WO2018218747A1 - Indoor positioning method and system - Google Patents

Abstract

Provided in the present invention are an indoor positioning method and system, the positioning method comprising: S1. a receiving step, receiving a footstep sound vibration signal which is collected by a probe; S2. a detection step, using a SWIM model to detect the footstep sound vibration signal and performing correction; S3. an estimation step, using a PCC algorithm to estimate a time delay; S4. a processing step, determining an accurate target location according to a positioning algorithm. The benefits of the present invention are as follows: the present invention collects footstep sound vibrations during walking by means of a seismic detector to achieve positioning of a person; footstep sound vibrations spread along a floor surface, thus not encountering furniture and like obstacles, which occurs when vibrations spread in the air and which produces multi-path reflection that influences positioning efficiency. In order to achieve accurate positioning, a SWIM model is designed in the present invention to detect footstep sounds and a PCC algorithm is designed to estimate time delays and to perform positioning.

Description

Indoor positioning method and system Technical field

The present invention relates to the field of information processing and indoor positioning technology, and in particular, to an indoor positioning method and system.

Background technique

The existing indoor positioning technologies include infrared technology, Bluetooth technology, computer vision technology, radio frequency identification technology, WIFI technology, ZigBee technology, UWB technology and ultrasonic technology. The technology with the highest positioning accuracy is UWB and laser technology, but the cost of UWB is too high. The laser is easily blocked. At present, indoor positioning systems mostly use RF or sound signals to achieve positioning, but in complex environments, these signals have serious multipath effects. Please refer to Figure 1, which illustrates the generation process of multipath effects; The multipath effect of the signal seriously interferes with the calculation of the distance, resulting in the positioning accuracy in the complex environment is generally not high; in order to reduce the multipath effect, the existing indoor positioning technology mostly requires an open indoor environment, but in reality, our There are many obstacles in the application scenario, so this has led to the delay in the application of indoor positioning technology to our lives. At the same time, in practical applications, due to the complex and varied indoor environment, many interference factors, signal detection It has also become a challenge, and even if the signal that needs to be located is successfully detected, it is a problem to accurately estimate the delay due to the influence of multipath effects.

Summary of the invention

The invention provides an indoor positioning method, comprising the following steps:

S1. receiving step, receiving a footstep sound vibration signal collected by the probe;

S2. detecting step, using SWIM model to detect footstep sound vibration signal and correcting;

S3. The estimating step uses a PCC algorithm to estimate the delay;

S4. Processing step, determining the precise position of the target according to the positioning algorithm.

As a further improvement of the present invention, the S2. detecting step comprises:

S21. The receiving head receives the footstep vibration signal and performs Wiener filtering on the signal;

S22. Calculating the short-time energy of the footstep vibration signal;

S23. Initially detecting the start and end points of the footsteps using a double threshold endpoint detection method based on short-term energy;

S24. Combining the footstep interval, assigning weights to re-adjust the starting point of the footsteps;

S25. Real-time correction of the SWIM model.

As a further improvement of the present invention, the S3. estimating step comprises:

S31. Detecting respective peaks of the footstep vibration signal x, and finding the position of the first peak, denoted as p ₁ ;

S32. Find the position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ;

S33. Take the smaller of p ₁ and p ₂ and record it as p _min ;

S34. According to experience, x takes n points from p ₁ forward and complements (p ₁ -p _min ) zeros for time synchronization. According to experience, n' points are taken backwards, and p ₁ constitutes signal 1. Recorded as sig ₁ ;

S35. Take n points from p ₂ forward, and make up time (p ₂ -p _min ) zeros for time synchronization, then take n' points backward, and p ₂ constitutes signal 2, recorded as Sig ₂ ;

S36. Find the lengths of sig ₁ and sig ₂ respectively, and record that the larger values of l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ;

. S37 rearwardly to fill _a sig (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align sig sig ₁ and ₂ of the length;

S38. Using GCC to perform delay estimation on sig ₁ and sig ₂ .

As a further improvement of the present invention, the S4. processing steps include:

S41. Determine, by using an S3. estimation step, a time difference that the footstep vibration signal reaches three or more receiving heads;

S42. The algorithm for positioning using the time difference of arrival calculates the precise coordinates of the footstep vibration signal.

