CN105636102A - Positioning method and device based on Bayes posterior probability - Google Patents

Abstract

The invention relates to the positioning field, in particular a positioning method and device based on Bayes posterior probability. Aiming at the problem existent in the prior art, the invention provides a processing method and device. The method comprises collecting training data in an environment without interference, the training data for modeling is worked out according to maximum likelihood method; since the model collected by the user is unknown, the possible signals under different mobile phones can be obtained through change (by translating the model); the positioning position is obtained through performing Bayes posterior probability maximization method on each extracted signal; furthermore, the indoor navigation path is obtained through the combination of the positioning method and cartographic information processing. Through the adoption of the method and device provided by the invention, the indoor positioning is finished through the cooperation of a network database module, an estimated value computing module and a Bayes posterior probability positioning module.

Description

A kind of localization method based on Bayes posterior probability and device

Technical field

The present invention relates to positioning field, especially a kind of localization method based on Bayes posterior probability and device.

Background technology

Existing algorithm for recognizing fingerprint, based on RSSI range measurement principle, is converted to signal intensity emission source to the distance between receptor, is then based on triangle positioning principle and positions.

Owing to WiFi signal anti-interference is poor, if there is the difference artificially blocking and receiving signal equipment, causing that the WiFi received has bigger error, thus causing error when signal intensity being converted to distance, finally positioning precision being produced impact.

Summary of the invention

The technical problem to be solved is: for prior art Problems existing, it is provided that a kind of localization method based on Bayes posterior probability and device, collects training data, set up model y in the environment not interfered with_ij=��_i+��_ilog(d_ij)+��_ij�� is calculated according to maximum likelihood method with training data_ij, ��_i, ��_i; What collect model due to user is unknown, and we are by changing ��_i, ��_i(it is translated) obtains signal possible under different mobile phone; The signal each time extracted is changed method the most significantly by Bayes posterior probability and obtains position location, further, processed by localization method combining cartographic information, obtain indoor navigation route.

The technical solution used in the present invention is as follows:

A kind of localization method based on Bayes posterior probability includes:

Step 1: in noiseless environment, it is some grid lattice points by field division, and be each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database about; Noiseless environment is collected the WiFi signal intensity level y of each grid lattice point by different mobile phone models_ij; The WiFi signal intensity level y that described mobile phone model, grid lattice point number value are corresponding with mobile phone_ijCorresponding;

Step 2: set up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; By changing ��^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^z; Wherein i is a certain mobile phone correspondence grid lattice point number value, and j is wireless router number value on a certain mobile phone correspondence grid lattice point, yi_jFor the WIFI signal that mobile phone receives from jth wireless router at i-th place; d_ijFor the i-th wireless router spacing to jth wireless router, for given value; ��_ijAt i-th grid lattice point, the measurement error of jth wireless router signal is obeyed N (�� for mobile phone_ij,��_ij) distribution, wherein ��_ijThe average of jth WIFI signal error, �� is collected for i-th grid lattice point_ijThe variance yields of jth WIFI signal error is collected for i-th grid lattice point;

Step 3: at positioning stage, collects m wireless router WIFI signal by mobile phone, sets wherein k WIFI signal noiseless, there is noisy signal in all the other m-k signals; Wherein,

$P\left(\mathrm{new}\right|i)=\underset{j=1}{\overset{k}{\mathrm{\Π}}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_{\mathrm{ij}}}\exp [-\frac{{(y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z-\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}})}^2}{2{\mathrm{\σ}}_{\mathrm{ij}}^2}\underset{j=k+1}{\overset{m}{\mathrm{\Π}}}\mathrm{\Φ}\left(\frac{y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}}}{{\mathrm{\σ}}_{\mathrm{ij}}}\right)$

Different �� is obtained for the training stage^zWith ��^z, from m ap, to take out k signal at random, substitute into Bayes posterior probability P (news | i), draw the maximized lattice point number value of posterior probability, the frequency of all lattice points numbering is designated as F; The mode of all lattice points numbering takes out, and is designated as P as deposit output one; Then mobile phone loads cartographic information, obtains indoor navigation route after all lattice point number value are processed.

Further, described step 2 sets up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij;

Specifically include:

Step 21: set up model y for mobile phone signal i_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, the signal data then collected according to different mobile phones sets up maximum likelihood function L: wherein Wherein t is different mobile phone models, and T is mobile phone quantitative value; N_tFor different mobile phone number of samples;

Step 22: with gradient descent method, it is thus achieved that ��^t, ��^t, ��_ij, ��_ij; Wherein ��^tFor the �� value of certain model mobile phone, ��^t�� value is received for certain signal;For receiving all wireless router signal values at a certain signal mobile phone at i point, for given value; I is total some bit number value, for given value; J is the sum of wireless router, for given value.

