A study on the application of sensor network for teaching rock climbing in college students’ health monitoring

Considering the design objectives and requirements of the system, we designed the constructed wearable real-time monitoring system [23] as shown in Fig. 1. The results show that the system is abstracted into five layers from the bottom up: hardware layer, data layer, system layer, function layer, and interface layer. The abstraction process follows the principle of high cohesion and low coupling. Each module tries to maintain clear responsibilities and functional independence, while the lower layer provides data or interface calls to the upper layer.

Figure 1.

Architecture of the wearable real-time monitoring system

The block diagram of the design scheme for extending the reserved interfaces is shown in Figure 2.

Figure 2.

Expanding reserved ports

Control unit: ST’s STM32F103 was selected, with high main frequency and support for floating-point operation, which facilitates the integration of complex heart rate and blood oxygen algorithms into the lower computer, taking into account the need to expand a variety of sensors in the later stage of the upper computer, to reduce the burden on the upper computer; and the MCU adopts an OFN package, which has a small volume and is suitable for wearable design.

Heart rate and blood oxygen sensor: NXP’s MAX30102 is used. MAX30102 is a module that features biosensors like pulse oximeter and heart rate monitor, and it also includes LEDs, opto-electronics detectors, optical components, and low-noise electronic circuits that suppress ambient light.

The MAX30102 uses a single 1.8V power supply, with a separate set of 3.3V as the internal LED power supply, to communicate with its standard I2C counterpart. A soft shutdown module is used with zero standby voltage to realize a wearable low power design. The specific block diagram is as follows: Display module:

the lower computer display unit of the prototype, choose OLED screen design, small size, and support pseudo-color display, can realize the real-time refresh display of heartbeat waveform.

In the field of health monitoring, the essence of smart wearable devices is to intervene and improve human health during long-term monitoring. For sensors that are easily affected by factors such as human movements and clothing cover-ups, the design should try to choose a wearable solution with less error. For this reason, the signal acquisition device in this project is designed to be both head-mounted and bracelet-mounted. The overall structure of the head-mounted device is shown in Figure 3.

Figure 3.

Overall structure of the first load equipment

Among them, the motion signal, environment signal and voice signal acquisition module is integrated in the head-hidden device, and the work of these signal-related sensors is easily affected by human body movements and clothing cover, and experiments show that compared with the widely used handheld and wristband devices, the head-hidden device can capture these signals more accurately in complex outdoor scenes. The bracelet device’s overall structure can be seen in Fig. 4. The physiological signal acquisition module is integrated in the bracelet device, and the contact temperature measurement and PPG signal acquisition need to be in contact with the skin.

Figure 4.

Overall structure of the bracelet device

The periodic systolic and diastolic beats of the heart propel blood through the blood vessels, resulting in a pulse wave, the waveform of which is related to basic factors such as the elasticity of the arterial wall, the viscosity of the blood pressure, the volume of circulating blood, the volume of the vascular system, etc., and which are interrelated and influence each other. Therefore, pulse wave pressure and waveform characteristics contain a great deal of information reflecting the important pathology and physiology of the heart and blood vessels. The circulatory system experiences periodic pulse waves due to the regular contraction and diastole of the heart, which leads to changes in blood volume and pressure. A typical pulse wave is shown in Figure 5. Point a is the highest point of the entire pulse map, 2 indicates the descending segment of the pulse wave, point b: the tidal wave, point c is the trough of the heavy wave, and point d is the peak of the heavy wave.

Figure 5.

Typical pulse wave pattern

The elasticity of the walls of arterial tubes primarily affects the wave velocity of the pulse wave. Assuming that blood is an ideal fluid and blood vessels are ideal purely elastic tubes, the wave velocity equation is as follows: 1C=E(R02−R12)3ρR02

Where E is the elastic modulus of the arterial vessel wall, R₀ is the outer radius of the arterial vessel, R₁ is the inner radius of the arterial vessel, ρ is the fluid density, and d is the inner diameter of the arterial vessel wall. From Eq. (1), it can be obtained that the wave velocity C is positively proportional to the elastic modulus of the vessel wall E.

