Thibaut Carpentier

Towards interactive authoring tools for composing spatialization

We present interactive tools designed to help music composers controlling sound spatialization processes in a computer-aided composition environment. We conducted interviews with composers to understand their needs and inform the design of new compositional interfaces. These interfaces support quick input, visualization and edition of three-dimensional trajectories, as well as the control of the temporal dimension in spatial scene descriptions.

Holophonic Sound in IRCAM's Concert Hall: Technological and Aesthetic Practices

This article presents a report on technological and aesthetic practices in the variable-acoustics performance hall, Espace de Projection, at the Institut de Recherche et Coordination Acoustique/Musique. The hall is surrounded by a 350-loudspeaker array for sound-field reproduction using holophonic approaches such as wave-field synthesis and higher-order Ambisonics. First we present the design and implementation of the audio system and discuss the challenges of both hardware and software architectures. This is followed by a discussion of spatial composition techniques, aesthetic approaches, and methodologies for composing computer music for high-density loudspeaker arrays, explored through the paradigmatic examples of pieces produced by two artist-in-research residencies.

Perceptual control of convolution based room simulators

Reverberation processing has been intensively studied in audio and acoustics research for many years now. Early approaches used feedback delay networks to control the temporal distribution of reflections and to simulate the statistical properties of room reverberation. Thanks to the increase in processing power and the development of low-latency convolution algorithms a new generation of reverberation processors has been developed. They apply room impulse responses (RIR) measured in real concert halls and thus guarantee naturalness and authenticity of reverberation. Extending this approach to the use of higher-order spherical microphone arrays provides the means for analyzing the spatiotemporal distribution of acoustic energy. This space-time-frequency representation of the acoustic wave field is also referred to as directional room impulse responses (DRIR) in literature. The objective of the presented work is to develop a perceptually motivated signal-processing environment based on the analysis and re-synthesis of DRIRs. It first extracts perceptual features from measured DRIRs (e.g. source presence and listener envelopment) and thus provides a perceptual signature of the measured room. The room acoustic behavior can then be modified along the various perceptual dimensions, preserving the microstructure of the original RIRs, before being re-synthesized for the use with reverberation processors.

Comparison of Different Egocentric Pointing Methods for 3D Sound Localization Experiments

This study evaluates several methods for reporting the perceived location of real sound sources. It is well known that the method used for collecting judgments in auditory-localization experiments has a strong influence on the accuracy of a subject’s response. Previous works on auditory-localization tasks revealed that egocentric pointing methods (which are based on a body-centered coordinate system) allow for more accurate judgments than verbal reporting or exocentric pointing techniques (which are based on a 2D or 3D reporting device). Three different egocentric methods are compared: the most commonly applied “manual pointing” and “head pointing” methods, and the “proximal pointing” method, which forces the participants to indicate the apparent direction by pointing in the proximal region of the head with a marker held at the fingertips. The two first methods involve a rotation of the body of the participant, whereas the third method only involves movements of the arm(s) and hand(s) with a fixed head. Sound stimuli were presented randomly over 24 loudspeakers that were uniformly distributed on the upper hemisphere around the subject. The merits of the different methods are compared and discussed with regard to localization errors and to practical considerations. Although they show similar trends, each of the different methods affects the pointing accuracy in a specific way. The proximal pointing method, for example, is more accurate for sources located at high elevation angles. However, at rear locations close to the median plane an increased bias appears due to difficulties in performing the motor task to reach these positions. The proximal pointing method shows faster response times, which may be advantageous when planning 3D sound localization experiments.