E-Locus - Institutional Repository of the University of Crete

Home Search

Results - Details

Search command : Author="Μουχτάρης" And Author="Αθανάσιος"

Current Record: 26 of 29

[Add to Basket]

Identifier

000381055

Title

Directional coding of audio signals using a circular microphone array

Alternative Title

Κωδικοποίηση κατευθυντικής πληροφορίας ηχητικών σημάτων χρησιμοποιώντας μαι κυκλική συστοιχία μικροφώνων

Author

Αλεξανδρίδης, Αναστάσιος Ι

Thesis advisor

Μουχτάρης, Αθανάσιος

Abstract

Microphone arrays have attracted great attention in the last decades. The main reason is their ability to perform sound source localization and beamforming. Moreover, based upon the knowledge of sound propagation, microphone arrays have great potential in the field of noiserobust speech capture and hands-free signal acquisition. Microphone arrays—and particularly circular arrays—are already used in several modern speech communication systems, such as teleconferencing and next generation hearing aids. In this thesis, we focus on recording and reproducing the spatial characteristics of an arbitrary sound field, and propose a new method for extracting and coding the directional information of audio and speech signals using a circular array of microphones. Our method is computationally efficient—it consumes approximately 50% of real-time—and thus is suitable for real-time implementations. We model the sound field based on estimating the Direction-of-Arrival (DOA) of all simultaneously active sound sources and separating the source signals through spatial filtering with a fixed superdirective beamformer. In contrast to previous work, our DOA estimation procedure is not based on a strict W-disjoint orthogonality assumption for the sound sources (i.e., we do not assume that each time-frequency element is dominated by only one sound source), which is expected to make the modelling of the sound field more accurate. The separated source signals are downmixed into one audio signal, and as a result, the sound field is encoded using one monophonic audio signal and side-information. To reduce the bitrate requirements, the monophonic audio signal can be encoded with any compression algorithm, such as MP3. We also propose an efficient lossless compression scheme for the side-information. The recorded sound field is reproduced using amplitude panning for loudspeaker reproduction or Head-Related Transfer Function (HRTF) filtering for reproduction via headphones. In order to evaluate the performance of our proposed method, we conducted listening tests using microphone array recordings in simulated and real environments. We considered both reproduction over headphones and loudspeakers, while the recordings included both speech and music signals of stationary and moving sources. Our approach was compared with other recently proposed microphone array based methods for spatial audio. Our listening test results reveal that our method achieves excellent reconstruction of the sound field while maintains the sound quality at very high levels. Lastly, we investigated the effects of coding the audio signal with an MP3 encoder and found that coding the downmixed signal at 64 kbps results in unnoticeable degradation in the spatial impression and sound quality of the reconstructed sound field, compared with not applying any compression.

Language

Greek, English

Subject

Beamforming

Spatial audio

Σχηματισμός λοβού

Χωρικός ήχος

Issue date

2013-07-19

Collection

School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses

Type of Work--Post-graduate theses