In the process of blood pressure generation, there are a variety of factors that affect arterial blood pressure, which are relatively complex. The following are discussions of basic factors such as cardiac output, heart rate, elasticity of arterial walls, circulating blood volume, and vascular system capacity, respectively. 1)

Cardiac output (CO) is the total amount of blood pumped out of the heart into the arteries every minute, and its value can be considered as equal to the product of heart rate and output per beat. When cardiac output decreases, arterial blood pressure decreases, keeping mean arterial blood pressure at a low level.

Hemoglobin in the blood is a binding protein responsible for transporting oxygen and carbon dioxide in higher organisms, and its primary function is to act as a buffer for blood pH. There are four types of hemoglobin in the blood: oxyhemoglobin (HbO₂), reduced hemoglobin (Hb), carboxyhemoglobin (CoHb), and methemoglobin (MetHb). Oxygen saturation is the percentage of oxygenated hemoglobin to oxygenated and reduced hemoglobin in human blood. Then there is: 2SaO2=cHbo2cHbo2+cHb×100%

Where c_HbO, and c_Hb are the concentrations of oxygenated and reduced hemoglobin.

According to Lombard’s one Beer’s law [24], assuming a wavelength of λ, the incident light intensity I₀ illuminates the human arterial blood perpendicularly, and the outgoing light intensity I through the arterial blood: 3I=I0e−εcV

Where ε is the absorbance of blood, c is the concentration of arterial blood, and V is the volume of arterial blood. When the arterial volume changes to ΔV, the corresponding outgoing light intensity changes to ΔI, equation (3) can be written as: 4I+ΔI=I0e−εc(V+ΔV)

The two equations (3) and (4) are divided and simplified to ΔII=−εcΔV to write equation (3) as: 5V=ln(ll0)(−εc)

Dividing the two equations gives the following equation: 6ΔV/V=Δll/ln(ll0)

That is, the rate of change of the volume of the fingertip artery ΔV / V is proportional to the rate of change of the light intensity through the volume ΔI / I. From the above equation, it can be seen that the transformation from light intensity signal to electrical signal, i.e., from the rate of change of light intensity can be detected changes in the fingertip blood volume this is the principle of blood flow detection of the photoelectric volumetric pulse wave tracing method.

The application of equation (3) to oxygenated hemoglobin and reduced hemoglobin can be written as the following line equation: 7ΔII=−(εHbO2cHbO2+εHbcHb)ΔV where ε_HbO2, c_HBO2 is the absorption coefficient and concentration of oxyhemoglobin and ε_Hb, c_Hb is the absorption coefficient and concentration of reduced hemoglobin, respectively.

Eq. (7) is simplified by combining with the oxygen saturation definition equation: 8ΔII=(εHbO2SaO2+εHb(1−SaO2)ΔV(cHbO2+cHb)

The light source used for the detection of oxygen saturation is composed of two beams of light of different wavelengths (λ₁ = 660nm of red light and λ₂ = 900nm of infrared light), for which Eq. (8) can be written in the following form for each of these two beams: 9ΔIλ1Iλ1=(εHBO2λ1SaO2+εHBλ1(1−SaO2))ΔV(cHBO2+cHB) 10ΔIλ2Iλ2=( εHbO2λ2SaO2+εHbλ2(1−SaO2) ) ΔV(cHbO2+cHb)

Divide the above two equations like to get: 11ΔIλ2/Iλ2ΔIλ2/Iλ2=εHBO2λ1Sa02+εHbλ1(1−Sa02)εHBO2λ2Sa02+εHbλ2(1−Sa02)

Using the absorbance curves of oxyhemoglobin and reduced hemoglobin, wavelengths near 900 nm can be considered as εHbλ1=εHbO2λ2 . Denoting ΔIλ1Iλ1Iλ1ΛIλ2Iλ22 as R, the equation for oxygen saturation can be written as: 12SaO2=εHbλεHbO2λ−εHbλ+εHbλ1εHbλ2−εHbO2λ2R

Finally, the formula for calculating the dual-wavelength pulse wave oxygen saturation based on the Lambert-Beer law is: 13SaO2=a+bR where a=εMbλ1εMcλ1−εMcλ1 , b=εMbλ1εMcλ1−εLcλ2 , εHbλ1 , εHbλ1 , εHbλ2 , εHbλ2 is related to the nature of oxyhemoglobin and reduced hemoglobin, independent of the measurement conditions, and both are constants.