Further, by changing �� in described step 2^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^zDetailed process is: build ��^zFor interval [��₀,��_q] in uniform point, then ��^z=��₀+z*(��_q-��₀)/p; Wherein ��₀=min{ ��^t: t=1,2...T}, ��_q=max{ ��^t: t=1,2...T}, ��^zThere is p+1, [��₀,��_q] for comprising described ��^tInterval; T=1,2...T; In like manner produceZ is for uniformly putting number value, and p is for uniformly putting number.

Further, described step 3 specifically includes:

Step 111: based on posterior probability P described in claim 3 (news | i), obtainIndividual posterior probability P (news | i) the grid lattice point number value that maximum is corresponding, the frequency of all lattice point number value is designated as F; The mode then taking out all lattice point number value takes out, and lays in output valve P as first; Wherein i=1,2,3 ... I; News is all WiFi signal intensity level y that a certain moment mobile phone is collected_ij;

Step 112: based on right 111, obtains n first deposit output valve and is designated as the first global variable [P₁,P₂...P_n], corresponding frequency is designated as the second global variable [F₁,F₂...F_n];

Step 113: based on step 111 and step 112, mobile phone loads cartographic information; Owing to there is aisle on map, it is assumed that object moves only along road, and there is the setting of the upper limit and network database in translational speed, then the frequency distribution vector signal being currently received being sampled is weighted with the second global variableDraw final frequency distribution F, it is the second deposit output valve to the grid lattice values that frequency in final frequency distribution F is maximum, in addition update the second global variable does from the frequency distribution that current time is farthest, if the second deposit output valve meets: in the first global variable within the scope of adjacent two grid lattice values of all first deposit output valves, then output the second deposit output valve, and by the second deposit output valve the first global variable done and update; Otherwise export the n-th point in the first global variable; Described X represents position two, current location of the distance grid lattice point at most of next second; F_nFor frequency distribution vector;

Step 114: lay in output valve according to second and obtain indoor navigation route.

A kind of positioner based on Bayes posterior probability includes:

Mesh data library module, for being some grid lattice points by field division in noiseless environment, and be each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database about; Noiseless environment is collected the WiFi signal intensity level y of each grid lattice point by different mobile phone models_ij; The WiFi signal intensity level y that described mobile phone model, grid lattice point number value are corresponding with mobile phone_ijCorresponding;

Estimated value computing module, for setting up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; By changing ��^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^z; Wherein i is a certain mobile phone correspondence grid lattice point number value, and j is wireless router number value on a certain mobile phone correspondence grid lattice point, y_ijFor the WIFI signal that mobile phone receives from jth wireless router at i-th place; d_ijFor the i-th wireless router spacing to jth wireless router, for given value; ��_ijAt i-th grid lattice point, the measurement error of jth wireless router signal is obeyed N (�� for mobile phone_ij,��_ij) distribution, wherein ��_ijThe average of jth WIFI signal error, �� is collected for i-th grid lattice point_ijThe variance yields of jth WIFI signal error is collected for i-th grid lattice point;

Bayes posterior probability locating module, at positioning stage, collecting m wireless router WIFI signal by mobile phone, set wherein k WIFI signal noiseless, there is noisy signal in all the other m-k signals; Wherein,

$P\left(\mathrm{new}\right|i)=\underset{j=1}{\overset{k}{\mathrm{\Π}}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_{\mathrm{ij}}}\exp [-\frac{{(y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z-\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}})}^2}{2{\mathrm{\σ}}_{\mathrm{ij}}^2}\underset{j=k+1}{\overset{m}{\mathrm{\Π}}}\mathrm{\Φ}\left(\frac{y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}}}{{\mathrm{\σ}}_{\mathrm{ij}}}\right)$

Different �� is obtained for the training stage^zWith ��^z, from m ap, to take out k signal at random, substitute into Bayes posterior probability P (news | i), draw the maximized lattice point number value of posterior probability, the frequency of all lattice points numbering is designated as F; The mode of all lattice points numbering takes out, and is designated as P as deposit output one; Then mobile phone loads cartographic information, obtains indoor navigation route after all lattice point number value are processed.