As a further improvement of the present invention, in the step S23, the method includes:

S231. Setting the maximum value of the short-time energy is the low threshold of the energy, and setting the maximum value of the footstep energy to 1/2 is the high threshold of the energy;

S232. Set two parameters: the longest length of the silence and the shortest length of the signal;

S233. The entire endpoint detection is divided into four segments: a silent segment, a transition segment, a signal segment, and an end;

S234. The program uses a variable Status to indicate the current state;

S235. The initial segment of the signal is a silent segment. If the short-term energy exceeds the low threshold, the starting point is marked and the transition segment is entered;

S236. In the transition section, it is not sure that it is in the signal segment, and if there is a short-term energy falling below the low threshold and exceeding the maximum mute length, it is restored to the mute state;

S237. If there is a high threshold in the transition section, be sure to enter the signal segment;

S238. If the length of the final segment is less than the minimum signal length, it is considered to be noise and discarded;

S239. The first and last sample points cut out are marked as x ₀ and x'₀;

S2310. Cut out the start and end points of the second footstep, labeled x ₁ and x'₁; and get the interval d _{1 of the} first two footsteps (d ₁ = x _{1 -} x' ₀ );

S2311. Obtaining the starting point x _{i of the} i+1 footsteps, the ending point x' _i and the interval d _i+1 , where i≥0;

In the step S24, in combination with the footstep interval d _i+1 , the weight is assigned to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows:

y _i =M _i-1 x' _i +N _i-1 d _i-1 (2)

_{M i-1 = (1/2)} ^ i (3)

M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the iterative process of EMA, the proportion of M decreases rapidly according to the exponential decay model;

In the step S25, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking speed greatly We let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N.

The invention also provides an indoor positioning system, comprising:

a receiving module, configured to receive a footstep vibration signal collected by the probe;

a detection module for detecting and correcting a footstep vibration signal by using a SWIM model;

An estimation module for estimating a delay using a PCC algorithm;

A processing module is configured to determine a precise position of the target according to the positioning algorithm.

As a further improvement of the present invention, the detecting module comprises:

a first detecting module, configured to receive a footstep vibration signal by the receiving head, and perform Wiener filtering thereon;

a second detecting module, configured to calculate a short-time energy of the footstep vibration signal;

a third detecting module, configured to initially detect a starting point and an ending point of the footstep sound by using a double threshold end point detection method based on short-time energy;

a fourth detecting module, configured to combine the footstep interval and assign a weight to re-adjust the starting point of the footstep sound;

The fifth detection module is used for real-time correction of the SWIM model.

As a further improvement of the present invention, the estimating module comprises:

a first estimating module, configured to detect respective peaks of the footstep vibration signal x, and find the position of the first peak, denoted as p ₁ ;

a second estimating module, configured to find a position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ;

a third estimation module for taking the smaller of p ₁ and p ₂ , denoted as p _min ;

The fourth estimation module is configured to take n points from p ₁ forward according to experience, and perform time synchronization by adding (p ₁ -p _min ) zeros forward, and take n' points backward according to experience, and p ₁ constitutes signal 1, recorded as sig ₁ ;

a fifth estimation module, configured to take n points from p ₂ forward for x′, and perform time synchronization by adding (p ₂ -p _min ) zeros forward, and then taking n′ points backward, and consisting of p ₂ Signal 2, denoted as sig ₂ ;

a sixth estimating module, configured to respectively determine the lengths of sig ₁ and sig ₂ , and the larger values recorded as l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ;

Seventh estimation module configured to complement the length of the rearwardly sig ₁ (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align the sig ₁ and sig _2;

An eighth estimation module is configured to perform delay estimation on sig ₁ and sig ₂ by using GCC.

As a further improvement of the present invention, the processing module includes:

a first processing module, configured to determine, by using the estimation module, a time difference that the footstep vibration signal reaches three or more receiving heads;

The second processing module is configured to calculate an accurate coordinate of the footstep vibration signal by using an algorithm for positioning the time difference of arrival.