In the model of two-wavelength pulse wave oximetry based on the Lambert-Beer law, firstly, the R value was calculated by finding the maximum and minimum values of the two sets of pulse waves, then the values of a, b were fitted by using the actual oximetry value and the R value, and finally, the linear regression model of the oximetry value and the R value was established.

The absorbance curves of oxygenated hemoglobin and reduced hemoglobin are shown in Figure 6. After many considerations, the 660nm red light at which the molecular absorption coefficients of oxyhemoglobin HbO₂ and reduced hemoglobin Hb are very different and the 940nm infrared light at which the absorption coefficients are almost equal were finally selected to realize the detection of blood oxygen saturation.

Figure 6.

Absorption curves of oxygenated hemoglobin and reduced hemoglobin

The technology for heart rate detection based on photoelectric volumetric pulse wave is a practical technology for non-invasive continuous detection. The principle is that when a certain wavelength of light irradiated to human tissue, due to the diastole and contraction of blood vessels, resulting in the absorption of light by the blood also shows periodic changes, so the received signal can reflect the periodic beating of the heart, so as to extract the heart rate. Heart rate detection methods from photoelectric volumetric pulse wave signals can be divided into two categories: time domain and frequency domain methods. The time-domain method mainly analyzes the photovoltaic volume pulse wave time-domain signal, extracts the characteristic points of the pulse wave, and finally obtains the heart rate. The common ones are differential threshold method and over-zero point method. The advantages of the time-domain method include small amount of calculation and high real-time performance, but it also has the disadvantages of very poor stability and anti-interference ability. The frequency domain method transforms the photovoltaic volume pulse wave time domain signal into the frequency domain by Fourier transform, and calculates the heart rate from the frequency components extracted from the spectrum that are related to the heart rate of the human body. The frequency domain method has a strong anti-interference ability.

Kalman filtering [25–26] is a kind of optimal filtering method based on the statistical characteristics of the signal, and its basic idea is that on the basis of the linear space of the state of the system, after giving the model of the information, the optimal error of the information is estimated by appropriate processing of the noisy inputs and measurements and by introducing a recursive formula in a certain time range, which leads to the actual state of the information system or the real value, and in the general case, it will be able to obtain a signal filtering estimate with higher accuracy than the actual measurement. In this paper, data processing is based on the Kalman filter algorithm, and inertial sensors are utilized to measure the movement posture and respiration rate of the wearer-carrier, so as to complete the monitoring of physical health.

The system state space model is constructed as follows: 14X(k)=AX(k−1)+BU(k−1)+W(k−1) 15Z(k)=HX(k)+V(k)

A and B denote the system parameter matrices; Z(k) is the measurement at moment k ; H is the parameter matrix of the measurement system; X(k) represents the state of the system at moment k ; U(k) is the amount of control over the system that is at moment k ; and W(k) and V(k) denote the noise of the process and the measurement, respectively, and are assumed to satisfy a normal distribution and to be independent of each other.

Time update equation: 16X(k|k−1)=AX(k−1|k−1)+BU(k) 17P(k|k−1)=AP(k−1|k−1)A′+Q

Correction: 18K=P(k|k−1)HTHP(k|k−1)HT+R 19HP(k|k−1)HT+RX(k|k)=X(k|k−1)+K(Z(k)−HX(k|k−1)) 20P(k|k)=(I−KH)P(k|k−1)

X(k | k) is the optimal estimate at moment k : U(k) is the amount of control over the system at moment k (which can be zero if there is no control); A is the state transfer matrix, B is the system parameters: Q is the covariance matrix of the process noise of the system, K is the Kalman gain, and H is the observation matrix; X(k – 1 | k – 1) is the value of the optimal outcome of the previous state; X(k | k – 1) is the result of updating the system by utilizing the predicted value of the previous state; P(k – 1 | k – 1) is the covariance of the previous covariance of state X(k – 1 | k – 1) ; P(k | k – 1) is the covariance corresponding to X(k | k – 1).

Since the gyroscope noise will have a great disturbance to the accuracy of the measurement, and choosing to output the signal directly without any processing will result in a great deviation, and it is impossible to get an accurate measurement, so Kalman filtering algorithm is used to measure the output of the signal in the static state.

