Abstract

Sound analysis and synthesis for music requires accurate analysis and high-quality synthesis of sounds. Some of the methods that were developed at the Institut de Recherches et de Communication Acoustique/Musique (IRCAM) for contemporary music research and production are described. The following analysis techniques have been implemented and used: FFT and phase-vocoder, wavelets, linear prediction algorithms (Levinson, global covariance, Burg, lattice covariance, lattice recursive), prony analysis, pitch detection, temporal envelope tracking (power tracking), detection of harmonic/noisy components, modal analysis (resonance modeling), formant-waveforms (FOF) coding, and formant trajectory coding. The phase-vocoder is used extensively for analysis and processing (filtering, cross-synthesis, time warping, etc.) of speech and music. Different LPC analysis algorithms have been studied and are used for precise estimation of spectral envelopes. These envelopes can be coded into (as well as recalculated from) formant parameters without loss of precision. Sounds and LPC residuals are coded into a harmonic/noise index for each of the frequency bands centered around harmonic partials of the signal, thus improving the quality of synthetic sounds. For sound processing, the following algorithms have been used: cascade filtering (LPC-vocoder), parallel filtering (in the program CHANT), frequency domain filtering (phase-vocoder), time-warping (phase-vocoder and LPC-vocoder), time-varying and arbitrary sample-rate conversion. Several synthesis methods have been developed, including LPC-derived synthesis filter, formant synthesis, formant-waveform (CHANT), parallel filtering, phase-vocoder, and additive synthesis. Synthesis-by-rule uses sophisticated object-oriented languages (PREFORM) for calculation, manipulation, and real-time control of synthesis parameters. For these studies, a real-time work-station is under development, based on SUN-3, Mercury array processor, and Sony PCM AD/DA.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call