Further, described estimated value computing module sets up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; Specifically include:

Step 21: set up model y for mobile phone signal i_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, the signal data then collected according to different mobile phones sets up maximum likelihood function L: wherein Wherein t is different mobile phone models, and T is mobile phone quantitative value; N_tFor different mobile phone number of samples;

Step 22: with gradient descent method, it is thus achieved that ��^t, ��^t, ��_ij, ��_ij; Wherein ��^tFor the �� value of certain model mobile phone, ��^t�� value is received for certain signal;For receiving all wireless router signal values at a certain signal mobile phone at i point, for given value; I is total some bit number value, for given value; J is the sum of wireless router, for given value.

Further, by changing �� in described estimated value computing module^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^zDetailed process is: build ��^zFor interval [��₀,��_q] in uniform point, then ��^z=��₀+z*(��_q-��₀)/p; Wherein ��₀=min{ ��^t: t=1,2...T}, ��_q=max{ ��^t: t=1,2...T}, ��^zThere is p+1, [��₀,��_q] for comprising described ��^tInterval; T=1,2...T; In like manner produceZ is for uniformly putting number value, and p is for uniformly putting number.

Further, described Bayes posterior probability locating module work process specifically includes:

Step 111: based on posterior probability P described in claim 3 (news | i), obtainIndividual posterior probability P (news | i) the grid lattice point number value that maximum is corresponding, the frequency of all lattice point number value is designated as F; The mode then taking out all lattice point number value takes out, and lays in output valve P as first; Wherein i=1,2,3 ... I; News is all WiFi signal intensity level y that a certain moment mobile phone is collected_ij;

Step 112: based on right 111, obtains n first deposit output valve and is designated as the first global variable [P₁,P₂...P_n], corresponding frequency is designated as the second global variable [F₁,F₂...F_n];

Step 113: based on step 111 and step 112, mobile phone loads cartographic information; Owing to there is aisle on map, it is assumed that object moves only along road, and there is the setting of the upper limit and network database in translational speed, then the frequency distribution vector signal being currently received being sampled is weighted with the second global variableDraw final frequency distribution F, it is the second deposit output valve to the grid lattice values that frequency in final frequency distribution F is maximum, in addition update the second global variable does from the frequency distribution that current time is farthest, if the second deposit output valve meets: in the first global variable within the scope of adjacent two grid lattice values of all first deposit output valves, then output the second deposit output valve, and by the second deposit output valve the first global variable done and update; Otherwise export the n-th point in the first global variable; Described X represents position two, current location of the distance grid lattice point at most of next second; F_nFor frequency distribution vector;

Step 114: lay in output valve according to second and obtain indoor navigation route.

In sum, owing to have employed technique scheme, the invention has the beneficial effects as follows:

1) it is some grid lattice points by field division, and be each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database and provide basis for follow-up location technology about;

2) owing to user mobile phone model is unknown, the present invention is by changing ��_i, ��_i(it is translated) obtains different mobile phone signal; Being positioned by the signal each time extracted, localization method is that Bayes posterior probability is turned to predictive value the most significantly so that positioning result is convenient accurately;

3) processed by localization method combining cartographic information, obtain indoor navigation route.

4) artificially blocking pure, using different mobile phone to collect under RST, locating accuracy is more than 90%.

Detailed description of the invention

All features disclosed in this specification, or the step in disclosed all methods or process, except mutually exclusive feature and/or step, all can combine by any way.

This specification (include any accessory claim, summary) disclosed in any feature, unless specifically stated otherwise, all can by other equivalences or there is the alternative features of similar purpose replaced. That is, unless specifically stated otherwise, each feature is an example in a series of equivalence or similar characteristics.

Embodiment one:

One, the training stage

1, it is some grid lattice points by field division, and is each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database about; Noiseless environment is collected the WiFi signal intensity level y of each grid lattice point by mobile phone_ij; The WiFi signal intensity level y that described mobile phone model, grid lattice point number value are corresponding with mobile phone_ijCorresponding;

2, WiFi signal fluctuation Normal Distribution

3, model y is set up for mobile phone t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; Wherein i is a certain mobile phone correspondence grid lattice point number value, and j is wireless router number value on a certain mobile phone correspondence grid lattice point, yi_jFor the WIFI signal that mobile phone receives from jth wireless router at i-th place; d_ijFor the i-th wireless router spacing to jth wireless router, for given value; ��_ijAt i-th grid lattice point, the measurement error of jth wireless router signal is obeyed N (�� for mobile phone_ij,��_ij) distribution, wherein ��_ijThe average of jth WIFI signal, �� is collected for i-th grid lattice point_ijThe variance yields of jth WIFI signal is collected for i-th grid lattice point. Specifically include:

