I was playing with the idea of performing integration on a sound signal, since I've never heard of where it'd be used.

Is it used somewhere?

Here's something:

  • 2
    $\begingroup$ yeah, it's done. maybe a leaky integrator (which is a one-pole LPF) would be more common because integrating DC is a problem. i've used this at the front-end of a pitch detector. works great with floating-point. if fixed-point, you might wanna put some AGC in it. $\endgroup$ – robert bristow-johnson Feb 5 '16 at 0:33

So, a sound signal $s$, in the DSP sense, is but a sequence of real-valued digital numbers:

$$ s = (1, 232, -12, 121, 0, 131, \dots) $$

Integrating that is nothing but adding these up, so the integrated signal $r$ would have the formula:

$$ r[n] = \sum\limits_{i=0}^n s[n] $$

or, writing that recursively, is the signal that it was at the last sampling time, plus the next input sample:

$$ r[n] = r[n-1] + s[n]\quad \text{ for }r>0\\ r[0] = 0$$

Now, that is what DSP people call a recursive system, or in this specific case, a single-pole IIR filter (IIR: infinite impulse response, because you send one pulse in, you see it in the output forever). It's transfer function $H(z)$ is

$$H(z) = \frac{1}{1+a\cdot z^{-1}} $$

Your $a=1$, here, because the last value of $r$ doesn't get multiplied with a dampening factor. You will find a lot of literature out there explaining the properties of IIR filters, but for all practical things, you could build a low-pass filter out of your integrator, if you use a factor $a<1$.


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.