Walk to Move Through a Virtual LabyrinthWalk to Move Through a Virtual Labyrinth

The authors list six movements the system must recognise: walking, rotating the body, and the head looking left, right, up and down. No single sensor covers all six. The gyroscope handles turning the head sideways, while looking up and down is derived from accelerometer data as a pitch angle.

Body movement was the harder half. The team combined two existing approaches. The first reads rotation rate and accumulates its magnitude, then compares that number against two thresholds: below the first the user is still, between them the user is walking, above the second the user is rotating. On its own that method reached 84.5 percent accuracy, and it slipped when someone walked much slower or faster than usual.

• The second approach compares accelerometer values between the current and the previous frame, and calls it walking when the difference lands between 0.9 and 0.994.
• Combining both approaches gave the best score on every measure, so the prototype runs the pair together.

Rendering was done in Unity3D for Android and iOS with two viewports, shifted a short distance apart for the left and right eye. The gyroscope reports attitude in a right handed coordinate system while Unity3D uses a left handed one, so the quaternion values need conversion. Spherical interpolation smooths the camera between frames so the view does not jump.

The speed study adds practical limits. Rotation was detected reliably between 18 and 45 degrees per second, and walking between 0.5 and 2 steps per second. Outside those bands the detector loses track of the player.

Thirty users played the game and all of them enjoyed the controls. The authors also name the obvious constraint. This kind of control needs room, because a player who has to walk and spin cannot do it at a desk.