As a further improvement of the present invention, the third detecting module comprises:

a first detection processing unit, configured to set a maximum value of the short-time energy of the noise as a low threshold of the energy, and set a maximum of 1/2 of the maximum value of the footstep energy as a high threshold of the energy;

a second detection processing unit for setting two parameters: a longest length of the silence and a shortest length of the signal;

The third detection processing unit is configured to divide the entire endpoint into four segments: a silent segment, a transition segment, a signal segment, and an end;

The fourth detection processing unit, the program uses a variable Status to indicate the current state;

The fifth detection processing unit, the initial segment of the signal is a silent segment, and if the short-term energy exceeds the low threshold, the marker starting point is entered and the transition segment is entered;

The sixth detection processing unit, in the transition section, cannot be sure that it is in the signal segment, and if there is a short-term energy falling below the low threshold and exceeding the maximum silence length, the mode is restored to the mute state;

The seventh detection processing unit, if there is a high threshold in the transition section, is sure to enter the signal segment;

The eighth detection processing unit considers that the noise is discarded if the length of the final segment is less than the minimum signal length;

The ninth detection processing unit, the first and last sample points cut out are marked as x ₀ and x'₀;

The tenth detection processing unit cuts out the start and end points of the second footstep, and marks them as x ₁ and x'₁; and simultaneously obtains the interval d _{1 of the} first two footsteps (d ₁ = x _{1 -} x' ₀ );

The eleventh detection processing unit obtains a starting point x _{i of the} i+1 footsteps, an end point x' _i and an interval d _{i+1 thereof} , wherein i≥0;

In the fourth detecting module, combining the footstep interval d _i+1 , assigning a weight to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows:

y _i =M _i-1 x' _i +N _i-1 d _i-1 (2)

M _i-1 =(1/2)^i (3)

M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the iterative process of EMA, the proportion of M decreases rapidly according to the exponential decay model;

In the fifth detection module, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking greatly At the speed, we let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N.

The invention has the beneficial effects that the invention can realize high-precision indoor positioning in a complex environment with a positioning accuracy of up to 7 cm; the proposed footstep sound detecting method (SWIM model) can realize a complicated environment without prior training. The detection and correction of the footstep vibration signal has a detection accuracy of up to 98%. The proposed time delay estimation algorithm (PCC) solves the problem that the generalized cross-correlation method (GCC) has a large delay estimation error due to multipath effect.

DRAWINGS

Figure 1 is a schematic diagram of the generation process of the multipath effect.

Figure 2 is a graph of various peaks of the footstep vibration signal.

Figure 3 is a schematic diagram of an indoor positioning method in a complex indoor environment by human footstep vibration signals.

detailed description

As shown in FIG. 3, in the present invention, three or more connections are arranged around the indoor floor. Receiving the probe, the invention discloses an indoor positioning method, comprising the following steps:

S1. receiving step, receiving a footstep sound vibration signal collected by the probe;

S2. detecting step, using SWIM model to detect footstep sound vibration signal and correcting;

S3. The estimating step uses a PCC algorithm to estimate the delay;

S4. Processing step, determining the precise position of the target according to the positioning algorithm.

In the S1. receiving step, the footstep vibration signal includes a human footstep vibration signal, an animal's footstep vibration signal, and a footstep vibration signal generated by the artificial machine.

The receiving probe can be a geophone Geophone or other vibration signal sensor.

In practical applications, the indoor environment positioning method and system of the present invention can be deployed on the ground of a complex indoor environment.

SWIM model: The walking weight based adaptive Weight Increment Model (SWIM).

PCC algorithm: Maximum cross-correlation algorithm Peak Cross Correlation (PCC).

Arrange the geophone Geophone, including:

Three Geophones are placed around the ground;

The next level of the Geophone signal amplifier is used to amplify the footstep vibration signal collected by the Geophone;

The lower stage of the signal amplifier is connected to the AD analog-to-digital converter for converting the acquired footstep vibration signal into a digital signal;

The next level of the AD analog-to-digital converter is connected to the Raspberry Pi to control the acquisition and preservation of footstep vibration signals.

The S2. detecting step includes:

S21. The receiving head receives the footstep vibration signal and performs Wiener filtering on the signal;

S22. Calculate the short-term energy of the footstep vibration signal. The formula for calculating the short-time energy is as follows:

S23. Initially detecting the start and end points of the footsteps using a double threshold endpoint detection method based on short-term energy;

S24. Combining the footstep interval, assigning weights to re-adjust the starting point of the footsteps;

S25. Real-time correction of the SWIM model.

