People are interested in traveling in an infinite virtual environment, but no standard navigation method exists yet in Virtual Reality (VR). The Walking-In-Place (WIP) technique is a navigation method that simulates movement to enable immersive travel with less simulator sickness in VR. However, attaching the sensor to the body is troublesome. A previously introduced method that performed WIP using an Inertial Measurement Unit (IMU) helped address this problem. That method does not require placement of additional sensors on the body. That study proved, through evaluation, the acceptable performance of WIP. However, this method has limitations, including a high step-recognition rate when the user does various body motions within the tracking area. Previous works also did not evaluate WIP step recognition accuracy. In this paper, we propose a novel WIP method using position and orientation tracking, which are provided in the most PC-based VR HMDs. Our method also does not require additional sensors on the body and is more stable than the IMU-based method for non-WIP motions. We evaluated our method with nine subjects and found that the WIP step accuracy was 99.32% regardless of head tilt, and the error rate was 0% for squat motion, which is a motion prone to error. We distinguish jog-in-place as “intentional motion” and others as “unintentional motion”. This shows that our method correctly recognizes only jog-in-place. We also apply the saw-tooth function virtual velocity to our method in a mathematical way. Natural navigation is possible when the virtual velocity approach is applied to the WIP method. Our method is useful for various applications which requires jogging.

In this paper, we propose to use the deep learning technique to estimate and predict the torso direction from the head movements alone. The prediction allows to implement the walk-in-place navigation interface without additional sensing of the torso direction, and thereby improves the convenience and usability. We created a small dataset and tested our idea by training an LSTM model and obtained a 3-class prediction rate of about 90%, a figure higher than using other conventional machine learning techniques. While preliminary, the results show the possible interdependence between the viewing and torso directions, and with richer dataset and more parameters, a more accurate level of prediction seems possible.