2.1.

Subjects

This study included seven males and three females (mean age $25.3\pm 2.3$ years and mean weight $66.3\pm 5.7\mathrm{kg}$) with no history of discernable gait abnormalities. The purpose was explained to each subject before they were asked to give their written informed consent for participation in this study. These subjects were selected from the postgraduate students of the National Engineering Laboratory for Fiber Optic Sensing Technology, Wuhan University of Technology, which approved our research.

2.2.

Measurement Principle of FBG Sensors

As shown in Fig. 1, two sets of FBG sensors (New port Inc., $400\mathrm{lp}/\mathrm{mm}$, $30\mathrm{mm}\times 30\mathrm{mm}$) are twisted separately around the ball and heel of the human foot. The dispersion parameter of each FBG sensor is $-11\mathrm{ps}/\mathrm{nm}$, and each FBG sensor possesses a high reflectivity of 97%. In addition, the FBG sensor includes a periodic modulation of the refractive index in the core of a single mode fiber optic cable. If the FBG sensor is illuminated with broadband light that meets the Bragg condition, a narrow spectrum is backreflected centered around the Bragg wavelength ${\lambda }_{\mathrm{B}}$, which depends on the periodic variation $\mathrm{\Lambda }$ of the FBG and the effective index $n_{\mathrm{eff}}$.¹ This condition can be described as follows:

Fig. 1

Schematic diagram of gait pattern detection systems.

Light that does not meet the Bragg condition transmits through the structure. In this paper, the reflected range of ${\lambda }_{\mathrm{B}}$ is from 1277 to 1292 nm, $n_{\mathrm{eff}}$ is about 1.45, and $\mathrm{\Lambda }$ is $\sim 443\mathrm{nm}$. The range of index modulations is in the ${10}^{-5}$ level, while the grating lengths range from 3 to 5 nm.

When the strain in an FBG sensor changes, the backreflected wavelength will shift such that the stain can be measured through the wavelength shift. Moreover, the length of the FBG sensor changes with the axial strain, and the change quantity of the FBG sensor changes with the axial strain is written as $\mathrm{\Delta }l$. As a result, this situation produces a change in the grating period (i.e., $\mathrm{\Lambda }$). Meanwhile, the axial strain induces a change in $n_{\mathrm{eff}}$, the effective index of the grating. Therefore, according to Eq. (1), the change in the center wavelength due to axial strain can be stated as

The FBG sensor used in this work is highly sensitive to axial strain, and small variability in the length of the FBG will lead to measurable change in the backreflected light spectrum. In addition, the temperature variability would also cause the length changes of FBG sensor.¹¹ Therefore, the work is carried out in a constant temperature laboratory, so the FBG sensors are only subjected to the axial strain. To determine the Bragg wavelength ${\lambda }_{\mathrm{B}}$, a micron optics (MOI) FBG static sensor demodulation instrument, as seen in Fig. 1, is used to capture the central wavelengths of the backreflected signal.

The central wavelengths are transmitted into a notebook computer through a shielded twisted pair with a Registered Jack 45 (RJ45) connector. One graphical user interface (GUI) compiled by LabVIEW is utilized to store the central wavelengths in synchronization with the computer clock. In addition, the central wavelength data from FBG sensor are filtered by Butterworth low pass filter with a cut-off frequency of 5 Hz.

2.3.

FSR Calibration

As shown in Fig. 1, two sets of FSR sensors (LSH-10, LOSON Instrumetation, Nankin, China) were placed separately inside the insole of the ball and heel of the human foot. The FSRs were utilized to capture the magnitude variations of GCFs in the plantar area, while the GCF data were collected at a frequency of 1000 Hz using a data acquisition module (National Instruments) with a universal serial bus (USB). The measurement range of each FSR is from 0 to 200 kg, and each FSR possesses a high accuracy of $\pm 0.5\%$ full scale. As the respondence of FSR shows the best linearity to the GCF exerted by the human foot in different walking postures, standard load cells (5, 10, 20, 25, 50, 100, and 200 kg) are used to calibrate the FSR. Because the output signal of each FSR is a week microvoltage, an amplification circuit is integrated to amplify the output signal of each FSR to 0 to 5 V, which correlates with the measured mass of 0 to 200 kg.

The GCF data collected by the data acquisition module are transmitted into the notebook computer through the USB connector. Another GUI compiled by Visual C++ 6.0 is used to store the GCF data in synchronization with the computer clock. Moreover, the GCF data from FSRs are filtered by Butterworth low pass filter with a cut-off frequency of 10 Hz.

2.4.

Description of Walking Experiments

After sensor calibrations, the experiments are carried out to test the reliability of FBG sensors for gait pattern detection. For reference platform, two sets of FSRs are mounted in different foot contact points such as the ball and the heel. For test platform, one FBG sensor is located in a twisted way around the ball of the human foot, while the other is mounted around the heel of the human foot. Then, each subject performed 10 trials including at least 50 gait cycles each time at a time. The first five trials were performed in a straight line, while the other five trials were performed in a curved line.

2.5.

