Menu:
Supervision of PhD. Thesis
Past PhD. Thesis
Title: Design and optimization of recursive digital filters by means of direct methods for audio applications
Student: German Ramos
Date of presentation: 18 Sep 2006
Title: Contributions to Discrete-Time Domain methods in Room Acoustic Simulations
Student: Jose Escolano
Date of presentation: 23 Jul 2008
Title: Analysis and enhancement of Multiactuator Panels for Wave-Field Synthesis reproduction
Student: Basilio Pueo
Date of presentation: 2 Sep 2008
Title: Contributions to the Implementation of Wave-Field Synthesis Systems
Student: Sergio Bleda
Date of presentation: 3 Sep 2009
Title: Application of Sound Source Separation Techniques to Spatial Audio Systems
Student: Maximo Cobos
Date of presentation: 28 Sep 2009
Current PhD. Thesis
Title: Hybrid Methods for Room Acoustic Simulation
(co-supervised with Dr. J. Escolano)
Student: Juan Miguel Navarro
Title: Real-Time Sound Source Localization in Videoconferencing Environments
(co-supervised with Dr. M. Cobos)
Student: Amparo Marti
Title: Application of MIMO Inverse Filters for Room Compensation in WFS Reproduction
(co-supervised with Dr. A. Gonzalez)
Student: Laura Fuster
PhD Abstracts
Design and optimization of recursive digital filters by means of direct methods for audio applications
Author: German RamosThe equalization of sound systems presents several challenges compared with other signal processing systems. One of these particularities is the fact that the audio band covers a broad range of nearly 10 octaves between 20 Hz and 20 kHz. On the other hand the transducers employed are far away to be considered perfect or ideal. To equalize its frequency response, analog filters and digital FIR and IIR filters are employed.
This Thesis starts with a detailed study of the most relevant methods used in the digital inversion and equalization of audio systems. All methods have been implemented and tested in order to detect and evaluate their advantages and disadvantages from different points of view: order of the filter needed, computational cost, introduced delay, stability of the solution and success of the final implementation. By this way, a scientific, and at the same time practical view has been obtained in order to clarify how an audio equalization and cross-over system must be designed to be employed in live applications with high power audio equipments. The generic IIR filter design methods have convergence problems and do not equalize the lower frequencies as desired. FIR filters have also problems with the low frequencies due to its frequency resolution, and will require very high filter orders to obtain enough resolution if the lower frequencies need to be equalized.
To solve these commented problems, this Thesis faces the problem from a different point of view looking for a low computational cost and low delay solution (so IIR filters are employed). The filter is designed directly as a chain of second order sections SOS, where each SOS is a concrete filter type defined by its parameters (frequency, gain, and Q). A combination of a direct search method to obtain the initial values of the parameters of each SOS, with a heuristic optimization process of the parameters is done where constraints are imposed on the values of the parameters to guarantee the success of the final implementation. The design is done minimizing a cost function defined over logarithmic axes (in frequency and in magnitude) obtaining better resolution at the lower frequencies, following at the same time a psicoacoustic criteria. The SOS are designed in order, first the ones that more correction perform on the response, obtaining a scalable solution of the filter. The proposed method obtains excellent results on the task of passive loudspeaker equalization, and it is extended to be used in active audio systems where the filter must work also as a cross-over.
The results obtained have been compared with other methods objectively with measurements, and subjectively with audition tests. The proposed method achieves better results in all cases for the same computational cost. Finally, the digital filters have been implemented in software in real time on a DSP system, and on a PC computer, and on hardware inside a FPGA.
Contributions to Discrete-Time Domain methods in Room Acoustic Simulations
Author: Jose EscolanoIn this thesis some aspects of discrete-time domain methods are investigated:
Coupling methods: it exits some methods to couple DWM and the FDTD method. They allow a spatial distribution of the method used in each part of the whole simulation, in order to use the more appropriated method for each area of the simulated room; however they are based in the knowledge of where each method is used. It would be desirable to define a common interaction to obtain a blind coupling between methods; this means that each method does not need to know the nature of the coupled method. This would be investigated.
Frequency-Dependent Boundary Conditions: Although it exist a popular use of the DWM to room acoustic simulation due its inherent simplicity to couple digital filters representing boundary conditions, it is not clear that the results correspond with any standard impedance model. It will be investigated the behaviour of this method and would be proposed possible solutions. Besides, boundary conditions in the FDTD method should be proposed, in order to obtain an efficienter method than the classical use of the DWM. Others alternative FDTD discretization scheme should be investigated.
Considered sources in these methods are always monopoles and/or dipoles. Combination of these sources will allow more complex source directivities. However, this is not enough for real sources, which directivity diagrams use to be very complex. So, it is necessary to propose and validate a new algorithm that allows to define real directivity diagram pattern.