CN104457781A - Self-adaption step number detection method based on single-axis accelerometer - Google Patents

Abstract

An adaptive step detection method based on a single-axis accelerometer, which uses a single acceleration sensor and fixes it on the back of the pedestrian's waist, uses this accelerometer to measure the pedestrian's acceleration in the vertical direction, and according to the periodic fluctuation of the human body's center of gravity in the vertical direction Change characteristics, by identifying the number of crests and troughs of the output signal of the accelerometer, the detection of the number of human walking steps can be realized. In order to effectively solve the problem of step detection rate decline caused by vibration and other factors, a dual-window step detection method is proposed, and a window length adaptive adjustment criterion is proposed according to the human body motion state, so as to realize the accurate detection of pedestrian steps at different walking speeds . The invention is simple, effective, easy to implement, and has certain robustness against interference such as vibration.

Description

An Adaptive Step Count Detection Method Based on Single-axis Accelerometer

technical field

The invention relates to an adaptive step detection method based on a single-axis accelerometer, which can effectively resist interference caused by factors such as vibration, and realize accurate counting of walking steps of a human body. It belongs to the field of pedestrian navigation and can be used to improve pedestrian navigation in pedestrian navigation. positioning accuracy.

Background technique

Pedestrian navigation based on microelectromechanical system (MEMS) inertial sensors is an emerging technology in the field of inertial technology in recent years. It is usually used in firefighting, disaster relief, blind guidance, large Personnel scheduling in buildings, etc. In the case of radio, the application of radio to pedestrian positioning is relatively mature and has been widely used. When radio signals cannot be obtained, motion sensors are usually used in combination with inertial technology and human dynamics to achieve autonomous positioning of pedestrians, and MEMS inertial sensors are commonly used motion sensors. Due to the large device error of the MEMS inertial sensor and the complexity of pedestrian walking movements, it is difficult to autonomously locate pedestrians based on the MEMS inertial sensor. Because the MEMS inertial sensor is placed on the waist for pedestrian positioning, it is easy to use, so the waist-based pedestrian positioning has become a research hotspot.

Aiming at the problem of how to accurately detect the number of steps in the waist-type pedestrian positioning, most of the existing research is to use the ups and downs of the center of gravity when the pedestrian walks to detect the number of steps, that is, to detect the number of steps through the output signal of the skyward accelerometer. However, various vibrations will be generated when the feet land on the ground and other disturbances during walking, which will lead to the introduction of various interference signals to the sky-to-air accelerometer, which will reduce the accuracy of detecting the number of steps. For waist-type pedestrian positioning, how to resist the interference of pedestrians' vibration when walking to improve the detection rate of steps is a difficult point in the field of pedestrian navigation and a problem that needs to be solved urgently.

Contents of the invention

Purpose of the present invention: In order to effectively solve the problem of step detection rate decline in pedestrian navigation caused by interference factors such as vibration, a double-window step detection method is proposed on the basis of studying the fluctuation and variation of the center of gravity of the human body under the state of motion, and according to the motion state of the human body A self-adaptive adjustment criterion for the window length is proposed to realize accurate detection of the number of steps of pedestrians at different walking speeds.

The technical solution of the present invention is to find out the characteristic quantity that can represent the number of walking steps by analyzing the variation rule of the center of gravity of the human body in the state of exercise. Further analyze the disturbance factors of this feature quantity in the actual movement, and propose an adaptive step detection method based on a single-axis accelerometer. This method is characterized by: introducing two one-dimensional Array (referred to as window), the length of the window L is the length of the array, that is, the number of sampling points contained, and the two windows overlap only one sampling point at the beginning and the end; the number of peaks of the acceleration in the direction of the sky when the pedestrian is walking is identified through the double window, and through the human body movement The state adaptively adjusts the window length, which can effectively resist the interference signal generated by vibration during walking, and realize the accurate detection of pedestrian steps.

Utilize the described self-adaptive steps detection method based on uniaxial accelerometer, comprise the following steps:

(1) Fix the uniaxial accelerometer on the back of the waist of the human body and use it to measure the celestial acceleration of pedestrians;

(2) When initializing, set the gait cycle (time for walking one step) T(1)=0.2 seconds, the maximum number of peaks of the accelerometer signal N1=0, the minimum number of valleys of the accelerometer signal N2=0, and the number of steps of walking N=0 ;

(3) Utilize the gait period T(N) and the system sampling frequency fs to adaptively adjust the length L(N) of the two windows;

(4) Calculate the maximum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the maximum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N1=N1+1, otherwise N1 remains unchanged , and record the timing j at which the maximum peak appears, and store it in maxpeak_time(N1);

(5) Calculate the minimum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the minimum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N2=N2+1, otherwise N2 remains unchanged , and record the timing j at which the minimum peak appears, and store it in minpeak_time(N2);

(6) When N1=N2 and N1>0, the step number N=N+1, and calculate the gait period T(N), otherwise N and T(N) remain unchanged, so as to realize the adaptive step detection.

