Sharpness

Sharpness is an attribute that is related to the Timbre of a sound. Sharpness is the subjective attribute describing the perception of the spectral balance of a sound. It is the perceptual equivalent of the spectral centroid.

Sounds with a major part of their energies located in the low frequencies have low sharpness values, while sounds with louder high-frequency contents have high sharpness values.

Sharpness depends on the shape of the spectral Envelope, the frequency distribution of the energy and the loudness. The unit of sharpness is the acum. One acum is defined as the sharpness of a narrow band of noise (one Bark) centered at one kHz at 60 dB SPL.

The Zwicker & Fastl sharpness model (1999) is basically a weighted barycenter of the specific loudness over the bark scale (which is the perceptual equivalent of the spectrum). The model also includes a weighting function g(z) giving more impact to high frequencies (typically above 16 Barks, roughly above 3000 Hz).

Where N’ is the specific loudness (or loudness density), z is the bark band number, and N is the total loudness. As for loudness, this model is suitable for stationary or near-stationary sounds.

Two sharpness models are supported in Ansys Sound: Analysis and Specification. Both models rely on Zwicker and Fastl's sharpness formula above, and only differ in how the function g(z) is calculated:

  • Sharpness

    and

  • DIN45692 Sharpness

    and

The difference in the g(z) weighting function between the Sharpness and DIN45692 Sharpness models is shown in the figure below:

Figure 1. Weighting function g(z) for Sharpness and DIN45692 Sharpness

Sharpness vs Time indicators are also supported. For these indicators, Ansys Sound: Analysis and Specification does the calculation at every 2 ms of the signal and a lowpass filter with a 8-Hz cutoff frequency is then applied to smooth out sharpness values over time. (See Value vs Time and Indicators for Non-Stationary Sounds.)