RC filtering algorithm: 21Xour0=Xout1+(Xin−Xout1)/num

X_in represents the input value; X_out1 represents the output value after the last filtering: X_out0 represents the result of this filtering:

num is called the filter coefficient, which is related to the RC value: the proportion of the sampled value is determined by the value of num in the current filtering result. The current sampled input value plays a corrective role in filtering, the smaller the filter coefficient num, the higher the sensitivity, the faster the response, but the filtering result is quite unstable; on the contrary, the response is slow, the filtering result is relatively smooth, but the sensitivity is relatively low.

The traditional Kalman filter, although very good in the field of filtering, is often tested experimentally for its computed Q and R, and is often assumed to be certain constants. However, in practical engineering, Q and R are usually time-varying. The adaptive Kalman filtering method mainly introduced in this paper, in order to reflect the difference between the measured and estimated values, can also be used to adaptively adjust the noise variance value using the sequence method, which in turn realizes the adaptive adjustment of the covariance matrix of the state estimation error under the Kalman filtering gain [27], ensures the filter performance and reduces the estimation error. Under the condition that the noise variance is unknown, real-time estimation of R can be performed.

Define the error between the measured and estimated values: 22λk=yk−(C^x^k|k−1+Duk) λ_k is the error between the measured and estimated values; y_k is the measured value; and Cx^k|k−1+Duk is the estimated value.

The theoretical covariance matrix can be defined as: 23δk=E[λkλkT]=CP¯kCT+Rk δ_k is the theoretical covariance matrix; E[] is the mean: P_k is the a posteriori estimation error covariance: R_k is the measurement noise average. In reality, the covariance is affected by modeling and measurement errors, so the actual covariance matrix is calculated as follows: 24δ^k=1M∑i=1M λk−1λk−1r

M represents the length of the moving window. After comparing the size of the theoretical covariance matrix δ_k with the actual covariance matrix δ^k , the adjustment of Q and R is realized. When δk>δ^k , it indicates that the set measurement noise variance is large in the Kalman filtering algorithm, which should be appropriately reduced R_k ; when δk>δ^k , it should be appropriately increased R_k, and in order to avoid filtering dispersion, R_k can be kept constant. The adjustment factor of definition R_k is: 25αk=max(1,tr(δ^k)tr(δk)) α_k is the adjustment factor for R_k ; k denotes the trace of the matrix. When updating the Kalman filter gain matrix, the following adaptive adjustments are made to R :

K_k = P_k|k–1C^T (CP_k|k–1C^T + a_kR_k)⁻¹ improves the estimation accuracy of the Kalman filtering No. method, the measurement noise variance matrix is generally adjusted inversely to the process noise variance matrix, and the following adaptive adjustment is made to Q when updating the a priori estimation error covariance matrix: 26Pk|k−1=APk−1AT+Qk-1/ak

Q is the covariance matrix of the system process noise.

The optimal state estimate x^k at the current moment is obtained by adaptively adjusting the noise variance through the sequence, thus realizing the adaptive adjustment of the Kalman filter gain and the state estimation error covariance matrix, ensuring the filter performance and reducing the estimation error, which reduces the interference of the gyroscope noise on the measurement accuracy.

In order to test the accuracy of the heart rate results of this system, 2 males and 2 females, aged between 24-32 years old, were selected for the test experiment, and the number of measurements was 4 times, and the measurement state was at rest. The system was worn on the right wrist, and the MAX30102 oximeter was placed at the fingertip of the right index finger, and the results of the heart rate experimental data measurements are shown in Table 1.

The results of the correlation analysis of the heart rate test are shown in Figure 7. The results show that the range of the tester’s heart rate is 66BPM-89BPM, and the correlation coefficient of this device and the comparison device is 0.9654, the mean value of the error is 0.2BPM, the absolute average error is 1.1BPM, and the 95% confidence interval is 0.19-1.045BPM.It can be concluded through the analysis that the correlation of the measurement data of this device and the comparison device is high and the data consistency is good, indicating that the measurement error of this device is ±2BPM. Equipment and the comparison equipment measurement data correlation is high, and the data consistency is good, the absolute average error is within 2BPM, indicating that the measurement error of this equipment is ±2BPM.