The principle of the present invention: walking one step in the text refers to taking one step forward with one foot, and az represents the acceleration in the sky direction when the pedestrian walks. See Figure 1 for the corresponding relationship between the pedestrian's walking gait and the corresponding azimuth acceleration az without considering the disturbance such as vibration.

When the span of the human body’s two feet is the largest, the center of gravity of the human body is at the lowest position, and az reaches the minimum valley value; when the person continues to move forward, the position of the center of gravity of the human body begins to move upward, and az increases monotonously during this time; when the human body is standing upright state, the legs are almost vertical at this time, the center of gravity reaches the highest position, and az reaches the maximum peak value; then the center of gravity of the human body moves from the highest position to the lowest position with the pace, completing the action of walking one step forward, forming a cycle of change .

According to the fluctuation of the center of gravity in the vertical direction, it can be seen that when the center of gravity of the pedestrian is at the highest position within one step, the az signal always appears a corresponding peak, as shown in the red circle mark in Figure 2. When the pedestrian walks one step without considering vibration and other disturbances, az correspondingly appears a maximum peak. Therefore, the number of peaks Num_crest and the number of steps Num_step in az satisfy the following relationship:

Num_crest=Num_step (1)

If there is no influence such as vibration, the number of steps the pedestrian walks can be calculated by detecting the number of occurrences of the az peak Num_crest. It can also be seen from Figure 1 that the number of az troughs is also equal to the number of steps, and when a pedestrian walks a step, there is always an az peak and az trough appearing. Using this feature, it can be considered that walking is only considered when az peaks and troughs appear successively One step, which can improve the step detection rate. However, there are always various interference signals in the actual walking process, among which the vibration when landing has the greatest impact on the output of the accelerometer signal. These interference signals cause false peaks to appear in az in each step, see the green circle mark in Figure 2. These false peaks make the accuracy of step detection using az signal decrease. How to effectively resist vibration interference and improve the detection rate of step detection. To solve the problem of vibration disturbance, an adaptive double-window step detection method is proposed.

Two adaptive overlapping windows are used to detect the peak value of the azimuth acceleration az in each step, including the maximum peak value and minimum trough value. The basic principle of detection: the window refers to the one-dimensional array used to extract the peak value of az; the length of the window L is the length of the array, that is, the number of sampling points contained; the beginning and end of the two windows overlap only one sampling point, if the maximum peak value of the two windows is For the same number under the same time series, the only maximum value of the two windows is considered to be the maximum peak value of az in this step, and the minimum valley value of az can be obtained in the same way. Accompanying drawing 2 is a schematic diagram of the principle of dual-window detection of the peak value of the output signal az of the accelerometer.

However, if the window length L is too long, the number of detected steps will be too small. Similarly, if the window length is too short, the number of steps will be too large, which will not achieve the effect of accurately calculating the number of steps. In order to adaptively detect the number of steps taken by different people at different speeds and the maximum and minimum peak values of az. Further analysis of the experimental data shows that the window length is related to pedestrian walking speed, and the window length can be adaptively adjusted according to the pedestrian walking speed. For the sake of simplicity, this paper uses the gait cycle T (time to walk one step) to characterize the walking speed. After analysis, the relationship between the window length L and the gait cycle T can be obtained as follows:

L=T*fs (2)

Where fs is the sampling frequency of the system, formula (2) is more applicable to different people walking at different speeds, and the length L of the two overlapping windows is adaptively adjusted through the difference in walking speed. After a large number of experiments, the real numbers of different pedestrians walking at different speeds were collected, and it was found that the stride frequency (number of steps per unit time) of pedestrians in the initial stage is generally not more than 5 steps per second. For different pedestrians at different starting speeds The test results further summarize the following empirical relationship:

L=0.2*fs (3)

Because the walking speed of pedestrians is small at the beginning, and the stride frequency is within 5 steps per second, and then further changes on this basis, so the initial window length can be set by formula (3), and then used in the process of walking from formula (1) to Adapt to adjust window length.

Use this method to detect the maximum peak value and the minimum valley value of az in each step and the corresponding timing j (the time point when they appear). The gait period T of pedestrians can be calculated by the timing of the maximum peak value and the minimum trough value, and then the window length L can be adjusted with the gait period T to realize the adaptive step detection based on the single-axis accelerometer.

