A New Comprehensive Approach of Surround Sound Recording

Many techniques have been developed concerning surround sound recording and the issue has turned out to be a challenge without a comprehensive theory. This paper presents an approach based on a full 3D acoustic field theory and the use of a 3D microphone array. Our research work has lead to a new spatial digital processing technique which allows the use of freely positioned capsules of any type such as omnidirectionals, bidirectionals, or cardioids. This technique can be seen as an extension of full sphere, generalized Ambisonics which provides a spatial resolution never achieved before, but requires neither high order directivity capsules nor assumes that all capsules are coincident. The theory has been validated with a full sphere 3rd-order prototype using 24 omnidirectional capsules. The theory also allows for a 5th order spherical harmonic, multichannel 5.0 microphone. 5717 LABORIE ET AL. A NEW COMPREHENSIVE APPROACH OF SURROUND SOUND RECORDING 0 INTRODUCTION Spatial sound recording has been an issue for a long time. Beginning with stereo, important steps have been made. However, high spatial resolution sound recording is still a challenge [9, 14, 18] mainly because of the lack of good high order elementary microphones and of a comprehensive theory [20]. Fundamental acoustics has been understood for a long time but has long been unused. It does not currently offer ready-to-use tools except for Wave Field Synthesis. Surround sound reveals the need for further application of this theory. In the early 70’s, Michael Gerzon [8] developed Ambisonics, a recording technique based on the encoding of the incoming direction of the sound within the sound itself [1, 2, 10, 13]. This approach considers the three dimensional sound field as a source distribution around a central point and was an innovative extension of stereo in the same spirit as its inventor, Blumlein. Ambisonics allows separation of the recording stage from the reconstruction stage via the so-called encoding and decoding processes. The Ambisonic microphone contains four cardioid capsules which deliver the A-format. An encoder transforms these signals to the B-format [7]. The B-format signals have the advantage of being independent of the characteristics of either the recording or the reconstruction systems, contrary to stereophonic, quadraphonic or today’s other multichannel schemes. During the reconstruction stage, a decoder generates the loudspeaker’s feed signals from the B-format. The decoder depends on the loudspeaker’s configuration and several decoders have been designed for various configurations, such as quadraphony or 5.1 multichannel format. The spatial resolution is limited, however, to first order spherical harmonics because the capsules used to record the sound are first order. More precisely, the tetrahedral microphone array known as the SoundField microphone, patented by Gerzon, has cardioid capsules [4]. Many extensions and developments concerning Ambisonics have been studied to reach higher orders [5, 6, 11, 12, 15, 16] but these extensions have been limited by the fact that there do not exist any higher order microphones with adequate characteristics. Moreover, the theory assumes that all microphones are physically at the same point, which leads either to imperfections or to requiring compensation of the signals. In a recent article [17], M. A. Poletti developed a new theory for a circular microphone array using only omnidirectional or cardioid capsules which leads to a technique that does not require that the capsules be all at the same point and also provides high order directivity in the horizontal plane. He achieves this by no longer considering a source distribution around the origin, but rather by considering the physical sound field in a horizontal plane near the origin. This has the advantage of not requiring the knowledge of high order derivatives of the sound field at the origin to obtain high order directivity in the horizontal plane. This theory still has some drawbacks. The capsules are necessarily on a circle, and omnidirectional capsules can not be used because they lead to infinite gains. Moreover, the capsule noise is not taken into account and could be amplified during the encoding process. Finally, this theory uses a two dimensional model, which constitutes a significant departure from reality. In this paper, a new spatial recording technique based on a three dimensional theoretical approach is presented. The goal of this technique is to extract and deliver as much spatial information as possible from the set of signals delivered by a microphone array. In order to do this, we use fundamental acoustics [3, 19] and develop new tools to apply it to spatial sound recording. This leads to a sound representation in a form compatible with full sphere, high order, generalized Ambisonics. However, this work is not derived from Ambisonics, as we do not consider the source repartition around the origin. This work can be seen as an extension to Ambisonics that overcomes its limitations. Our technique makes it possible to obtain high order Ambisonic signals from an array of capsules which can be of any type (omnidirectional, bidirectional, cardioid, etc.) and can have any position and orientation in space. Nothing is presumed a priori about the position of the capsules, so they are not necessarily on a plane or on a sphere. The resulting microphone consists of an array of capsules and a processing system based on a DSP or a PC. This paper will first present the theoretical principle including the spatial sampling and encoding aspects. The practical results we have arrived at will then AES 114 CONVENTION, AMSTERDAM, THE NETHERLANDS, 2003 MARCH 22–25 2 LABORIE ET AL. A NEW COMPREHENSIVE APPROACH OF SURROUND SOUND RECORDING