Loudness Compensation

Ever noticed that "Loudness" button on your music system, home theatre, on car audio system. Do you know what it does?

"Loudness compensation" is the correction required during playback of an audio source that is reproduced at a 'Listening level' different from its 'Mastering level'. The intent of this compensation is to ensure that the spectral balance of the audio source is same as that intended by the artist (or the composer), when played back at any volume level.

Put differently, the essence of "Loudness Compensation" is to maintain the perceived spectral balance of audio content irrespective of playback volume level. When in use, loudness compensation corrects for our reduced sensitivity at bass and treble (very low and high frequencies regions) when playing back at a low volume levels. For instance, while watching a movie at low volume levels late at night, you might have noticed the sound becomes "dialogue only" and other environmental sounds like footsteps, thunder are not audible. The need for this correction in an audio playback system arises due to the inherent non-linearity in the human aural perception. This fact has been established in Equal Loudness Contours (ELC), which is standardised in ISO 226:2003.

I formulated a closed form expression for loudness compensation starting from the equations in ISO 226:2003 standard. However, to achieve an ideal loudness compensation would require a computationally complex implementation which is not feasible on resource constrained hand held devices.

Next, I proposed an efficient approximation for loudness compensation by quantifying the trade-off between computational complexity and the approximation error.

Here, the below graph shows the proposed loudness compensation across frequencies as a function of user applied attenuation.

Reference: Pradeep Prasad, "A low complexity approach for Loudness Compensation", 129th Audio Engineering Society (AES) Convention, vol. 1, pp. 310-319, (no. 8263) Nov 4-7, 2010, San Francisco

Please find a copy of the published paper below