Article Preview
TopIntroduction
As an important device in VR environment, the head mounted display has become a mainstream product in VR environment. To simulate a more realistic VR environment, more and more manufacturers begin to develop the device for foot motion recognition. However, according to Gerd Bruder (2013), there are few devices used for foot motion recognition. As the main device for foot motion recognition in VR environment, the VR treadmill requires the user to wear a pair of special shoes which can catch the foot motion, and the user should lean on a ring fence to moves their foot on a treadmill, which is an unnatural and uncomfortable way to move in VR environment. The treadmill as shown in Figure 1.
There are many devices and methods that focus on foot motion recognition. Visual-based approaches are effective techniques which directly obtain foot motion data from video streams using stereo vision (Zheng, 2014). Muqing Deng (2016) used the camera optical system to track the foot motion, however it is limited to laboratory conditions due to its requirement of large space and high capacity processing and recording storage. Wang (2019) used plantar pressure imaging pre-process and segmentation techniques to extract the region of interests in the captured images from the optical pressure sensors. However, this method is limited to its application environment, not allowing user to freely move in a VR environment. Da-Un Jung (2009) provided a new motion-capture method by extracting the foot object from sequence images taken by stereo cameras. However, this method has difficulty in extracting required features and implementing in real-time because of its environment limitations for features and colors, and they are very sensitive to the foot material and environment light and very susceptible to the shelter from other objects. Tri-Nhut Do (2011) proposed a new motion capture system to track foot motion using a camera and IR-LEDs. However, this method is also very sensitive to the foot material and environment light and very susceptible to the shelter from other objects. Ju-Hwan Seo (2012) introduced a new system consists of a laser scanner that can detect user’s foot motion, but this system suffers from its limited ability of scanning different foot motion, which cannot support foot motion recognition in a large-scale VR environment. J. Shotton (2011) propose the depth-image based technique with the techniques of real-time foot motion capture and reconstruction. However, this approach suffers from the same drawbacks of all vision-based motion capture systems. Parwekar (2014) proposed a wearable system that can be put into a mount which can be attached to a shoe. Wang (2017) attempted to develop a fusion model for plantar pressure distribution images, which is expected to contribute to feature points construction based on shoe-last surface generation and modification. According to Liu Rong (2007) and Fei Wang (2011), the algorithms of Dynamic Time Warping (DTW), Principal Component Analysis (PCA), SVM and Fast Fourier Transform (FFT) are classical algorithms for foot motion recognition. However, these algorithms usually need to compare with the database, as the comparison time is relatively long, it is not suitable for the foot motion recognition in VR environment. Casas (2017) reviewed the most important strategies for the generation of motion cues in simulators, listing the advantages and drawbacks of the different solutions.
To overcome the problems of image-based methods, and the unnatural and uncomfortable experiment of VR treadmill, we designed a light and portable device with a gyroscope(Birsel Ayrulu-Erdem (2011), Paola Catalfamo (2010) and Zheng (2014)) to collect the altitude data of foot motion and proposed an new adaptive algorithm based this device for foot motion recognition, and user’s foot motion can be recognized only using this device without any cumbersome or external tracker system, which could provide the user much natural and easy ways to move in the VR environment. Proposed method could also completely overcome the problems of visual-based methods, which provide a more reliable recognition effect, higher recognition accuracy and speed.