FBG Wavelength Shift During Different Walking Poses

As reported by Li et al., the FBG sensor is also sensitive to temperature changes, so the entire experiments are conducted in a laboratory with constant temperature under a controlled condition.² To determine the wavelength change of the FBG, the Bragg wavelength shift was recorded using an MOI FBG static sensor demodulation instrument. It is quantitatively illustrated in Table 1 where the original and after-mounted central wavelengths of two FBG sensors are presented. Figure 2 shows the FBG wavelength changes during human motion process. It can be seen that the FBG spectrum would be slightly shifted away from its original central wavelength as the FBG sensor is laid still on the human body. In addition, the FBG spectrum would continue to be shifted when subjects begin to take a step. Figure 3 shows the curve of FBG central wavelength change during gait cycles. The cyclic characteristic of the curve suggests that the central wavelength changes correlate to human walking and exhibit the possibility of application to gait pattern detection.

Table 1

Central wavelength (nm) of the FBG sensor.

Fig. 2

Central wavelength shift on different motion conditions.

Fig. 3

Curve of central wavelength from FBG sensors: (a) mounted on the ball and (b) mounted on the heel.

2.6.

Gait Patterns Detection Algorithms

In human walking, a whole gait cycle can be divided into stance-phase and swing-phase. The transitions between gait phases are gait events.¹² Specifically, the transition from swing-phase to stance-phase is the gait event of heel-strike, while the transition from stance-phase to swing-phase is the gait event of heel-off. Stance-phase is defined as when the human foot is in full contact with the ground. Swing-phase is defined as when the human foot totally leaves the ground. Heel-strike is defined as when the heel of the human foot contacts the ground but the ball leaves the ground. Heel-off is defined as when the heel of the human foot rises from the ground but the ball touches the ground. In this paper, two types of sensor platforms are employed to detect these gait patterns, and the detection methods are presented in the following.

2.7.

Evaluation of the Results

Many studies^19–21 in the literature compared their research results with the standard force platforms as reference to investigate the accuracy and reliability of their gait pattern detection systems. Hence, the force platform in this work is chosen as the reference system. The reliability of this study can be determined by comparing the detection results between the reference system and the FBG sensor platform.

3.1.

Selection of Coefficients

To determine the reliability of the FBG sensor platform, the golden standard force platform was chosen as the reference. Then, the detection results of the FBG sensor platform were compared with the reference platform.

For the force platform, the threshold value for the status division was determined by the method of defining 5% of subject body weight.⁸ For the FBG sensor platform, $\alpha$ in Eq. (5) was chosen to be 0.1.

3.2.

Results of Gait Pattern Detection

The FBG sensor platform was considerably reliable in gait pattern detection when compared with the golden standard force platform (Table 4). The average reliabilities were 83.50% and 83.28% in the first five and other five trials.

Table 4

Reliability of the proposed method.

The GCFs collected by FSRs from the ball and heel were divided into on-ground status and off-ground status by setting thresholds (Fig. 4). Similarly, the central wavelengths collected by FBG sensors from the ball and heel were divided into on-ground status and off-ground status (Fig. 5). For the force platform, the gait patterns can be detected through the rules in Table 2 based on the status division of GCFs (Fig. 6). For the FBG sensor platform, the rules in Table 3 can distinguish the gait patterns based on the status division of central wavelengths (Fig. 7).

Fig. 4

A threshold divides the GCF into on-ground and off-ground statuses: (a) GCFs from the ball, (b) GCFs from the heel, (c) status of GCFs from the ball, and (d) status of GCFs from the heel.

Fig. 5

A threshold divides the central wavelength into tension and no-tension statuses: (a) central wavelengths from the ball, (b) central wavelengths from the heel, (c) status of central wavelengths from the ball, and (d) status of central wavelengths from the heel.

Fig. 6

The gait patterns are distinguished based on the detection rules for force platform.

Fig. 7

The gait patterns are distinguished based on the detection rules for FBG sensor platform.

4.1.

Advantages of the Research

The FBG sensor is easily mounted on human body parts and can measure the tension on the body surface. When a person is in ill health, the tension on the affected part of a patient should be different than that of normal people. Therefore, except for the application to gait pattern detection, the FBG sensor can also be used to monitor people’s health.

As applied to gait pattern detection, the FBG sensors need no amplifying circuit, but the FSR does. The central wavelength data can be directly transmitted into a computer using the MOI FBG static sensor demodulation instrument. In some experiments,^2,3,22 the central wavelengths should be calibrated into specific forces or angles. However, the central wavelengths can be directly processed into gait pattern without calibration.

4.2.

Limitations of the Research

For some body parts, there is little tension in a whole gait cycle such that the central wavelengths of the FBG sensor vary a little. Therefore, the FBG sensor should be located on a human body part where a strong tension occurs.

Sometimes, the subject was sweating in the experiments, so the FBG sensor fell off. As a result, the experiment was stopped until the subject cooled.

4.3.

Future Work

In fact, the curve of the central wavelengths from the FBG sensor mounted on the shank represents more information, so the subphases of swing can be detected. These subphases include preswing,²³ toe-off,²⁴ and terminal-swing.²⁵ Specifically, when the central wavelengths are divided into tension status, the whole leg is swinging in the air and the curve changes back and forth. The curve characteristic correlated with the subphase detection of the swing-phase.²⁶ This hypothesis can be affirmed in the future.