An Introduction To Gait Analysis, 4th Ed. |BEST|
Recently, inertial sensor based systems are becoming popular. In [3], inertial sensors are installed on a leg and gait phases are identified by computing angles of leg segments. In [4], inertial sensors are used to estimate upper-limb orientation. More relevant results are given in [5,6], where inertial sensors are installed on a shoe or a leg and the foot movement is estimated using inertial navigation algorithms. Step length is estimated in [5] and a foot movement is estimated in [6]. These techniques can estimate step length or foot movement without range limitations. However, due to double integration errors, the accuracy tends to degrade as time goes by.
An Introduction to Gait Analysis, 4th Ed.
One possible application of the proposed system is the clinical gait assessment of patients, where long walking ranges are desirable. Another application is parameter estimation of some pedestrian navigation algorithms [10]. In those algorithms, relationship between step length and one step walking time needs to be identified; thus step length and one step walking time should be accurately measured. We note that step length measurement using vision only [11] and partial combinations of vision and inertial sensors are reported in [12]. In this paper, vision and inertial sensor data are tightly coupled in the proposed algorithm. The proposed system can be applied only for the flat ground. For example, the proposed system cannot be used for stairs.
Recently, studies are also focusing on head-worn devices for gait analysis. For instance, earbuds and smart glasses are of great interest because they are well-known to consumers and provide a highly seamless user experience. Head mount displays (HMDs), in addition, are directly engaged in virtual reality (VR) and augmented reality (AR) applications. The idea of using head-worn devices in gait analysis is supported by literature. It is reported that head acceleration is helpful to analyze gait events and gait patterns [11]. Researchers have succeeded in embedding gait analysis features on ear-worn devices [12,13,14] or head-worn inertial measurement unit (H-IMU) solutions [15]. These sensors have been developed to provide more gait parameters independently.
In terms of gait symmetry analysis, however, single head-worn sensors have not been developed as an independent solution. In these cases, temporal-spatial gait parameters (e.g., step time) have been measured and differences between the odd and even data periods are compared [13]. However, these kinds of sensors need support from additional sensors because of the recognition of left or right steps. Without independent solutions with single head-worn sensors, their advantages, such as cost effectiveness, spatial freedom, and long product lifetime, are not expected in gait symmetry analysis. No previous studies, nevertheless, have been published for independent solutions using single head-worn sensors in gait symmetry analysis. The development of the independent solutions for gait symmetry analysis has recently been required more often due to the increasing number of head-worn devices such as earbuds, smart glasses, and HMDs.
This paper, therefore, starts to develop independent head-worn sensor solutions for gait symmetry analysis. On the one hand, the head is swaying and oscillating during walking. On the other hand, the hip obliquity (the pelvic rotation about the anterior-posterior axis, influencing the trunk lateral movements) is a gait determinant [1]. As the trunk and head are in coordination during walking [16], the head swaying and oscillating trajectory might be used as a gait determinant. In gait analysis, diagrams have been used, such as the sagittal plane joint angles [17], hip-knee angle-angle diagram [18], and butterfly diagram [19]. These analysis methods are, however, normally used with angular kinematics of low limbs or forces between the foot and ground. Thus, these diagrams are not fit for the linear kinematic analysis of the head trajectory and, moreover, a symmetric format of these diagrams is necessary to display the symmetricity of the data. One of the most widely used diagrams to confirm symmetricity is the eye diagram (also called an eye pattern) in digital communications [20]. With the eye diagrams, circuit designers and testers can intuitively notice asymmetric patterns [21], even with their naked eye. The eye diagram helps circuit designers to decide quickly whether they need to modify the circuit design in terms of the size (or the ratio in the size) of n-type and p-type transistors which are responsible for falling and rising signals, respectively. Secondly, to build modification plans or to decide exact sizes of transistors, they look up more detailed parameters of the eye diagram, such as the eye height, eye width, and eye crossing percentage. Finally, the eye diagram can be used for a proof of the system reliability.
Box and Whisker diagrams of (a) the average eye height at three events: lateral peak (EH.LP), vertical valley (EH.VV), and heel strike (EH.HS), and the step width measured by foot sensors (SW.FT), as well as (b) their three gait symmetry indices (RI, SI, and GA).