The S3. estimating step includes:

S31. Detecting respective peaks of the footstep vibration signal x, as shown in FIG. 2, and finding the position of the first peak, denoted as p ₁ ;

S32. Find the position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ;

S33. Take the smaller of p ₁ and p ₂ and record it as p _min ;

S34. According to experience, x takes n points from p ₁ forward and complements (p ₁ -p _min ) zeros for time synchronization. According to experience, n' points are taken backwards, and p ₁ constitutes signal 1. Recorded as sig ₁ ;

S35. In the same manner as for the x 'taken from p ₂ n points forward, forward and fill (p ₂ -p _min) zero for time synchronization, again after taking n' points, and the constituent signal p ₂ 2 , recorded as sig ₂ ;

S36. Find the lengths of sig ₁ and sig ₂ respectively, and record that the larger values of l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ;

. S37 rearwardly to fill _a sig (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align sig sig ₁ and ₂ of the length;

S38. Using GCC to perform time delay estimation on sig ₁ and sig ₂ , the time difference is recorded as □t ₁ ;

S39. In the same way, we can estimate the time difference between the arrival of the footsteps to Geophone 1 and Geophone 3, recorded as □t ₂ .

The S4. processing steps include:

S41. Determine, by using an S3. estimation step, a time difference that the footstep vibration signal reaches three or more receiving heads;

S42. Calculate the precise coordinates of the footstep vibration signal by using the TDOA, TOA three-point positioning algorithm, or other algorithms that use the time difference of arrival to locate.

In the step S23, the method includes:

S231. Setting the maximum value of the short-time energy is the low threshold of the energy, and setting the maximum value of the footstep energy to 1/2 is the high threshold of the energy;

S232. Set two parameters: the longest length of the silence and the shortest length of the signal;

S233. The entire endpoint detection is divided into four segments: a silent segment, a transition segment, a signal segment, and an end;

S234. The start of the acquired footstep vibration signal is a silent segment, and the program uses a variable Status to indicate the current state;

S235. The initial segment of the signal is a silent segment. If the short-term energy exceeds the low threshold E _s , the starting point is marked and the transition segment is entered;

S236. In the transition section, it is not sure that the signal segment is in the signal segment. If the short-term energy falls below the low threshold E _s and exceeds the maximum silent length S _m , the mute state is restored;

S237. If there is a high threshold E _h in the transition section, be sure to enter the signal segment;

S238. If the length of the final segment is less than the minimum signal length L _m , it is considered to be noise and discarded;

. S239 first and the last point of a sample cut out, labeled as x ₀ and x _'0;

S2310. Cut out the start and end points of the second footstep, labeled x ₁ and x'₁; and get the interval d _{1 of the} first two footsteps (d ₁ = x _{1 -} x' ₀ );

S2311. Obtain the starting point x _{i of the} i+1 footsteps, the end point x' _i and its interval d _i+1 , where _i ≥ 0.

In the step S24, in combination with the footstep interval d _i+1 , the weight is assigned to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows:

y _i =M _i-1 x' _i +N _i-1 d _i-1 (2)

M _i-1 =(1/2)^i (3)

M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the EMA iterative process, the proportion of M decreases rapidly according to the exponential decay model.

In the step S25, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking speed greatly We let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N.

而一个人的步幅一般不超过一米，所以把脚步声间距设成一个向量

模长为1，如果定位的预测坐标与实际坐标的欧拉距离E_d大于

时，则认为行人突然大幅度改变行走速度，重新分配M、N的权重，恢复初始值。欧式距离的计算公式如下：Real-time correction of the SWIM model. In order to judge whether the pedestrian suddenly changes the walking speed and causes the cutting error, we introduce the vector.

And a person's stride usually does not exceed one meter, so set the footstep spacing to a vector.

The modulus length is 1, if the Euler distance E _d between the predicted coordinates of the positioning and the actual coordinates is greater than

At that time, it is considered that the pedestrian suddenly changes the walking speed greatly, redistributes the weights of M and N, and restores the initial value. The formula for calculating the Euclidean distance is as follows:

The invention also discloses an indoor positioning system, comprising:

a receiving module, configured to receive a footstep vibration signal collected by the probe;

a detection module for detecting and correcting a footstep vibration signal by using a SWIM model;

An estimation module for estimating a delay using a PCC algorithm;

A processing module is configured to determine a precise position of the target according to the positioning algorithm.

