Results - Details
Search command : Author="Μουχτάρης"
And Author="Αθανάσιος"
Current Record: 5 of 29
|
Identifier |
000416964 |
Title |
Multiple sound source location estimation and counting in wireless acoustic sensor networks |
Alternative Title |
Εκτίμηση θέσης και αριθμού πολλαπλών ηχητικών πηγών σε δίκτυα ακουστικών αισθητήρων |
Author
|
Αλεξανδρίδης, Αναστάσιος Ι
|
Thesis advisor
|
Μούχταρης, Αθανάσιος
|
Reviewer
|
Τσακαλίδης, Παναγιώτης
Moonen, Marc
|
Abstract |
Wireless acoustic sensor networks (WASNs) represent a new paradigm for acoustic
signal acquisition. Multiple acoustic nodes that feature processing and
communication capabilities are distributed in the environment where typically
multiple sound sources are active. In such a setup, inference of location information
has always been an attractive research problem. Enabling machines to estimate the
locations of the multiple simultaneously active sound sources from their acoustic
emissions is crucial in many applications, such as wildlife monitoring and speech
enhancement for robust signal acquisition.
Throughout the years, different localization methods have been proposed with the
everlasting goal to achieve the lowest possible localization error. Although significant
steps have been made towards this direction, another unexplored field concerns the
practical limitations posed by the sensor network that restrict the application of such
methods to real-life WASNs. Such limitations include the limited processing power and
battery life of the nodes, the communication bandwidth that has to be attained at low
levels, the real-time requirements and synchronization issues. In this thesis, we
consider the problem of multiple source localization and we investigate the
development of methods that not only achieve high accuracy in realistic scenarios but
also attain low communication bandwidth, tolerate unsynchronized input and are
computationally efficient to facilitate their application in real-life WASNs.
We consider a WASN where each node is a microphone array that estimates and
transmits information related to the directions of arrival (DOAs) of the active sound
sources. Such a scheme attains very low communication bandwidth, as only the DOAs
need to be transmitted. Moreover, DOA-based localization methods can tolerate
unsynchronized input, thus the acoustic signals need not be perfectly synchronized.
We first focus on the single source case and propose a computationally efficient nonlinear
least squares estimator that can accurately estimate the source's location using
an iterative grid-based approach. We then proceed to the multiple sources case,
assuming that the number of sources is known. In this case, a core problem for DOAbased
approaches is that the fusion center that receives the multiple DOA estimates
from the nodes cannot know to which source each DOA belongs. This is known as the
data-association problem. To address this problem we propose two solutions: the first
concerns the extension of our grid-based approach to multiple sources and the second
utilizes additional information, apart from the DOA estimates, in order to find the
correct association of DOAs from the nodes to the sources. We then relax the
assumption of known number of sources and propose another method that can jointly
perform source counting and location estimation. Our method is based on clustering
narrowband per-frequency location estimates which are inferred using narrowband
per-frequency DOA estimates from the nodes.
Since a determinant factor that affects localization performance is the accuracy in
which the DOA estimates are obtained, we also investigate how we can improve DOA
estimation performance and we propose a methodology to infer more accurate and
reliable DOA estimates. Finally, we investigate the potential use of location
information to audio processing applications. We provide two examples of how
location information can be used for spatial audio capturing and for the design of
beamformers that leverage location information in order to estimate the steering
vector of the target source. Our preliminary results reveal the potential of locationbased
approaches to provide improved performance.
|
Language |
English |
Subject |
Direction-of-arrival estimates |
|
Localization |
|
Location information |
|
Microphone arrays |
|
Εκτιμήσεις κατεύθυνσης άφιξης |
|
Εύρεση θέσης |
|
Πληροφορία θέσης |
|
Συστοιχίες μικροφώνων |
Issue date |
2018-03-23 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Doctoral theses
|
|
Type of Work--Doctoral theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/a/3/c/metadata-dlib-1530864967-475045-5324.tkl
|
Views |
486 |