0
$\begingroup$

Hello I am working on audio watermarking in MATLAB and am trying to choose specific frames of the audio signal where the watermark will be embedded. Through research, I have read that the best way to do this is to choose parts/samples of the audio signal that are the loudest, or above a certain threshold. By using wavread in MATLAB, the result is an array of the audio data in type double precision, which represent the sound pressure at a given sample. My question is, since the numbers represent sound pressure, will the loudest portions of the audio signal be the largest numbers? If not, is there another way to achieve what I am trying to do? Thank you!

$\endgroup$
1
$\begingroup$

Loudness is defined as human perception of volume. In general, there is a very good correspondence between the two. Therefore, you can often use one as a proxy for the other.

In this case, you can easily measure volume: the portions of the audio signal with the highest signal intensity (or volume) correspond to the parts with the greatest absolute value (or, if you prefer, square value). With a few assumptions, you can say that the loud sections are also the sections with high volume.

$\endgroup$
  • $\begingroup$ Okay yes I should of worded it as volume or greatest sound pressure. But that is what I thought I would be able to do regarding the greatest absolute value. Im happy to hear that this would work. This may be a stupid question but why are the smallest negative values also considered to be of high volume? What do negative values represent in this context compared to positive numbers? $\endgroup$ – Math244 Jul 10 '13 at 21:34
  • $\begingroup$ in simple terms, sound is fluctuations in air pressure. Both positive and negative numbers are excursions from the mean air pressure, negative represents reductions in pressure, positive represents increases. $\endgroup$ – Bjorn Roche Jul 11 '13 at 0:55
1
$\begingroup$

As a general rule the parts of the signal containing the largest numbers will correspond to the parts of the signal with the greatest perceived loudness.

When you inspect individual samples within the signal you are measuring the instantaneous peak amplitude - this is useful but not the most robust way to determine perceived loudness. A better approach is to measure the power or energy of the signal - for example by calculating the RMS Power.

To demonstrate the difference consider signal with 1000 samples, in which 999 samples have a value of 0.0 and one sample has a normalised value of 1.0 - the peak amplitude of this signal is very high, whereas the power of this signal is extremely low.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.