[edit: if you read my previous notes here, ignore them. I was having a brain malfunction; what I originally said is correct AFAICT, based on my reading of the standard described in NMRA S-9.1. Therefore, my original message below stands. Some of what I said in my edits would have been idiotically wrong even if that hadn't been the case.

]
I've seen this said a few times, and as I understand it, it's incorrect. While it is not sinusoidal AC like what you expect from a wall outlet, it is
definitely an AC signal. The sine wave isn't actually required for it to be AC.
As to the negative voltage issue, that's a matter of reference; there is no such thing as "absolute ground" or "absolute 0V" in this context, and thus whether a voltage is positive or negative is entirely relative to what you're measuring it against. This is one thing that took me a
long time to come to terms with when it came to electricity and electronics.
What's important -- what makes it an AC signal -- is that the current flow reverses direction periodically.
If I understand the protocol correctly, then in a DCC system, it does that. During one phase, where (for example) the left rail is at 12V and the right rail is at 0V, current flows from the left rail to the right rail. In the next phase, the situation is reversed, and the current is flowing from the right rail to the left rail. This continually alternates, hence alternating current.
[Aside: even the direction of current flow between positive and negative is apparently an arbitrary assumption, though; there are multiple schools of thought on which direction current is actually flowing between positive or negative, or so said my electronics 101 teacher. This Is A Dark Art, IMO.]
So it's completely correct to say that DCC is an AC signal. It's just not a 60Hz sinusoidal one.
Caveat: I've had only a single college-level course on the topic and a bit of personal circuit design experience, and that's it, so if someone with a degree in this stuff wants to chime in and correct me, feel free. And if that's you,
@ctclibby, please explain where I got it wrong, because I definitely want to know!
This is also my understanding. The signal to the decoder is carried in the timing of the pulses (or if you prefer, the timing of the direction switches - it's a matter of taste how you think of it), similar to "typical" PWM signals, for example to control a servo.
The IMPORTANT difference is that the PWM signal to a servo is an on/off DC signal: sometimes the voltage is +5v (or +3V3, depending on the system), other times it is 0V, i.e. the +ve and -ve leads are at the same voltage.*
If DCC were implemented that way, the voltage delivered to the track would be constantly fluctuating between 0V and +24V, which obviously is not good for powering things. Instead, DCC provides a constant voltage either in one direction or in the other, i.e. an AC signal, which is then rectified by the decoder to provide constant DC power. This allows both power and signal to be delivered by a single pair of wires. (The oscilloscope pictures are a bit misleading here.)
Bottom line: DCC uses AC power, with current always flowing in one direction or the other.
----
As to the other stuff insomniville said, I too had the same challenges. I took S Level physics in England (comparable to a first year college course if you're left-ponded, I think) and it was only many years later that I grokked that voltage is always measured relative to some reference point, usually the -ve terminal of the battery or a DC power supply or the GND pin on an Arduino, Pi, etc. This is why when you are building electronic circuits, all of the components need to be connected to the same ground, i.e. they all agree on what 0V is. And that's why you'll sometimes see circuits refer to it as "common", i.e. all the components have that ground "in common".***
And yes, I was also confused for many years about the difference between 0V, "ground"**** and the -ve terminal. For example, my battery/power supply has terminals labeled +ve and -ve, whereas my Arduino has pins labeled 5V (or just V) and GND. For our purposes, they are the same thing: the reference point from which we measure voltage.
Eventually I trained my brain not to think in terms of "voltage" but rather "voltage drop", with a mental image of water flowing downstream: there is no "absolute" water level*****, only how far the water drops between one point and another.
This page has a helpful explanation in more detail, with pictures.
And finally, wrt direction of current flow: historically, the early investigators of electricity arbitrarily defined which way current was flowing, and labeled +ve and -ve accordingly.**** This convention is called... uh... conventional current or flow. Many years later, it was discovered that electrons were negatively charged, so physically they are moving in the opposite direction to conventional current.
Any circuit diagram you ever see will use conventional current direction unless it very clearly says otherwise. This is critical for getting your LEDs (and other diodes) the right way around.
----
*This is why with a DC servo, constant power is delivered by two other leads separate from the signal.
**Confusingly, electricians also use "ground" as shorthand for "earth ground", which is a connection directly to the Earth itself. That's the third pin in a three-pin plug, and it protects against shocks e.g. from a short ciruit by routing the power into the Earth, which can safely absorb it. And to make life really confusing, they use the same symbol for both senses of "ground".
***For example, with a DC servo, the +ve power voltage and the signal voltage are both measured relative to the common GND lead (typically black in low-voltage electronic circuits, but for some reason often brown in servo connectors. *shrug*).
****Not even sea level is the same everywhere in the world, it turns out.
*****This is where the "water level" metaphor breaks down. Water has a very specific downward direction thanks to gravity. The voltage direction of electricity is an arbitrary convention.