The metric was initially developed for quantifying the degree of granularity in idealized modulated sound as perceived by a person with normal hearing, which is
similar to the effects of rumble. However, the calculations are performed on both
sound pressure and vibration data obtained in this experimental study. In the latter
case, we get a measure of the apparent roughness as opposed to the classical meaning
of roughness. Fig. 11a and b consists of two bar charts comparing the classical and
apparent roughness levels of the measured sound pressure and vibration signatures
respectively in the noisy and quiet vehicles. From Fig. 11a, the roughness metric
consistently distinguishes the intensity of rumble response measured by microphones
A±H and the binaural acoustic head in the front seat of the two vehicles. The calculations using the data from the binaural acoustic head in the rear seat are not as
conclusive. This is due to the limitation of the existing roughness algorithm, which
was primarily developed for idealized sound. Hence, it may not work as well in the
presence of lower signal to noise ratios due to a higher relative masking component
and clutter, which can also contribute to the general sensation of roughness (from
the randomness in the data) as detected by the algorithm. However, the overall
trends are consistent and suciently accurate to justify the use of this technique in
spite of its restriction. In the case of the measured vibration data as shown in Fig.
11b, the apparent roughness of the noisy vehicle is clearly higher in amplitude than
the quiet one. The predicted overall trend matches well with the subjective assessment too. Similarly, it also shows a high level of rumble transmission across the rear
and left engine mounts. In addition, other types of existing time-varying-based
metrics such as loudness, ¯uctuation strength, etc. were also applied during this
study, but since they did not produce a clear trend, none is presented here.