Step 21: set up model y for mobile phone signal i_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, the signal data then collected according to different mobile phones sets up maximum likelihood function: Wherein t is different mobile phone models, and T is mobile phone quantitative value; N_tFor different mobile phone number of samples;

Step 22: with gradient descent method, it is thus achieved that ��^t, ��^t, ��_ij, ��_ij; Wherein ��^tFor the �� value of certain model mobile phone, ��^t�� value is received for certain signal;For receiving all wireless router signal values at a certain signal mobile phone at i point, for given value; I is total some bit number value, for given value; J is the sum of wireless router, for given value.

Two, positioning stage:

At positioning stage, collect m wireless router WIFI signal by mobile phone, set wherein k WIFI signal noiseless, all the other m-k signals exist noisy signal;

$P\left(\mathrm{new}\right|i)=\underset{j=1}{\overset{k}{\mathrm{\Π}}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_{\mathrm{ij}}}\exp [-\frac{{(y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z-\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}})}^2}{2{\mathrm{\σ}}_{\mathrm{ij}}^2}\underset{j=k+1}{\overset{m}{\mathrm{\Π}}}\mathrm{\Φ}\left(\frac{y_{\mathrm{ij}}-{\mathrm{\α}}^z-{\mathrm{\β}}^z\log \left(d_{\mathrm{ij}}\right)-{\mathrm{\μ}}_{\mathrm{ij}}}{{\mathrm{\σ}}_{\mathrm{ij}}}\right)$

Different �� is obtained for the training stage^zWith ��^z, m ap takes out k signal at random, substitutes into Bayes posterior probability P (news | i), draws the maximized lattice point number value of posterior probability, and described lattice point number value is indoor positioning position.

Step 111: based on posterior probability P described in claim 3 (news | i), obtainIndividual posterior probability P (news | i) the grid lattice point number value that maximum is corresponding, the frequency of all lattice point number value is designated as F; The mode then taking out all lattice point number value takes out, and lays in output valve P as first; Wherein i=1,2,3 ... I; News is all WiFi signal intensity level y that a certain moment mobile phone is collected_ij;

Step 112: based on right 111, obtains n first deposit output valve and is designated as the first global variable [P₁,P₂...P_n], corresponding frequency is designated as the second global variable [F₁,F₂...F_n];

Step 113: based on step 111 and step 112, mobile phone loads cartographic information; Owing to map existing aisle, assume that object moves only along road, and there is the setting of the upper limit (position of next second is two, current location of distance grid lattice point at most) and network database in translational speed, the frequency distribution the vector then signal being currently received being sampled is weighted with the second global variable that (in my experiment, the frequency distribution of current demand signal is 0.5, five bars weights of global variable two everybody 0.1)(F_nFor frequency distribution vector) draw final frequency distribution F, it is the second deposit output valve to the grid lattice values that frequency in final frequency distribution F is maximum, in addition update the second global variable does from the frequency distribution that current time is farthest, if the second deposit output valve meets: in the first global variable within the scope of adjacent two grid lattice values of all first deposit output valves, then output the second deposit output valve, and by the second deposit output valve the first global variable done and update; Otherwise export the n-th point in the first global variable;

Step 114: lay in output valve according to second and obtain indoor navigation route

Four, the debugging stage:

1, the WiFi signal in place collected by the mobile phone utilizing different model.

2, different extraction signal number k is changed, and ��^t, ��^t, reach the highest precision.

The invention is not limited in aforesaid detailed description of the invention. The present invention expands to any new feature disclosed in this manual or any new combination, and the step of the arbitrary new method disclosed or process or any new combination.

