Relative pitch and absolute pitch are 2 different skills. While most people can learn relative pitch, some people say that absolute is genetic and can’t be learnt. I don’t believe this is true (at least in my case). My technique is to learn songs that start on a specific pitch (C for example) and internalise the pitch, imagine the sound and try to sing it before comparing with the original. Try this exercise every morning, take notes and see if you improve over time.
From my experience as a singer, relative pitch exercises are much more difficult when the notes are a few octaves apart. So making sure the notes jump around over a large range would probably help.
There are a variety of apps available on Google's Play store. Do you have tried them and found them to fail?
Otherwise I can imagine creating a bunch of Anki cards for different notes that are mixed into your normal Anki cards.
Chris Aruffo has done some work on this: http://www.aruffo.com/eartraining/
There is one dubious study about sodium valproate making it easier to learn AP: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3848041/
As a fallback plan, you can measure and remember what's the lowest pitch your voice can reach. It's not totally stable, but it can be helpful in some extreme emergency situations.
Assuming that, in order to learn perfect pitch, one must train differently than to learn relative pitch:
Absolute pitch is the ability to correctly identify any musical note. It is close to another ability, relative pitch, which is the ability to identify any interval correctly, although relative pitch is usually described as the ability to correctly identify any note, once the subject has been given a "reference tone". An important fact here is that relative pitch is not as rare as absolute pitch, which may hint that absolute pitch is harder to acquire/train.
What kind of learning procedure would you design to learn absolute pitch?
In particular, I have two strategies in mind, and I don't think either would work. The first one is simple: a computer produces a note, the user identifies the note, the computer corrects the user. This is flawed because the first note in pair with the correct answer given by the computer provides a reference tone, therefore, after the first note, the user only trains his/her relative pitch. The second strategy would be to not correct the user immediately, but instead wait for example 10 notes before correcting. This seems flawed too, because of how crucial a quick feedback is to learning. Note that, with either strategies, it is not possible to make long learning sessions anyway, because it soon becomes a relative pitch training strategy.
Any clever idea?