AES Munich 2009
Poster Session P20
P20 - Signal Processing
Saturday, May 9, 13:30 — 15:00
P20-1 Subjective Audio Quality with Multiple Description Coding in a WLAN 802.11-Based Multicast Distribution Network—Marcus Purat, Tom Ritter, TFH Berlin - Berlin, Germany
This paper presents a number of results of a study of different methods to mitigate the impact of packet loss on the subjective quality in a wireless distribution network of compressed high fidelity audio. The system was simulated in Matlab based on parameters of an 802.11a WLAN in multicast-mode and the Vorbis codec. The performance of multiple description coding in terms of expected subjective audio quality and its impact on the data rate is quantified and compared to other receiver-based packet loss concealment methods. The optimum set of parameters for a multiple description coder that achieves the best subjective audio quality for a given packet loss rate in the simulated system is presented.
Convention Paper 7755 (Purchase now)
P20-2 A Joint Approach to Extract Multiple Fundamental Frequency in Polyphonic Signals Minimizing Gaussian Spectral Distance—Francisco J. Cañadas-Quesada, Pedro Vera-Candeas, Nicolás Ruiz-Reyes, Julio Jose Carabias-Orti, D. Martínez-Muñoz, University of Jaén - Jaén, Spain
This paper presents a joint estimation approach to extract multiple fundamental frequency (F0) in monaural polyphonic music signals. In a frame-based analysis, we generate a spectral envelope for each combination of F0 candidates, from non-overlapped partials, under assumption that a harmonic sound is characterized by a Gaussian mixture model (GMM). The optimal F0 candidates combination minimizes a spectral Euclidean distance measure between the original spectrum and Gaussian spectral models. Evaluation was carried out using several piano recordings. Evaluation shows promising results.
Convention Paper 7756 (Purchase now)
P20-3 A Mixture-of-Experts Approach for Note Onset Detection—Norberto Degara, Antonio Pena, Manuel Sobreira-Seoane, Universidade de Vigo - Vigo, Spain; Soledad Torres-Gijarro, Laboratorio Oficial de Metroloxía de Galicia (LOMG) - Tecnópole, Ourense, Spain
Finding the starting time of events (onsets) is useful in a number of applications for audio signals. The goal of this paper is to present a combination of techniques for automatic detection of events in audio signals. The proposed system uses a supervised classification algorithm to combine a set of features extracted from the audio signal and reduce the original signal to a robust detection function. Onsets are obtained by using a simple peak-picking algorithm. This paper describes the analysis system used to extract the features and the details of the neural network algorithm used to combine them. We conclude by comparing the performance of the proposed algorithm with the system that obtained the first place in the 2005 Music Information Retrieval Evaluation eXchange.
This paper presented by Soledad Torres-Gijarro.
Convention Paper 7757 (Purchase now)
P20-4 Automatic Adjustment of Off-the-Shelf Reverberation Effects—Sebastian Heise, Michael Hlatky, Hochschule Bremen (University of Applied Sciences) - Bremen, Germany; Jörn Loviscach, Fachhochschule Bielefeld (University of Applied Sciences) - Bielefeld, Germany
Virtually all effect units that process digital audio—software plug-ins as well as dedicated hardware—can be controlled digitally. This allows subjecting their settings to optimization processes. We demonstrate the automatic adaptation of reverberation plug-ins to given room impulse responses. This facilitates replacing computationally expensive convolution reverberation units with standard ones, which also are amenable to easier parameter tweaking after their overall setting has been adjusted through our method. We propose optimization strategies for this multi-dimensional nonlinear problem that need no adaptation to the particularities of each effect unit, are sped up using multicore processors and networked computers. The optimization process evaluates the difference between the actual response and the targeted response on the basis of psychoacoustic features. An acoustic comparison with effect parameter settings crafted by professional human operators indicates that the computationally optimized settings yield comparable or better results.
Convention Paper 7758 (Purchase now)
P20-5 Improvements on Automatic Parametric Equalization and Cross-Over Alignment of Audio Systems—German Ramos, Pedro Tomas, Technical University of Valencia - Valencia, Spain
The idea and algorithm implementation of an automatic parametric equalizer and cross-over alignment of audio systems was proposed previously by one of the authors with proven success. This method designed Infinite Impulse Response (IIR) equalization and cross-over filters directly in a series of second-order-sections (SOS) employing peak filters, and pre-initialized high-pass and low-pass filters, defined by its parameters (frequency, gains, and Q). The method supported the inclusion of constraints (maximum and minimum parameter values) and designed the SOS in order of importance in the equalization, providing thus a scalable filter implementation. In order to lower the order of the needed filter, and looking also for an automatic decision on the selection of filter types and initialization, several improvements are presented. It is now possible for the algorithm to select, configure, and use Shelving filters in the SOS chain for equalization. Also, the decision and initialization of the needed high-pass and low-pass SOS filters could be automatic, helping in the cross-over design stage for active audio systems.
Convention Paper 7759 (Purchase now)
P20-6 Low Noise Transformer Input Preamp Design—A Solution that Eliminates CMID—Milan Kovinic, MMK Instruments - Belgrade, Serbia; Dragan Drincic, Advanced School for Electrical & Computer Engineering - Belgrade, Serbia; Sasha Jankovic, OXYGEN-Digital, Parkgate Studio - Sussex, UK
This paper examines more closely the advantages of input transformer-op amplifier configurations, especially those implemented in low-noise designs. Usual transformer input stage topology works in non-inverting architecture, since it allows the transformer to work with optimum loading, to maximize the signal-to-noise ratio. However, this configuration is subject to Common-Mode voltage Induced Distortion—CMID. The susceptibility is further increased if the amplifier source impedance is not perfectly matched. This is illustrated by tests on popular audio op amps. Advanced transformer input stage topology proposed in this paper completely prevents this kind of distortion. Noise performance remains unaffected, yet listening tests in practical application confirm the sound to be more pleasant.
Convention Paper 7760 (Purchase now)