It is hypothesized that sound quality metrics, particularly loudness, sharpness, tonality, impulsiveness, fluctuation strength, and roughness, could all be possible indicators of the reported annoyance to helicopter noise. To test this hypothesis, a psychoacoustic test was recently conducted in which subjects rated their annoyance levels to synthesized helicopter sounds [Krishnamurthy, InterNoise2018, Paper 1338]. After controlling for loudness, linear regression identified sharpness and tonality as important factors in predicting annoyance, followed by fluctuation strength. Current work focuses on multilevel regression techniques in which the regression slopes and intercepts are assumed to take on normal distributions across subjects. The importance of each metric is evaluated one-by-one, and the variation among subjects is evaluated using simple models. Then, more complete models are investigated, which include the combination of selected metrics and random effects. While the conclusions from linear regression analysis are affirmed by multilevel analysis, other important effects emerge. In particular, a random intercept is shown to be more important than a random slope. In this framework, the relative importance of sound quality metrics is re-examined, and the potential for the modeling of human annoyance to helicopter noise based on sound quality metrics is explored.

No abstract

In acoustical spaces, room acoustics parameters are often predicted using energybased geometrical acoustics. For smaller rooms, interference among coherent reflections is taken into account by phased geometrical acoustics, which improves results for lower frequencies. The use of a spherical wave reflection coefficient improves the results further, yet the impact on room acoustics parameters is not fully known. This work focuses on the differences in predicted reverberation time when using plane or spherical wave reflection coefficients. The differences are analyzed for a variety of boundary conditions, including nonuniform distributions of absorption, in medium-sized rooms using a phased image source model. Since calculated differences are greater than the conventional just-noticeable-difference of 5% for reverberation time, a laboratory listening test is performed to confirm audibility of the modeled differences. Two narrow band noise stimuli (octave bands with central frequency 125 and 250 Hz) with a duration of 1 s were used for comparisons of 18 acoustic scenarios by means of a three-alternative forced choice method (3AFC). More than half of the listeners could hear the differences in all 36 cases. Statistically significant results (chi-squared test was used) were found in two thirds of the cases, corresponding to those with longer reverberation times.

The multimodal radiation from the open end of a cylindrical waveguide with arbitrary wall thickness is solved by deriving algebraic solutions of the radiation impedance matrix, without restrictive hypothesis on the frequency range. The basic idea of the method is to turn the original unbounded problem into the problem of a cylindrical waveguide embedded in an infinite waveguide with an annular perfectly matched layer (PML) on its wall. Then, using a multimodal formalism of the guided wave propagation and a complex coordinate stretching PML, algebraic expressions are derived for the continuity and radiation conditions in this coupled system.

Geometrical acoustics provides a correct solution to the wave equation for rectangular rooms with rigid boundaries and is an accurate approximation at high frequencies with nearly hard walls. When interference effects are important, phased geometrical acoustics is employed in order to account for phase shifts due to propagation and reflection. Error increases, however, with more absorption, complex impedance values, grazing incidence, smaller volumes and lower frequencies. Replacing the plane wave reflection coefficient with a spherical one reduces the error but results in slower convergence. Frequency-dependent stopping criteria are then applied to avoid calculating higher order reflections for frequencies that have already converged. Exact half-space solutions are used to derive two additional spherical wave reflection coefficients: (i) the Sommerfeld integral, consisting of a plane wave decomposition of a point source and (ii) a line of image sources located at complex coordinates. Phased beam tracing using exact half-space solutions agrees well with the finite element method for rectangular rooms with absorbing boundaries, at low frequencies and for rooms with different aspect ratios. Results are accurate even for long source-to-receiver distances. Finally, the crossover frequency between the plane and spherical wave reflection coefficients is discussed.