Claims (8)

1. the localization method based on Bayes posterior probability, it is characterised in that including:

Step 1: in noiseless environment, it is some grid lattice points by field division, and be each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database about; Noiseless environment is collected the WiFi signal intensity level y of each grid lattice point by different mobile phone models_ij; The WiFi signal intensity level y that described mobile phone model, grid lattice point number value are corresponding with mobile phone_ijCorresponding;

Step 2: set up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; By changing ��^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^z; Wherein i is a certain mobile phone correspondence grid lattice point number value, and j is wireless router number value on a certain mobile phone correspondence grid lattice point, y_ijFor the WIFI signal that mobile phone receives from jth wireless router at i-th place; d_ijFor the i-th wireless router spacing to jth wireless router, for given value; ��_ijAt i-th grid lattice point, the measurement error of jth wireless router signal is obeyed N (�� for mobile phone_ij,��_ij) distribution, wherein ��_ijThe average of jth WIFI signal error, �� is collected for i-th grid lattice point_ijThe variance yields of jth WIFI signal error is collected for i-th grid lattice point;

Step 3: at positioning stage, collects m wireless router WIFI signal by mobile phone, sets wherein k WIFI signal noiseless, there is noisy signal in all the other m-k signals; Wherein,

Different �� is obtained for the training stage^zWith ��^z, from m ap, to take out k signal at random, substitute into Bayes posterior probability P (news | i), draw the maximized lattice point number value of posterior probability, the frequency of all lattice points numbering is designated as F; The mode of all lattice points numbering takes out, and is designated as P as deposit output one; Then mobile phone loads cartographic information, obtains indoor navigation route after all lattice point number value are processed.

2. a kind of WIFI localization method based on Bayes posterior probability according to claim 1, it is characterised in that set up model y for mobile phone model t in described step 2_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; Specifically include:

Step 21: set up model y for mobile phone signal i_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, the signal data then collected according to different mobile phones sets up maximum likelihood function L: wherein Wherein t is different mobile phone models, and T is mobile phone quantitative value; N_tFor different mobile phone number of samples;

Step 22: with gradient descent method, it is thus achieved that ��^t, ��^t, ��_ij, ��_ij; Wherein ��^tFor the �� value of certain model mobile phone, ��^t�� value is received for certain signal;For receiving all wireless router signal values at a certain signal mobile phone at i point, for given value; I is total some bit number value, for given value; J is the sum of wireless router, for given value.

3. based on a kind of WIFI localization method based on Bayes posterior probability described in claim 1 or 2, it is characterised in that by changing �� in described step 2^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^zDetailed process is: build ��^zFor interval [��₀,��_q] in uniform point, then ��^z=��₀+z*(��_q-��₀)/p; Wherein ��₀=min{ ��^t: t=1,2...T}, ��_q=max{ ��^t: t=1,2...T}, ��^zThere is p+1, [��₀,��_q] for comprising described ��^tInterval; T=1,2...T; In like manner produceZ is for uniformly putting number value, and p is for uniformly putting number.

4. a kind of WIFI localization method based on Bayes posterior probability according to claim 1, it is characterised in that described step 3 specifically includes:

Step 111: based on posterior probability P described in claim 3 (news | i), obtainIndividual posterior probability P (news | i) the grid lattice point number value that maximum is corresponding, the frequency of all lattice point number value is designated as F; The mode then taking out all lattice point number value takes out, and lays in output valve P as first; Wherein i=1,2,3...I; News is all WiFi signal intensity level y that a certain moment mobile phone is collected_ij;

Step 112: based on right 111, obtains n first deposit output valve and is designated as the first global variable [P₁,P₂...P_n], corresponding frequency is designated as the second global variable [F₁,F₂...F_n];

Step 113: based on step 111 and step 112, mobile phone loads cartographic information; Owing to there is aisle on map, it is assumed that object moves only along road, and there is the setting of the upper limit and network database in translational speed, then the frequency distribution vector signal being currently received being sampled is weighted with the second global variableDraw final frequency distribution F, it is the second deposit output valve to the grid lattice values that frequency in final frequency distribution F is maximum, in addition update the second global variable does from the frequency distribution that current time is farthest, if the second deposit output valve meets: in the first global variable within the scope of adjacent two grid lattice values of all first deposit output valves, then output the second deposit output valve, and by the second deposit output valve the first global variable done and update; Otherwise export the n-th point in the first global variable; Described X represents position two, current location of the distance grid lattice point at most of next second; F_nFor frequency distribution vector;

Step 114: lay in output valve according to second and obtain indoor navigation route.