The detection module includes:

a first detecting module, configured to receive a footstep vibration signal by the receiving head, and perform Wiener filtering thereon;

a second detecting module for calculating short-term energy of the footstep vibration signal, short-time energy Calculated as follows:

a third detecting module, configured to initially detect a starting point and an ending point of the footstep sound by using a double threshold end point detection method based on short-time energy;

a fourth detecting module, configured to combine the footstep interval and assign a weight to re-adjust the starting point of the footstep sound;

The fifth detection module is used for real-time correction of the SWIM model.

The estimation module includes:

a first estimating module for detecting respective peaks of the footstep vibration signal x, as shown in FIG. 2, and finding the position of the first peak, denoted as p ₁ ;

a second estimating module, configured to find a position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ;

a third estimation module for taking the smaller of p ₁ and p ₂ , denoted as p _min ;

The fourth estimation module is configured to take n points from p ₁ forward according to experience, and perform time synchronization by adding (p ₁ -p _min ) zeros forward, and take n' points backward according to experience, and p ₁ constitutes signal 1, recorded as sig ₁ ;

a fifth estimation module, configured to take n points from p ₂ forward for x′, and perform time synchronization by adding (p ₂ -p _min ) zeros forward, and then taking n′ points backward, and consisting of p ₂ Signal 2, denoted as sig ₂ ;

a sixth estimating module, configured to respectively determine the lengths of sig ₁ and sig ₂ , and the larger values recorded as l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ;

Seventh estimation module configured to complement the length of the rearwardly sig ₁ (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align the sig ₁ and sig _2;

An eighth estimation module is configured to perform delay estimation on sig ₁ and sig ₂ by using GCC.

The processing module includes:

a first processing module, configured to determine, by using the estimation module, a time difference that the footstep vibration signal reaches three or more receiving heads;

The second processing module is configured to calculate an accurate coordinate of the footstep vibration signal by using an algorithm for positioning the time difference of arrival.

The third detecting module includes:

a first detection processing unit, configured to set a maximum value of the short-time energy of the noise as a low threshold of the energy, and set a maximum of 1/2 of the maximum value of the footstep energy as a high threshold of the energy;

a second detection processing unit for setting two parameters: a longest length of the silence and a shortest length of the signal;

The third detection processing unit is configured to divide the entire endpoint into four segments: a silent segment, a transition segment, a signal segment, and an end;

The fourth detection processing unit, the program uses a variable Status to indicate the current state;

The fifth detection processing unit, the initial segment of the signal is a silent segment, and if the short-term energy exceeds the low threshold, the marker starting point is entered and the transition segment is entered;

The sixth detection processing unit, in the transition section, cannot be sure that it is in the signal segment, and if there is a short-term energy falling below the low threshold and exceeding the maximum silence length, the mode is restored to the mute state;

The seventh detection processing unit, if there is a high threshold in the transition section, is sure to enter the signal segment;

The eighth detection processing unit considers that the noise is discarded if the length of the final segment is less than the minimum signal length;

The ninth detection processing unit, the first and last sample points cut out are marked as x ₀ and x'₀;

A tenth detection processing unit, the second cut-out start and end of footsteps, labeled x ₁ and x _'1; footsteps simultaneously obtain the first two spacing _{d 1 (d 1 = x 1} -x'0);

The eleventh detection processing unit obtains a starting point x _i , an ending point x' _{i of} i+1 footsteps, and an interval d _{i+1 thereof} , where _i ≥ 0.

In the fourth detecting module, combining the footstep interval d _i+1 , assigning a weight to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows:

y _i =M _i-1 x' _i +N _i-1 d _i-1 (2)

_{M i-1 = (1/2)} ^ i (3)

M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the EMA iterative process, the proportion of M decreases rapidly according to the exponential decay model.

In the fifth detection module, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking greatly At the speed, we let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N.

The invention realizes the positioning of the human by collecting the footstep vibration of the walking by the geophone. The footstep vibration propagates along the ground, and does not encounter obstacles such as furniture when traveling in the air to cause multipath reflection to affect the positioning effect. In order to achieve accurate positioning, the present invention designs a SWIM model to detect footsteps, and designs a PCC algorithm to estimate delay and position.

The above is a further detailed description of the present invention in connection with the specific preferred embodiments, and the specific embodiments of the present invention are not limited to the description. It will be apparent to those skilled in the art that the present invention may be made without departing from the spirit and scope of the invention.

