There are already commercial headsets available that can read brainwaves, like Emotiv.
Let us start with such a headset. And let us now add voice recognition capabilities.
Suppose you are in a state happiness and you think, I want to experience this state again in the future.
So you say:"Input
happy".
So the headset then reads the occuring set of brainwaves and saves this profile of brainwaves in a file named happy.
A few days later, you want to feel this state again, so you say: "Command Happy"
Luckily this headset also has a way to activate the brain. For example with tDCS or TMS.
So the headset then starts an algorithm by which the present brainwaves are read and then electrical (or EM or ultrasound) signals are send to the brain in order to adjust the present brainwaves towards the saved profile.
And miraculousy, in only a few seconds, that state of happyness is recreated.
Just as one would look at a picture to recreate a view once experienced
It should be noted however that it may still be unfeasible with present technology.