160 A.R. Ortiz et al. techniques. The Author used a kinematic gait analysis system to determine the three dimensional motion of the center of the pelvis during walking. The measurement of the center of the pelvis was synchronized with a force platform, and these data were integrated to eventually determine the motion of the center of gravity of the body. Recently, Dang and Zivanovic in 2015 [6] used a more advanced motion capture system, which consists of twelve video-based optoelectronic cameras and sensors that were used to capture displacements of 34 markers attached to human anatomical landmarks. The Authors considered four different markers’ arrangement to increase accuracy of the estimation of such parameters of walking as: the pacing frequency, step length, step width, attack angle, end-of-step angle, trunk rotation and first harmonic of the dynamic loading factor. An example of an application of an inertial motion tracking system for the analysis the walking behavior of pedestrians is the paper by van Nimmen et al. [7]. The Authors used the Xsens MTw Development Kit measurement system, which consisted of multiple wireless inertial units incorporating 3D accelerometers. The system consisted of six sensors, one of which was placed close to the body’s CoM located at the level of the fifth lumbar vertebra. To estimate the trajectory of the pedestrian’s center of mass during walking the above-mentioned measurement system has to be combined with appropriate computational algorithms. Usually, two types of methods are utilized: dynamic or kinematic. The kinematic methods can be further subdivided into minimalistic marker methods and segmental analysis methods. The latter suffer from uncertainties arising from missing information about limb motion. The former, on the contrary, require full-body marker sets and calculate the CoM trajectory by assuming the masses and CoM locations of each segment [8]. The accuracy of the dynamic methods is limited by the precision of the measured forces and the precision of the integration constants. It was observed by Maus et al. [9] that inaccuracies in each method, dynamic or kinematic, are related to different parts of the Fourier spectrum. As a result of this, they proposed a new approach to compute CoM motion based on a reliable frequency range of force and kinematic measurements. An extension of the previous method has been presented by Carpentier et al. [10]. The modification was related to adding information about the center of pressure, also called Zero Moment Point (ZMP). However, according to the Authors’ conclusion, the proposed filtering approach with the ZMP measure does not improve the vertical component of the CoM, because ZMP provides only two-dimensional information. Recently, researchers started to consider infrared cameras combined with video cameras as an alternative to the abovementioned marker-based systems. An example of this new measurement technology is the Microsoft Kinect sensor. After finding its popularity among gamers, MS Kinect has received a lot of attention from biomechanical, mechanical and civil engineers. Jun et al. [11] performed a comparative analysis of motion data from two alternative human motion-capture systems (Vicon vs Kinect). They employed an 8-camera Vicon MX system that was synchronized with the Kinect system via a video synchronizer (Kistler 5610). They concluded that direct application of the Kinect system to clinical or research work (without post-processing of raw data) tends to be limited. However, they pointed out that with suitable post processing there is a potential for clinically relevant use. Another paper related to the accuracy of the MS Kinect was published by Galna et al. [12]. They proposed to use MS Kinect as a potentially low-cost solution for clinical and home-based assessment of movement symptoms in people with Parkinson’s disease. Another application of the MS Kinect sensor was proposed by Seer et al. [13]. In their study, the Authors proposed an algorithm for human detection and tracking. The algorithm is based on agglomerative clustering of Kinect depth data captured from an elevated view in contrast to the lateral view used for gesture recognition in Kinect’s gaming applications. They combined measurement signals from three different Kinect sensors. Yet another comparison of MS Kinect with a marker-based motion capture system was presented by Zerpa et al. [14]. The Authors used both Vicon Peak Motus version 9 and the Microsoft Kinect system with customized skeleton software to collect data about a subject sitting on a platform moving horizontally at the speed of 2.4 m/min. The results of their study support the findings in literature and indicate that the Kinect system has a potential to be used as a tool to measure and analyze human movement kinematics. The last example of an application of MS Kinect to human movement analysis is the paper by Chen et al. [15]. In their article, the Authors proposed an efficient pedestrian detection approach for crowded scenes by fusing the RGB and depth images from the Kinect. First they extracted pedestrian contour regions from the RGB images using background subtraction and then, they applied a region clustering algorithm to extract pedestrians from the contour regions using depth information. Finally, a tracking and counting algorithm was employed to acquire pedestrian volumes. All the above examples indicate that a depth sensor combined with a video camera can be an attractive alternative to the classical expensive motion capture measurement systems. The present paper is organized as follows: Sect. 20.2 presents the model used for describing the trajectory of the pedestrian and offers an introduction to the probabilistic model updating technique, which is the tool used in this paper for updating the parameters of the trajectory model. Section 20.3 describes the experimental setup to obtain the trajectory of a pedestrian walking on a rigid surface as well as the mobile sensor developed on MS Kinect technology. Section 20.4 presents the experimental trajectory and the parameters obtained after being updated. Conclusions and future work are detailed in Sect. 20.5.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==