About project

December,3 2017 - Posted by.Henri

Effective implementations for active noise control in adaptive distributed networks

The proposal aims to address a daily life problem in entertainment and communication such us noise control. Nowadays, we witness the boom of applications for personal use devices and their massive extension in parallel with the advances of communications and signal processing. In general terms, this proposal is focused on signal processing algorithms and systems which interact with sound environments and use the network of acoustic nodes in an efficient way.

Scientific context and motivation

The wireless sensor networks (WSNs) are a cheap and flexible solution for environmental, habitat and industrial monitoring. The proposed active noise control (ANC) project involves “smart” sound processing techniques that include intelligent generation and/or capture of sound signals by using signal processing devices potentially heterogeneous and distributed. In the centralized system the error and reference signals are available for the multichannel controller. The principal investigator (PI) has proposed multiple LMS, Affine Projection, Affine Projection Like algorithms or RLS algorithms for centralized ANC systems. Most of them obtained a good performance for ANC systems, having low communication and computational requirements. However, there is still a clear need for better noise reduction (NR) performance of the distributed algorithms for ANC. We expect that the proposed adaptive distribution algorithms would achieve similar performance with the centralized versions, having low computational requirements and improved scalability in a ring network. Simulations with real acoustic channel responses would be made for a WASN with N single-channel nodes that will support an ANC system composed of N error sensors and N secondary sources and adapt the internal parameters based on latency, real-time computational possibilities and acoustic coupling levels. Time domain or frequency domain techniques, partitioned frequency or partitioned time domain techniques, partial update techniques, proportionate versions will be investigated toot

Objectives

1. To develop new active noise control algorithms for adaptive distributed networks. New approaches will be investigated by using sparse adaptive algorithms, the division-less and multiplier-less Dichotomous Coordinate Descent (DCD) method, Gauss-Seidel (GS) method, various affine projection (AP) algorithms, AP-like algorithms, proportionate AP or RLS algorithms.

2. To propose new algorithms integrating noise reduction and active noise control approaches for various system identification applications such as digital hearing aids. Also, to investigate the suitability of kernel adaptive algorithms for nonlinear identification.

3. To develop and evaluate an integrated solution for adaptive distributed networks by examining the required number of microphones and loudspeakers that enable real-time implementation and weighting advantages and disadvantages of the investigated algorithms.

This work was supported by CNCS-UEFISCDI