Figure 7.

Correlation analysis results of heart rate test

The results of the oximetry experimental data tests are shown in Table 2. The arterial oxygen saturation measurement comparison data concluded that the single-point arterial oxygen saturation error between this system and the comparison device was within ±1%, with extremely high correlation.

The human blood pressure values (placed in the left hand) were collected using this device and measured ten times continuously (the calculated values were kept to 2 decimal places), while the blood pressure values in the right hand were measured using a pulse wave electronic sphygmomanometer. The results of the experimental data testing of blood pressure parameters are shown in Table 3. From the comparison data of arterial blood pressure measurement in Table 3, it is concluded that the absolute maximum value of single-point systolic blood pressure error of this system and the comparison equipment is 5, and the absolute maximum value of diastolic blood pressure error is 11. The average absolute error of systolic blood pressure is 3.6 mmHg, and the standard deviation is 3.34 mmHg. The average absolute error of diastolic blood pressure is 6.8 mmHg, and the standard deviation is 6.64 mmHg. It can be seen that the systolic blood pressure and diastolic blood pressure The AAMI American Association for the Advancement of Medical Instrumentation (AAMI) standard: the standard deviation of the error is not more than 8 mmHg.

The results of the upper limb strength tests of the experimental and control groups before and after the experiment are shown in Table 4. The results show that, from between groups, the number of pull-ups of the experimental group and the control group before the experiment was 21.45 and 21.12 in turn, and the hanging feeler height (mean value of right and left hands) was 51.46 cm and 51.88 in turn.T-testing the data of the two groups before the experiment, we got the P-value of 0.664 and 0.781 in turn, which were greater than 0.05, that is to say, the maximal upper limb strength of the two groups before the experiment and the explosive strength were not significantly different and were at the same level.

Post-experimental T-test between the experimental and control groups for pull-ups and hanging feelers (left and right hand means) yielded P-values of 0.000 and 0.024, respectively, showing significant differences. From the perspective of the group, after 12 weeks of training, the experimental group’s post-test improvement values for the two indexes were 10.71 and 9.79 cm, respectively; the control group’s post-test improvement values for the two indexes were 5.33 and 3.66 cm, respectively; and the average scores of the upper specialized strength of the experimental group and the control group showed an upward trend as a whole. The results of the experimental group and the control group before and after the experiment were tested by the paired sample test: the P-value of the two test indexes of the experimental group and the control group were less than or equal to 0.01, which indicated that after 12 weeks of rock-climbing training, the upper extremity strength of the experimental group and the control group showed a very significant difference with the experimental group before the experiment and there was a significant increase in the strength of upper extremity specialization. The experimental group had a significantly higher increase rate of maximal weight-bearing pull-ups and hanging reach than the control group after the experiment, which resulted in significant differences.

The results of the comparison of the scores of the technical assessment items between the experimental group and the control group before and after the experiment are shown in Table 5. The results show that the scores of the specific technical assessment items of the experimental group and the control group before the experiment were statistically analyzed one by one by intergroup t-test, and the p-value of intergroup t-test for the six technical assessment items was obtained to be greater than 0.05, which indicated that these six types of techniques of the two groups were basically at the same level before the experiment. After the completion of the experiment, the students of the experimental group grew 17.68%, 16.75%, 15.55%, 17.70%, 15.35% and 16.27% in the 6 categories of technology in turn; the growth rate of the control group in the 6 categories of technology after the experiment was not more than 7.5%, indicating that the specific technical evaluation technology of the players of the two groups were improved after the experiment. Using the intergroup paired test to process and analyze the technical evaluation data of the two groups after the experiment, the P value of the intergroup test of the 6 types of technical items was obtained (P < 0.05), i.e., there was a significant difference between the experimental group and the control group after the experiment in the timing of the application of force and the control of the strength and the appropriateness of the adjustment of the center of gravity technique, and there was a very significant difference in the degree of stabilization of the fixed point, the accuracy of the probing technique, the skillfulness of the continuous movement and the smoothness of the course of the movement. Obviously, the sensing network application of college students’ health monitoring in rock climbing teaching proposed in this paper is effective.