5. the positioner based on Bayes posterior probability, it is characterised in that including:

Mesh data library module, for being some grid lattice points by field division in noiseless environment, and be each grid lattice point numbering, obtain grid lattice point number value i, by grid lattice point number value and the grid lattice point number value of walking position can do a network database about; Noiseless environment is collected the WiFi signal intensity level y of each grid lattice point by different mobile phone models_ij; The WiFi signal intensity level y that described mobile phone model, grid lattice point number value are corresponding with mobile phone_ijCorresponding;

Estimated value computing module, for setting up model y for mobile phone model t_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; By changing ��^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^z; Wherein i is a certain mobile phone correspondence grid lattice point number value, and j is wireless router number value on a certain mobile phone correspondence grid lattice point, y_ijFor the WIFI signal that mobile phone receives from jth wireless router at i-th place; d_ijFor the i-th wireless router spacing to jth wireless router, for given value; ��_ijAt i-th grid lattice point, the measurement error of jth wireless router signal is obeyed N (�� for mobile phone_ij,��_ij) distribution, wherein ��_ijThe average of jth WIFI signal error, �� is collected for i-th grid lattice point_ijThe variance yields of jth WIFI signal error is collected for i-th grid lattice point;

Bayes posterior probability locating module, at positioning stage, collecting m wireless router WIFI signal by mobile phone, set wherein k WIFI signal noiseless, there is noisy signal in all the other m-k signals; Wherein,

Different �� is obtained for the training stage^zWith ��^z, from m ap, to take out k signal at random, substitute into Bayes posterior probability P (news | i), draw the maximized lattice point number value of posterior probability, the frequency of all lattice points numbering is designated as F; The mode of all lattice points numbering takes out, and is designated as P as deposit output one; Then mobile phone loads cartographic information, obtains indoor navigation route after all lattice point number value are processed.

6. a kind of positioner based on Bayes posterior probability according to claim 5, it is characterised in that set up model y for mobile phone model t in described estimated value computing module_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, then according to maximum likelihood function calculate obtain ��^t, ��^t, ��_ij, ��_ij; Specifically include:

Step 21: set up model y for mobile phone signal i_ij=��^t+��^tlog(d_ij)+��_ij, obtain ��_ij����^tAnd ��^tRelation, the signal data then collected according to different mobile phones sets up maximum likelihood function L: wherein Wherein t is different mobile phone models, and T is mobile phone quantitative value; N_tFor different mobile phone number of samples;

Step 22: with gradient descent method, it is thus achieved that ��^t, ��^t, ��_ij, ��_ij; Wherein ��^tFor the �� value of certain model mobile phone, ��^t�� value is received for certain signal;For receiving all wireless router signal values at a certain signal mobile phone at i point, for given value; I is total some bit number value, for given value; J is the sum of wireless router, for given value.

7. a kind of positioner based on Bayes posterior probability according to claim 5, it is characterised in that by changing �� in described estimated value computing module^t, ��^tObtain the parameter alpha that different mobile phone is corresponding^zAnd ��^zDetailed process is: build ��^zFor interval [��₀,��_q] in uniform point, then ��^z=��₀+z*(��_q-��₀)/p; Wherein ��₀=min{ ��^t: t=1,2...T}, ��_q=max{ ��^t: t=1,2...T}, ��^zThere is p+1, [��₀,��_q] for comprising described ��^tInterval; T=1,2...T; In like manner produceZ is for uniformly putting number value, and p is for uniformly putting number.

8. a kind of positioner based on Bayes posterior probability according to claim 5, it is characterised in that described Bayes posterior probability locating module work process specifically includes:

Step 111: based on posterior probability P described in claim 3 (news | i), obtainIndividual posterior probability P (news | i) the grid lattice point number value that maximum is corresponding, the frequency of all lattice point number value is designated as F; The mode then taking out all lattice point number value takes out, and lays in output valve P as first; Wherein i=1,2,3...I; News is all WiFi signal intensity level y that a certain moment mobile phone is collected_ij;

Step 112: based on right 111, obtains n first deposit output valve and is designated as the first global variable [P₁,P₂...P_n], corresponding frequency is designated as the second global variable [F₁,F₂...F_n];

Step 113: based on step 111 and step 112, mobile phone loads cartographic information; Owing to there is aisle on map, it is assumed that object moves only along road, and there is the setting of the upper limit and network database in translational speed, then the frequency distribution vector signal being currently received being sampled is weighted with the second global variableDraw final frequency distribution F, it is the second deposit output valve to the grid lattice values that frequency in final frequency distribution F is maximum, in addition update the second global variable does from the frequency distribution that current time is farthest, if the second deposit output valve meets: in the first global variable within the scope of adjacent two grid lattice values of all first deposit output valves, then output the second deposit output valve, and by the second deposit output valve the first global variable done and update; Otherwise export the n-th point in the first global variable; Described X represents position two, current location of the distance grid lattice point at most of next second; F_nFor frequency distribution vector;

Step 114: lay in output valve according to second and obtain indoor navigation route.