Difference Thresholds for Timbre related to Spectral Centroid
| Many studies have found a mapping between the spectral center of gravity and timbre perception. | |
| In our own work, spanning some 10 years, we have found a consistent relationship between the main perceptual dimension of timbre and spectral distribution measures for continuant (as opposed to impulse) signals. | |
| This research included musical contexts, amusical contexts, natural instrument signals, and various emulated and synthetic signals. | |
| It appears to us that if we are to continue to use spectral centroid as a basic measure, it is necessary to understand its psychophysical properties with some measure of clarity and detail. | |
| Spectral tilt, total harmonic distortion, and spectral centroid are all suitable for making this mapping, however, spectral centroid has been employed recently by a number of researchers, and is the subject of the present experiments. | |
| Spectral centroid is
variously defined, so we pause here to clarify. Some
researchers have called the centroid spectral
brightness, an obvious confusion between frames of
reference, for spectra are neither bright nor dark. |
|
| The following formulation
is used in the present experiments:
|
|
| Note
that this is a unitless measure since the denominator is
weighted by the fundamental F1 Centroids
calculated in this manner are directly comparable. |
|
| Some researchers remove the weighting, producing a centroid value in Hz. In this case, a given spectrum with a fundamental at 1 kHz would have a lower centroid than its equivalent at 2 kHz, for example. |
 |
 |
||
| Previous Page | Next Page |