Ole Kirkeby

Fast deconvolution of multichannel systems using regularization

A very fast deconvolution method, which is based on the fast Fourier transform (FFT), can be used to control the outputs from a multichannel plant comprising any number of control sources and error sensors. The result is a matrix of causal finite impulse response filters whose performance is optimized at a large number of discrete frequencies. The paper is particularly aimed at multichannel sound reproduction and more specifically reproducing the sound field from a set of loudspeakers.

Reproduction of plane wave sound fields

The problem of reproducing a desired sound field in space, not just at a number of discrete points, but over a continuous two‐dimensional area, is investigated. In theory, any sound field can be reconstructed perfectly in a given region by using a continuous monopole/dipole layer, but this is obviously not possible in practice. This paper attempts to give some quantitative measures of the extent to which a given sound field can be reproduced by using a number of discrete monopole sources. Some of the physical limitations that apply to any sound reproduction system are illustrated by studying a simple model. The desired sound field is a plane wave, the sources are ideal monopoles in a free field, and the optimal source accelerations are calculated using the traditional least‐squares method. All calculations are undertaken in the frequency domain, and three different loudspeaker arrangements are studied. The results clearly demonstrate that the quality of the reproduced sound field is mainly determined by t...

Local sound field reproduction using digital signal processing

This work shows how an acoustic wavefront can be reconstructed locally by using only a few loudspeakers. The loudspeaker inputs are calculated by passing a set of signals recorded by only a few microphones through a matrix of causal digital filters having finite impulse responses. These filters, referred to as the inverse filters, are calculated by inverting (in the least‐squares sense) a matrix which contains the electroacoustic transfer functions from the loudspeakers to the microphones. In practice, it is crucial to use a modeling delay and a regularization factor in order to achieve an accurate inversion. The technique is illustrated with an example that shows how well four loudspeakers can reproduce a sound field that has been recorded with three microphones. When the recorded field does not contain energy at frequencies whose acoustical wavelengths are shorter than the distance between adjacent microphones, the original field is reproduced remarkably accurately in the vicinity of the microphones reg...

Local Sound Field Reproduction using Two Closely Spaced Loudspeakers

When only two loudspeakers are used for the reproduction of sound for a single listener, time domain simulations show that it is advantageous that the two loudspeakers are very close together. The sound field reproduced by two loudspeakers that span 10 degrees as seen by the listener is simpler, and locally more similar to the sound field generated by a real sound source, than that reproduced by two loudspeakers that span 60 degrees. The basic physics of the problem is first explained by assuming that the sound propagates under free-field conditions. It is then demonstrated that when the influence of the listener on the incident sound waves is taken into account by modeling the listener’s head as a rigid sphere, the results are qualitatively the same as in the free-field case. Consequently, two closely spaced loudspeakers are capable of accurately reproducing a desired sound field, not only at the ears of the listener but also in the vicinity of the listener’s head. This result, although counter-intuitive...

Resolution of front–back confusion in virtual acoustic imaging systems

A geometric model of the scattering of sound by the human head is used to generate a model of localization cues based on interaural time delay (ITD). The ITD is calculated in terms of the interaural cross-correlation function (IACC) for sources placed at a series of azimuthal angles in the horizontal plane. This model is used to simulate the pressures generated at the ears of a listener due to real sources and due to a two-channel and a four-channel virtual source imaging system. Results are presented in each case for the variation of ITD with head rotation. The simulations predict that the rate of change of the ITD with head rotation produced by a real source and replicated by the four-channel virtual source imaging system, cannot be replicated by the two-channel system. These changes to the ITD provide cues which allow resolution of front-back confusion. The results of subjective experiments are also presented for the three cases modeled. These results strongly support the findings from the modeling work indicating that, for the systems described here, front-back confusion is resolved through changes to the ITD arising from head motion.

A multiple microphone recording technique for the generation of virtual acoustic images

A new recording technique based on multichannel digital signal processing is suggested. The system uses a dummy-head that is modeled as a rigid sphere with two pairs of microphones mounted on opposite sides of the sphere in the horizontal plane. Reversals—front back confusion, is a well-known phenomenon when localizing virtual acoustic images produced by either headphones or loudspeakers. Reproduction with two loudspeakers to the front of the listener causes rear virtual acoustic images to be perceived primarily at “mirrored” angles in the frontal hemisphere. The problem is tackled here by using a multichannel signal processing technique rather than by mimicking accurately the acoustomechanical properties of a human head. The acoustic signals which are recorded at the microphones are filtered by a 4×4 matrix of digital filters before being transmitted via four loudspeakers. The performance of the system is investigated by means of computer simulations, objective measurements, and also by subjective experiments in an anechoic environment, where the listeners are asked to localize the perceived angle of the signals which were prerecorded with the sphere dummy-head. Successful discrimination of reversals is achieved primarily due to the dominant role of the interaural time delay (ITD) for localization at low frequencies, but the accuracy with which listeners can localize virtual acoustic images is reduced in comparison to a conventional two-ear dummy-head (e.g., KEMAR) with a two-loudspeaker arrangement. The system is robust with respect to head rotations—virtual acoustic images do not disappear and localization ability improves when listeners use small head rotations.

SIMULATION OF THE TRANSFER FUNCTION FOR A HEAD-AND-TORSO MODEL OVER THE ENTIRE AUDIBLE FREQUENCY RANGE

In this study, a method for simulating the transfer function of a head-and-torso model over the entire audible frequency range is introduced. The simulation method uses the ultra-weak variational formulation (UWVF) which is a finite element type method tailored for wave problems. In particular, the UWVF uses plane wave basis functions which better approximate the oscillatory field than a polynomial basis used in the standard finite element methods (FEM). This leads to reduction in the computational complexity at the high frequencies which, accompanied with parallel computing, extends the feasible frequency range of the UWVF method. The accuracy of the new simulation tool is investigated using a simple spherical geometry after which the method used for preliminary HRTF simulations in the geometry of a widely used head-and-torso mannequin.

Subjective investigation of a new sound reproduction system (stereo dipole)

Recently, a new transaural sound reproduction system, referred to as a ‘‘stereo dipole’’ (SD) reproduction system, has been proposed, in which a closely spaced pair of loudspeakers is located in front of a listener. Theoretical investigation was carried out by computer simulations and it was reported that the SD system would be capable of providing a relatively large equalized area to the listener and robustness with respect to a listener’s head movement compared with those given by the standard loudspeaker arrangements. This paper deals with a subjective investigation of the SD system. It is known that the ordinary transaural system has a disadvantage in the quality of reproduced sound; that is, the reproduction of unnatural and antiphase virtual sound from the psychoacoustic point of view. Therefore, subjective listening tests were carried out, which aimed to compare the SD system with the ordinary system. The significance of the SD system on reproduced sound perception is suggested.

Acoustics and Hearing

This article reviews Acoustics and Hearing by Peter Damaske , Berlin, Heidelberg, 2008. 120 pp. Price:

Design of a filter matrix used for stereo dipole transaural systems

79.95 (paperback). ISBN: 978-3-540-78227-8