Abstract
BACKGROUND:
Due to the problem of aging societies, there is a need for smart buildings to monitor and support people with various disabilities, including rheumatoid arthritis.
OBJECTIVE:
The aim of this paper is to elaborate on novel techniques for wireless motion capture systems for the monitoring and rehabilitation of disabled people for application in smart buildings.
METHODS:
The proposed techniques are based on cross-verification of distance measurements between markers and transponders in an environment with highly variable parameters. To their verification, algorithms that enable comprehensive investigation of a system with different numbers of transponders and varying ambient parameters (temperature and noise) were developed. In the estimation of the real positions of markers, various linear and nonlinear filters were used. Several thousand tests were carried out for various system parameters and different marker locations.
RESULTS:
The results show that localization error may be reduced by as much as 90%. It was observed that repetition of measurements reduces localization error by as much as one order of magnitude.
CONCLUSIONS:
The proposed system, based on wireless techniques, offers a high commercial potential. However, it requires extensive cooperation between teams, including hardware and software design, system modelling, and architectural design.
Keywords
Introduction
Smart buildings are usually understood as systems that offer various facilities that increase the comfort of life, make daily life safer and optimize the costs of living. However, due to the problem of aging societies, buildings of this type should also support medical healthcare and diagnostics, especially of persons with motion disabilities such as long-standing rheumatoid arthritis, Parkinson’s disease, cerebral palsy, etc. One of the most characteristic, physically debilitating diseases is rheumatoid arthritis (RA). During the course of the disease, persistent inflammation often leads to bone erosions and joint damages and eventually causes several physical impairments and significant, irreversible disability [1]. In many cases, disease progression handicaps or even precludes the performance of everyday living activities. RA, since its earliest stages manifests itself by pain, joint deformities, stiffness and swelling, thus resulting in motion difficulties [2, 3, 4]. This paper focuses on the development of techniques and algorithms suitable for motion capture systems, which record the trajectories of moving segments of the body and includes an analysis of the recorded data, enabling medical diagnosis and rehabilitation. The system records the pattern of the human gait in a 3-D environment in real time, thus belonging to the group of real time localization systems (RTLS). The most important parameters include localization precision, and system capacity, i.e. the number of markers that may be tracked simultaneously with a given frequency. From the architectural perspective, the problem may be investigated through the perspective of the building in which the system is applied. The layout of the living space and its equipment (furniture, materials used, etc.) may cause reflections and occlusions. Another problem is the influence of the environment (noise, temperature variation) on the system performance. All these factors have a strong impact on the number of devices used in the system and the complexity of the algorithms used. Alternative state-of-the-art technologies include: ultrasound (Zebris) [5], magnetic [6], inertial (NORAXON, Aminian) [7, 8], and – recently – wireless technologies [9, 10]. The most advanced motion capture system is offered by VICON company [11] with a mean absolute error of 0.15 mm and variability below 0.025 mm. This approach, however, is too expensive to be widely used in the home environment. For this reason, there is a need to develop alternative solutions based on other technologies. In this paper, the assumption is made that the hardware layer is either a typical one or will be custom-designed in the future, and is therefore excluded from the analysis. The focus is on the enhancement algorithm, which is the core of the developed software system. The aim of this paper is to elaborate on novel techniques for wireless motion capture systems for the monitoring and rehabilitation of disabled people including rheumatoid arthritis for application in smart buildings [12, 13, 14]. It was hypothesized that the proposed system may be used directly in the place of residence of disabled people.
Methods
Test methodology
The system consists of both the hardware and the software subsystems. The software plays the role of a model, which allows to modify such parameters as the number of transponders, the noise levels, the range of the temperature variation, as well as the parameters of the algorithm itself. Over 5000 simulations were performed for different combinations of the described parameters. The simulation environment allows to test very complex and fully repeatable scenarios, even with heavily exaggerated negative phenomena. For example, location errors in the IR-UWB technology are typically at the level of 1 to 10 centimeters, while the system was tested for as much as 100 centimeters to assess its robustness. In the model, such tests can be planned with precision (exact trajectories of the markers – ground truth data). Simulations can be performed for exactly the same trajectories and for different levels of environment imperfections, which allows for a direct comparison.
Enhancement algorithm
The proposed algorithm is based on repeated measurements between particular transponder-marker pairs (TMPs), which enables statistical improvements. Individual transponders successively send their signals to particular markers, identified by distinct identification numbers (IDs) contained in these signals. A given marker needs a certain amount of time to make a comparison between the received number and its own ID. If the numbers match, a given marker sends a feedback signal that is recorded by a given transponder. Based on the measurement of the difference between the transmission and the reception times, a distance is determined. In a single measurement session, it is impossible to gain information about the direction the marker sent back the feedback from. For this reason, more transponders are paired with each marker. Based on the knowledge of their positions (the system is calibrated before measurement), the framework of transponders localizes the marker. One of the sources of errors is temperature variation, which impacts the response time of the markers. The measured time may be longer or shorter than a nominal value, which results in a set of apparent positions (APs) of the marker around its real position. This effect creates an uncertainty region (UR) that resembles a circle, as shown in Figs 1 and 2. Let it be named an apparent circle (AC). In the following steps, from the set of APs, subsets of three different APs are selected, on the basis of which we estimate parameters (
Results
Example results demonstrating the impact on temperature are shown in Fig. 1, for two example resultant UR cases: 225 (a, b) and 750 mm (c).
Selected results for two example URs of: (a) 225 and 750 mm (b, c).
The errors between real marker positions (
The performance of the algorithm for a UR radius of 375 mm after the addition of white noise (10% of UR radius) and different numbers of repetitions of measurements between particular TMPs: (a) 1 repetition; (b) 2 repetitions; and (c) 5 repetitions.
These values are exaggerated in comparison with real conditions, to better illustrate the performance. Large blue circles (radius) represent distances between particular TMPs. Under nominal conditions, they should cross at the real position of the marker (a grey circle); however, due to temperature variation they cross in APs, i.e. small dark circles at the UR border. The small blue solid circle illustrates the position of the marker after estimation. Individual APs represent measurements performed by particular transponders (located here at the red square plan). The dark “dashed” circles are IACs. Due to the aforementioned negative conditions, the centers of the ACs are distributed over the scene. Figure 2 presents the errors between real marker positions (
Figure 3 illustrates the performance of the algorithm for a UR radius of 375 mm after the addition of white noise (10% of UR radius) and different numbers of repetitions of measurements between particular TMPs.
The results presented in Figs 1 and 2 show that, as expected, measurement error increases together with UR radius in an approximately linear manner. The results shown in Fig. 3 show that repetition of results between particular TMPs enables to eliminate noise to levels comparable with situations with noise absent.
The paper presents an enhancement algorithm that improves the results of measurement of the position of markers in wireless motion capture systems. The algorithm makes it possible to improve measurement results by over 90%. Given the accuracy of the current location systems, it is anticipated that a wireless technology-based system can achieve an accuracy of about 1–2 mm, which is comparable with other solutions such as Zebris. At the same time, the cost will be substantially reduced. This creates the opportunity of a common use of the proposed system in intelligent buildings.
Footnotes
Conflict of interest
None to report.