In terms of methods, as a pilot study, twelve participants were involved, by wearing an IMU-based motion capture system (XSENS). Nevertheless, data were valuable because it can be found that parameters are reliably in a certain range and trends in terms of straight gait under the flat ground condition. It is also remarkable that their shapes of head trajectory are all different, which could be one source of personal identification using gait patterns.
Although SNR in data communications has a unit of decibel (dB) as a log scale parameter, the SNR concept in gait analysis does not result in a log scale value in this study. The first reason is to reduce unnecessary computations. Secondly, the final goal is to compare left and right values, so that it was not important if the parameter is a log scale or not. The comparison results might be different in different groups and conditions. In future studies, it can be figured out if the log scale is more informative or not.
In addition, the eye crossing percentage in data communications plays an important role in the analysis of signal symmetry (e.g., rising and falling signals). The concept of eye crossing percentage is, however, not mentioned in this study because the gait symmetry indices of the eye height can replace it to show the level of gait symmetry. In further research with different conditions, the concept can be additionally proved, by measuring the distance between AEs of the left (positive) and right (negative) side.
Two main approaches are used for gait assessment: kinematics and kinetics. The kinematics studies the description of motion of the body without considering the causes of motion [12], such as joint angles, center of mass (CoM) displacement and velocity, and spatio-temporal gait parameters. The spatio-temporal gait parameters describe the gait relating the foot placement, gait events timing and velocity variables [3], and their measurement forms the basis of any gait analysis [1]. Among the technological devices commonly used for kinematic analysis of the human gait there are camera-based optical motion capture systems based on image processing, capable of collecting in three dimensions body joint positions and estimate spatio-temporal gait parameters and joint angles [13]. The marker-based optical motion capture systems are considered to be the gold standard for gait analysis and commonly used as reference [12]. Other solution for kinematic analysis are the inertial sensors, which are attached onto the surface of the human body, collecting the linear acceleration and angular velocities, allowing the combination of these data to estimate the joint angles and spatio-temporal gait parameters [2].
The proposed smart carpet can measure the plantar pressure along the device and to estimate the spatio-temporal gait parameters using the sensors distributed throughout carpet. To enable the side-coupling of the light source and, at the same time, increase the sensor sensitivity, a lateral section is made on the fiber, where the cladding and part of the core are removed, creating the sensitive zones demonstrated in Figure 1 inset. The lateral section length, depth and surface roughness were made through abrasive removal of material following the guidelines presented in [32]. The sensitive zone is created using a sandpaper connected to a rotary tool. The POF is positioned on a fixed support where the rotary tool advancement is limited by another part over this support to guarantee the desired length and depth of the lateral section in the POF sensor [32].
Lastly, the third experimental protocol consisted of three walking tests, in which the volunteers started the tests with right foot. The goal of this protocol is to validate the previous characterizations and to estimate the GRF and spatio-temporal gait parameters during walks.
Figure 5 shows a significant difference between the response of sensor 3, where the force was applied, and the responses of the adjacent sensors (sensors 1, 2, 4, and 5). Considering the low power variation of the other sensors, the crosstalk between sensors is negligible. The cross-sensitivity between both photodetectors is only observed in sensor 10 due to the proximity of this sensor to photodetectors P1 and P2. However, human gait comprises of sequential contralateral steps [1], in which the foot will be placed in one fiber at a time. For this reason, we can identify if the sensor 10 variation was caused by the left or right region of the fiber by analyzing the previous steps. It is noticeable that the power variation decreases as the weight increases. This is due to the exponential behavior of the sensor with saturation tendency, showed in Figure 3.
Table 1 presents the results of spatio-temporal gait parameters. The volunteers were asked to perform four steps on each walking test. Since the tests were performed by young adults, the step and stride lengths were shorter than the ones commonly obtained in other gait experiments [3]. For walking tests applied to kids or older people, the step and stride lengths would be naturally shorter than the obtained results and can be analyzed with the same system due to the system modular configuration and high spatial resolution. The version of the POF Smart Carpet has 180 cm, which presumably results in a mean of step length of about 45 cm. It is worth noting that the step lengths presented in Table 1 generally are close to 45 cm, with a few deviations from this mean value due to the intrinsic variability of the gait [34]. The cadence variability can be related to self-selected pace, which results in a different velocity patterns for each test. However, it is possible to observe similarities on the cadence if the tests of each volunteers are analyzed (see Table 1). 041b061a72