The computer doesn’t know what is “head” and what is “background” in the footage. All it sees are lights, darks, and colours.
So you have to tell it what is head, and what is background. If the person were shot on a greenscreen, you could tell the computer “green is background, everything else is foreground — cut it out”.
However, this was not shot on greenscreen, so you have to use the pen tool to draw a mask around the person’s head. This is called “rotoscoping”, or “roto”. The process is in the help under “animating masks”, or something like that.
Once you read up on that, here’s a hint: only set mask shape keyframes when the direction of shape movement changes. let AE interpolate between.