Filippo M. Fazi graduated in Mechanical Engineering from the University of Brescia (Italy) in 2005, with a master thesis on room acoustics. In the same year he joined the Institute of Sound and Vibration Research of the University of Southampton, initially as a master student and then as a doctoral student. He obtained his PhD in Acoustics in 2010, with a thesis on Sound Field Reproduction. He is currently a research fellow at the ISVR, after he has been awarded a fellowship by the Royal Academy of Engineering and by the Engineering and Physical Sciences Research Council. Dr Fazi has been working for the last 6 years on Acoustics, Audio technologies, Electroacoustics and Digital Signal Processing, with special focus on acoustical inverse problems, multi-channel systems (including Ambisonics and Wave Field Synthesis), virtual acoustics, microphone arrays and room acoustics. During his postdoctoral studies he has been involved in several national and international projects, in collaboration with industrial and academic partners, including nine months of research activity at the University of California San Diego. He has been awarded the ANC prize for best paper by a young person by the Institute of Acoustics in 2008 and the Sacerdote prize for best doctoral thesis by the Associazione Italiana di Acustica in 2010. He is a member of the Audio Engineering Society and of the Acoustical Society of America.
Research Projects
Electroacoustics, Virtual Acoustics, Imaging, and Inverse Methods
3D audio technologies for virtual reality
Systems for virtual reality are becoming increasingly relevant in a wide range of industrial applications. Such systems generally consist of audio and video devices, which aim at providing the user with a realistic perception of a three dimensional virtual environment. Ultimate 3D audio technologies are the focus of this project. New approaches to the reconstruction of three dimensional acoustic fields have been developed from rigorous mathematical and physical theories. These methods generally rely on the use of systems constituted by a multiple number of transducers (microphones and loudspeakers). The rapid development of multi-channel audio technologies and the new commercially available systems for digital audio processing provide ideal platforms enabling the application of the most innovative and computationally demanding sound field reconstruction theories. These devices should be controlled by algorithms that allow real time processing and enhanced user interaction. The crucial transition point from the theory to the practical implementation is at the core of this project. The latter is carried out in collaboration with the CALIT2 of the University of California, San Diego. The aim of the long term project proposed is to promote a scientific dialog and interaction between researchers from both institutions, with the goal of developing and realising audio technological solutions for Virtual Reality applications.
Application of compressed sensing to acoustical engineering
Compressed sensing (CS) is a novel signal analysis technique, which allows the acquisition and reconstructionof a signal utilizing prior knowledge on its structure (more specifically, its sparseness). This technique has been successfully applied to several engineering applications. Our primary aim is to demonstrate how CS theory can be applied to acoustic engineering problems: 3D sound recording, audio surveillance, room acoustic imaging, active sound control, and near-field acoustic holography. All these engineering applications are of much interest for industrial applications.
The project is carried out in collaboration with the CARlab of the University of Sydney.
Electroacoustical inverse problems
The proposed research is aimed at the application of the mathematical theory of inverse problems to the study of new possibilities arising in electroacoustics. The project includes both theoretical and experimental work to be undertaken in parallel, with constant input and feedback from industry, which is expected to provide fundamental guidelines for the research activity. The project proposed is also aimed at defining new technologies and to transfer this knowledge to national and international industrial companies operating in the fields of telecommunications, professional audio and virtual reality. The main objectives are:
1) To study new technological solutions for multi-channel audio applications, with special focus on the development of novel methods for the reproduction of a desired sound field and on the techniques for rendering multi-channel audio material.
2) To develop the theory of microphone arrays for sound recording and to develop technological applications for video conferencing.
3) To develop innovative solutions for audio systems for applications in Virtual Reality systems and to carry out technological research for the realisation of high-end multi-channel auralisation systems for the simulation of the reverberant acoustic field of concert halls and other real or virtual environments.
4) To apply the theory of inverse problems to DSP based control of large loudspeaker arrays for sound reinforcement.
This five-year project is supported by the Royal Academy of Engineering and by the EPSRC.
Generalized sampling theorem with Sperhical Harmonics
The generalized sampling expansion introduced by Papoulis allows the reconstruction of a band limited function sampled at a frequency lower than the Nyquist frequency using the outputs of several linear time-invariant systems. The work carried out in the framework of this project includes the formulation of the generalized sampling expansion for functions defined on the sphere, using spherical harmonics decomposition, thus facilitating sub-Nyquist sampling without aliasing error. An analysis of spherical convolution and the aliasing phenomenon in the spherical harmonic domain is undertaken, and examples are studied to of performance and limitations of the method. The project aims at the application of this technique to engineering problems such as the DSP for spherical microphone arrays.
Innovative microphone arrays with an unconventional geometry
The project involves the study of the underlying theory of microphone arrays, and consequently the formulation of a novel approach for array signal processing for microphone arrays with non-conventional shapes. This is also likely to involve the use of recently developed results on the application of the theory of integral equations to multi-channel DSP. The research starts from an idealized model of a continuous distribution of microphones, which should then be discretized in order to take into account the finite number of transducers. A further objective of the project involves the study of technological solutions for implementing the method developed in a practical system, with real-time signal processing. The research should involve the design and realization of an advanced array prototype with non-conventional geometrical arrangement. The project includes also experiments to evaluate the performance of the developed technology and of the prototype. The technology under study can be used in engineering applications such as acoustic holography and source identification as well as in audio and telecommunication applications such as multi-channel 3D sound recording and teleconferencing.
Perception of a virtual 3D acoustic space over headphones
The project investigates and develops techniques for the simulation of 3D virtual sound environment through headphone reproduction. The DSP system recreates a spherical virtual array of loudspeakers, combining Ambisonic techniques and binaural audio processing. In the initial phase of the project, depersonalized Head Related Transfer Functions are used. The research is also aimed at developing strategies to overcome one of the most common problems in headphone reproduction, namely the internalization of the virtual sound source. Some techniques are investigated, mainly head tracking and the use of virtual reverberation. Subjective listening experiments are required in order to validate the results obtained.
Perceptual models for sound field analysis and synthesis
Recent work undertaken in ISVR (Park et al Acustica 94, (2008), 825-839) has led to the development of an auditory signal processing model that gives significant accuracy in the prediction of human localization of acoustic sources in the horizontal plane. The models are based as far as possible on the known behaviour of the human auditory periphery (outer and inner ears, cochlea, inner hair cells, auditory nerve), a physiologically plausible model of binaural processing, and a central processor based on a pattern matching hypothesis.
The overarching goal of this study is the extension of the human perceptual model to be able to localize the position of a sound image in 3D space and deal with other auditory signal processing tasks, such as “cocktail party processing”. Like the 2D model, the EI cell pattern matching approach could be used to identify the spatial locations of individual sources and thus influence the binaural processing system in the extraction of individual source signals.The model would find application in the design of novel multimedia/auditory displays and multichannel audio rendering techniques which are of interest in both domestic and professional contexts.
This project is carried out in collaboration with Meridian Audio.
Publications
List of publications
Other links
Fluid Dynamics and Acoustics Group
Virtual Acoustics and Audio Engineering
Email: fmf205@soton.ac.uk
Database Updated by: S J Brindle on 23 Jun 2011
Dr Filippo Fazi