I guess the most straightforward way of adding an artificial echo to a sound signal is to add its time-delayed version to the original signal. However, I am not sure how realistic that would sound like.
I was reading papers on room geometry estimation from impulse responses (see for example this paper). So, perhaps one should have several echoes instead of just one.
I guess something also happens to a sound signal when it is reflected (kind of filtering).
What would you suggest here?