A human-machine interaction method, system and apparatus of the present invention enables a user to interact with an electronic device by turning the user's head to implement a mouse's point-and-click functionality. A main characteristic is to detect the user's head rotation (moving direction, displacement and nod/shake one's head action) following the direction of eye watching by adding an angular velocity sensor in an earphone, so as to realize a more natural and convenient human-machine interaction, where you are watching is where the cursor will point, look-and-nod instead of point-and-click with mouse. With the supports of the virtual keyboard on the screen of the electronic device, information can be inputted. Combined with the earphone microphone sound input, it can realize more complicated control combination with head movement input and sound input; making use of the earphone sound output as user output, user input and user output are integrated in one.