Claims (10)

An indoor positioning method, comprising the steps of: S1. receiving step, receiving a footstep sound vibration signal collected by the probe; S2. detecting step, using SWIM model to detect footstep sound vibration signal and correcting; S3. The estimating step uses a PCC algorithm to estimate the delay; S4. Processing step, determining the precise position of the target according to the positioning algorithm. The positioning method according to claim 1, wherein the S2. detecting step comprises: S21. The receiving head receives the footstep vibration signal and performs Wiener filtering on the signal; S22. Calculating the short-time energy of the footstep vibration signal; S23. Initially detecting the start and end points of the footsteps using a double threshold endpoint detection method based on short-term energy; S24. Combining the footstep interval, assigning weights to re-adjust the starting point of the footsteps; S25. Real-time correction of the SWIM model. The positioning method according to claim 1, wherein the S3. estimating step comprises: S31. Detecting respective peaks of the footstep vibration signal x, and finding the position of the first peak, denoted as p ₁ ; S32. Find the position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ; S33. Take the smaller of p ₁ and p ₂ and record it as p _min ; S34. According to experience, x takes n points from p ₁ forward and complements (p ₁ -p _min ) zeros for time synchronization. According to experience, n' points are taken backwards, and p ₁ constitutes signal 1. Recorded as sig ₁ ; S35. Take n points from p ₂ forward, and make up time (p ₂ -p _min ) zeros for time synchronization, then take n' points backward, and p ₂ constitutes signal 2, recorded as Sig ₂ ; S36. Find the lengths of sig ₁ and sig ₂ respectively, and record that the larger values of l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ; . S37 rearwardly to fill _a sig (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align sig sig ₁ and ₂ of the length; S38. Using GCC to perform delay estimation on sig ₁ and sig ₂ . The positioning method according to claim 1, wherein the S4. processing step comprises: S41. Determine, by using an S3. estimation step, a time difference that the footstep vibration signal reaches three or more receiving heads; S42. The algorithm for positioning using the time difference of arrival calculates the precise coordinates of the footstep vibration signal. The positioning method according to claim 2, wherein: In the step S23, the method includes: S231. Setting the maximum value of the short-time energy is the low threshold of the energy, and setting the maximum value of the footstep energy to 1/2 is the high threshold of the energy; S232. Set two parameters: the longest length of the silence and the shortest length of the signal; S233. The entire endpoint detection is divided into four segments: a silent segment, a transition segment, a signal segment, and an end; S234. The program uses a variable Status to indicate the current state; S235. The initial segment of the signal is a silent segment. If the short-term energy exceeds the low threshold, the starting point is marked and the transition segment is entered; S236. In the transition section, it is not sure that it is in the signal segment, and if there is a short-term energy falling below the low threshold and exceeding the maximum mute length, it is restored to the mute state; S237. If there is a high threshold in the transition section, be sure to enter the signal segment; S238. If the length of the final segment is less than the minimum signal length, it is considered to be noise and discarded; S239. The first and last sample points cut out are marked as x ₀ and x'₀; S2310. Cut out the start and end points of the second footstep, labeled x ₁ and x'₁; and get the interval d _{1 of the} first two footsteps (d ₁ = x _{1 -} x' ₀ ); S2311. Obtaining the starting point x _{i of the} i+1 footsteps, the ending point x' _i and the interval d _i+1 , where i≥0; In the step S24, in combination with the footstep interval d _i+1 , the weight is assigned to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows: y _i =M _i-1 x' _i +N _i-1 d _i-1 (2) M _i-1 =(1/2)^i (3) M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the iterative process of EMA, the proportion of M decreases rapidly according to the exponential decay model; In the step S25, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking speed greatly We let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N. An indoor positioning system, comprising: a receiving module, configured to receive a footstep vibration signal collected by the probe; a detection module for detecting and correcting a footstep vibration signal by using a SWIM model; An estimation module for estimating a delay using a PCC algorithm; A processing module is configured to determine a precise position of the target according to the positioning algorithm. The positioning system according to claim 6, wherein the detecting module comprises: a first detecting module, configured to receive a footstep vibration signal by the receiving head, and perform Wiener filtering thereon; a second detecting module, configured to calculate a short-time energy of the footstep vibration signal; a third detecting module, configured to initially detect a starting point and an ending point of the footstep sound by