Saturday, October 1, 1:30 pm — 3:00 pm (Rm 409B)
Leslie Gaston-Bird, University of Colorado Denver - Denver, CO, USA
P21-1 Automatic Design of Feedback Delay Network Reverb Parameters for Impulse Response Matching—Jay Coggin, University of Miami - Coral Gables, FL, USA; Will Pirkle, University of Miami - Coral Gables, FL, USA
Traditional reverberation algorithms generally fall into two approaches: physical methods, which involve either convolving with room impulse responses (IRs) or modeling a physical space, and perceptual methods, which allow the use of practically any reverberation modules in various combinations to achieve a perceptually realistic reverberation sound. Perceptual reverberator algorithms are typically “hand tuned” where many of their parameters are found empirically. In this paper we present an automatic method of matching Feedback Delay Network parameters to real room impulse responses so that we may produce computationally efficient reverberation algorithms that perceptually match linear convolution with the target room IRs. Features are extracted from the target room IR and used to guide a Genetic Algorithm search to find the reverberator parameters.
Convention Paper 9666 (Purchase now)
P21-2 Perceptually Alias-Free Waveform Generation Using the Bandlimited Step Method and Genetic Algorithm—Francisco Valencia, codigoriginal - Medellin, Colombia; Samarth Behura, University of Miami - Coral Gables, FL, USA; Will Pirkle, University of Miami - Coral Gables, FL, USA
Quasi-Bandlimited waveforms may be synthesized by smoothing the discontinuities of trivial waveforms using the Bandlimited Step Method (BLEP) that produces excellent results with low computational overhead . The correction scheme first starts with a sinc function—the impulse response of a low-pass filter—and uses it to generate offset values that are applied to the points around the discontinuity. Windowing the sinc function prior to use is found to reduce aliasing at the expense of the harmonic envelope, whose shape is no longer ideal. In this paper we explore two methods for generating the initial sinc function in an effort to achieve a perceptually alias-free waveform. Both approaches involve using the Genetic Algorithm as a search method.
Convention Paper 9667 (Purchase now)
P21-3 Preserving Reverberation in a Sinusoidally Modeled Pitch Shifter—Sarah R. Smith, University of Rochester - Rochester, NY, USA; Mark F. Bocko, University of Rochester - Rochester, NY, USA
Many pitch shifting algorithms suffer when the signal contains reverberation. In general, it is possible to preserve the spectral envelope of the original sound, however, an appropriate phase response can only be estimated for minimum phase systems such as vocal formants. This paper presents a pitch shifting algorithm that preserves the reverberant qualities of the original signal by modifying the instantaneous amplitude and frequency trajectories of a sinusoidal model. For each overtone, the sinusoidal trajectories are decomposed into correlated and uncorrelated components and a deviation spectrum is calculated. To synthesize the modified sound, the uncorrelated components are adjusted to preserve the deviation spectrum. The resulting trajectories and sounds are then compared with those of a standard pitch shifter.
Convention Paper 9668 (Purchase now)