Amin Hassani

Research

Distributed node-specific signal enhancement algorithms for single-task and multi-task wireless acoustic sensor networks

In this research project, algorithms for node-specific signal enhancement problems in a Wireless Acoustic Sensor Network (WASN) are being developed. In general, if each node would have access to the microphone signals of all the other nodes (e.g., via a centralized unit), optimal node-specific estimation performance could be achieved. However, this would require a large communication bandwidth and computational power. We alternatively consider a distributed processing in which each node only broadcasts a fused signal to the other nodes and hence the centralized computational cost is shared between the different nodes. The resulting algorithms aim to iteratively obtain the same performance as the centralized unit in a block-adaptive fashion.

The first class of such algorithms have been focusing on single-task networks, i.e., a setting where all nodes cooperate to obtain different but coupled solutions of the same Signal Processing (SP) task. Distributed algorithms for node-specific Direction-of-Arrival (DOA) estimation, node-specific signal estimation (speech enhancement) based on Generalized EigenValue Decomposition (GEVD), and GEVD-based network-wide signal subspace estimation have been developed in this class.

The second class of algorithms, currently under development, deal with heterogeneous multi-task networks, where the nodes are interested in solving different but interrelated SP tasks. Hence each node may apply different SP techniques (e.g., node-specific spatial/spectral filters or beamformers) in order to fulfill the particular performance requirement of its application layer. In this class, again the ultimate goal for each node is to converge to the corresponding centralized solution, despite the fact that each node solves a different local task.

PATENTS

US Provisional Patent Application US62908483 “Active Noise Equalisation Apparatus and Method”, 30 September 2019

Finished projects

• Cooperative Signal Processing Solutions for IoT-based Multi-User Speech Communication Systems
• Distributed Network of Active Noise Equalize - D-Noise

Publications