using a double threshold end point detection method based on short-time energy; a fourth detecting module, configured to combine the footstep interval and assign a weight to re-adjust the starting point of the footstep sound; The fifth detection module is used for real-time correction of the SWIM model. The positioning system according to claim 6, wherein the estimating module comprises: a first estimating module, configured to detect respective peaks of the footstep vibration signal x, and find the position of the first peak, denoted as p ₁ ; a second estimating module, configured to find a position of the first peak of the same footstep x' received by another receiving head, denoted as p ₂ ; a third estimation module for taking the smaller of p ₁ and p ₂ , denoted as p _min ; The fourth estimation module is configured to take n points from p ₁ forward according to experience, and perform time synchronization by adding (p ₁ -p _min ) zeros forward, and take n' points backward according to experience, and p ₁ constitutes signal 1, recorded as sig ₁ ; a fifth estimation module, configured to take n points from p ₂ forward for x′, and perform time synchronization by adding (p ₂ -p _min ) zeros forward, and then taking n′ points backward, and consisting of p ₂ Signal 2, denoted as sig ₂ ; a sixth estimating module, configured to respectively determine the lengths of sig ₁ and sig ₂ , and the larger values recorded as l ₁ , l ₂ , l ₁ , and l ₂ are denoted as l _max ; Seventh estimation module configured to complement the length of the rearwardly sig ₁ (l _max -l ₁₎ zeros to fill rearwardly sig ₂ (l _max -l ₂₎ zeros to align the sig ₁ and sig _2; An eighth estimation module is configured to perform delay estimation on sig ₁ and sig ₂ by using GCC. The positioning system according to claim 6, wherein the processing module comprises: a first processing module, configured to determine, by using the estimation module, a time difference that the footstep vibration signal reaches three or more receiving heads; The second processing module is configured to calculate an accurate coordinate of the footstep vibration signal by using an algorithm for positioning the time difference of arrival. The positioning system of claim 7 wherein: The third detecting module includes: a first detection processing unit, configured to set a maximum value of the short-time energy of the noise as a low threshold of the energy, and set a maximum of 1/2 of the maximum value of the footstep energy as a high threshold of the energy; a second detection processing unit for setting two parameters: a longest length of the silence and a shortest length of the signal; The third detection processing unit is configured to divide the entire endpoint into four segments: a silent segment, a transition segment, a signal segment, and an end; The fourth detection processing unit, the program uses a variable Status to indicate the current state; The fifth detection processing unit, the initial segment of the signal is a silent segment, and if the short-term energy exceeds the low threshold, the marker starting point is entered and the transition segment is entered; The sixth detection processing unit, in the transition section, cannot be sure that it is in the signal segment, and if there is a short-term energy falling below the low threshold and exceeding the maximum silence length, the mode is restored to the mute state; The seventh detection processing unit, if there is a high threshold in the transition section, is sure to enter the signal segment; The eighth detection processing unit considers that the noise is discarded if the length of the final segment is less than the minimum signal length; The ninth detection processing unit, the first and last sample points cut out are marked as x ₀ and x'₀; The tenth detection processing unit cuts out the start and end points of the second footstep, and marks them as x ₁ and x'₁; and simultaneously obtains the interval d _{1 of the} first two footsteps (d ₁ = x _{1 -} x' ₀ ); The eleventh detection processing unit obtains a starting point x _{i of the} i+1 footsteps, an end point x' _i and an interval d _{i+1 thereof} , wherein i≥0; In the fourth detecting module, combining the footstep interval d _i+1 , assigning a weight to re-adjust the starting point of the footstep sound to y _i ( _i ≥ 2), and the adjustment formula is as follows: y _i =M _i-1 x' _i +N _i-1 d _i-1 (2) M _i-1 =(1/2)^i (3) M _i +N _i =1 (4)

Where M and N are weights, and i is an increasing function with a step number of 1, increasing by 1 each time it is cut, because human walking is not absolutely uniform, and the latest footstep interval is better predictive of the next footstep. Interval, so we choose EMA function to calculate d _i , as shown in formula (5), and because the more backward, the more stable and reliable the speed, with the iterative process of EMA, the proportion of M decreases rapidly according to the exponential decay model; In the fifth detection module, the SWIM model is corrected in real time, because when the pedestrian suddenly changes the walking speed greatly, for example, from the walking state to the running, the segmentation causes a deviation, so when the person suddenly changes the walking greatly At the speed, we let M and N recover the initial values and start to re-use the exponential growth model to assign the weights of M and N.

Images