignment and parameter selection where the algorithm choice is perceived as a parameter as well.
- Different roles (e.g., sensing, storing, transmitting, processing) have to be assigned to various nodes
- Algorithm parameters (e.g., sampling rates, FFT parameters) have to be selected for mechanisms, algorithms, and protocols on all system layers.
Distributed versions of algorithms will be developed that are tuned to the limitations of the wireless network. We investigate how data streams have to be organized for an optimal trade-off between acoustic signal processing performance and resource efficiency of the communication network.
The Computer Network Research Group of Paderborn University is mainly involved in the following work packages of the overall Research Group.
Work package 2: Quality-resource tradeoff of distributed algorithms
Using different distributed algorithm versions, we shall characterize their tradeoffs between obtained result quality and invested resources. For example, we will check how different packet error rates, limited data rate, limited processing power for iterative algorithms or – in particular – different choices of acoustic sensors (microphones) impact the obtained quality
Work package 3: Role assignment and parameter control
Decisions have to be taken on the basis of the current network topology, the source locations in the network, and the available distributed algorithms’ tradeoff profiles. Decisions shall select the components of the distributed algorithms which have to be deployed onto concrete nodes, apply route configuration, and select the concrete acoustic sources.
Work package 5: Cross-layer interoperability and interfacing
This work package addresses the need for a formal description by specifying which data and parameters of one layer have to be made available to which processing task of the other layer using typical network configuration formalisms. We will design ways to organize feedback loops between lower-layer mechanisms (like the role assignment) and higher-layer acoustic algorithms to exchange information to explore an option space at runtime
Work package 7: Experimental Setup
This work package will port and integrate the developed algorithms along with proposed schemes on a concrete hardware platform.