The advantage of the present invention compared with prior art is:

(1) Use dual windows to identify characteristic signals, effectively resist vibration and other interference, and achieve accurate detection of steps.

(2) Adaptively change the window length according to the different walking speeds of pedestrians, and then adaptively detect the number of steps.

Description of drawings

Figure 1 is a schematic diagram of the relationship between gait and accelerometer output signal az;

Fig. 2 is a schematic diagram of the principle of double-window detection accelerometer output signal az peak value;

Fig. 3 is a flow chart of the self-adaptive step count detection based on the single-axis accelerometer of the present invention.

Detailed ways:

Fig. 3 is the flow chart of self-adaptive steps detection of the present invention, mainly is divided into following several steps:

(1) Fix the uniaxial accelerometer on the back of the human body to ensure that the accelerometer and the human body are fixed as one without relative movement, and it is used to measure the acceleration of pedestrians;

(2) When the system is initialized, set the gait cycle (time for walking one step) T (1) = 0.2 seconds, the maximum number of peaks of the accelerometer signal N1 = 0, the minimum number of valleys of the accelerometer signal N2 = 0, and the number of walking steps N = 0; and use the formula (1) to obtain the double window length L(1) when the pedestrian starts;

(3) During walking, utilize gait cycle T (N) and system sampling frequency fs to adjust the length L (N) of two windows adaptively according to formula L=T*fs;

(4) Calculate the maximum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the maximum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N1=N1+1, otherwise N1 remains unchanged , and record the timing j at which the maximum peak appears, and store it in maxpeak_time(N1);

(5) Calculate the minimum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the minimum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N2=N2+1, otherwise N2 remains unchanged , and record the timing j at which the minimum peak appears, and store it in minpeak_time(N2);

(6) When N1=N2 and N1>0, the number of steps N=N+1, and use the gait cycle calculation formula: T(N+1)=0.5*(maxpeak_time(N1)-maxpeak_time(N1-1)+ minpeak_time(N2)-minpeak_time(N2-1))*(1/fs) calculates the gait cycle T(N+1), otherwise N and T(N) remain unchanged, so as to realize adaptive step detection.

The content not described in detail in this specification belongs to the prior art known to those skilled in the art.

Claims (4)

1. a kind of adaptive steps detection method based on uniaxial accelerometer, it is characterized in that comprising the following steps: (1) Fix the uniaxial accelerometer on the back of the waist of the human body and use it to measure the celestial acceleration of pedestrians; (2) When initializing, set the gait cycle (time for walking one step) T(1)=0.2 seconds, the maximum number of peaks of the accelerometer signal N1=0, the minimum number of valleys of the accelerometer signal N2=0, and the number of steps of walking N=0 ; (3) Utilize the gait period T(N) and the system sampling frequency fs to adaptively adjust the length L(N) of the two windows; (4) Calculate the maximum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the maximum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N1=N1+1, if not, then N1 unchanged, and record the timing j of the maximum peak value, and store it in maxpeak_time(N1); (5) Calculate the minimum peak value of the accelerometer output signal contained in each window in the two windows, and judge whether the minimum peak value of the two windows is the same number corresponding to the same time sequence, if yes, then N2=N2+1, otherwise N2 remains unchanged , and record the timing j at which the minimum peak appears, and store it in minpeak_time(N2); (6) When N1=N2 and N1>0, the step number N=N+1, and calculate the gait period T(N), otherwise N and T(N) remain unchanged, so as to realize the adaptive step detection. 2. the self-adaptive steps detection method based on uniaxial accelerometer according to claim 1, is characterized in that: in described step (1), require accelerometer to be placed on the back side of human body waist, and measure pedestrian's direction with accelerometer acceleration. 3. the adaptive steps detection method based on uniaxial accelerometer according to claim 1, is characterized in that: when initial, get gait cycle T (1)=0.2 second; In the step detection process, pedestrian's gait The period T(N) takes the period average value of the acceleration peak and trough signals; and adjusts the window length L(N) with the gait period T(N), which can realize the adaptive change of the window length according to the different walking speeds of pedestrians, thus Perform adaptive step detection. 4. the adaptive steps detection method based on uniaxial accelerometer according to claim 1, is characterized in that: the number of periodic peaks and valleys of the output signal of uniaxial accelerometer is used to characterize the number of walking steps, and is identified by double windows Peak and trough signals, effectively resist vibration and other interference, in order to achieve accurate detection of steps.