Gesture recognition lets a person control a screen by moving, rather than by touching it: a hand wave to advance, a point to select, a body position that triggers a response, read by a camera or depth sensor.
It found real demand when touch became unwelcome, and it still earns its place where touching is impractical: behind glass in a shop window, in a food preparation area, or where the screen is out of reach by design.
The caveat is discoverability, and it is close to fatal. A touchscreen announces what it is; a gesture interface looks like a normal screen playing content, and nobody gestures at a screen unprompted. Every deployment needs an explicit invitation (an attract loop showing a hand doing the gesture), and even then the first attempt is usually wrong.
Second, gestures are tiring and imprecise. Holding an arm up for more than a few seconds is uncomfortable, and fine selection is unreliable, so the interaction must be coarse and short. Three large targets work; a menu does not.
Where touch is acceptable, it is almost always the better interface. Reach for gestures when there is a reason touch is impossible, not for novelty.