CN105792351A - Wireless fingerprint matching method based on unequal length sequence similarity - Google Patents

Abstract

The invention discloses a wireless fingerprint matching method based on unequal length sequence similarities. The method is based on a plurality of deployed Wi-Fi hotspots, a Wi-Fi detection module for detecting Wi-Fi signals in real time and a database for storing Wi-Fi fingerprint information. The Wi-Fi detection module is arranged on the mobile terminal. The method comprises two stages of determining candidate sequences and calculating unequal length sequence similarities. The stage of determining the candidate sequences is used for obtaining a set of candidate sequences related to a to-be-identified sequence. The set of candidate sequences is taken as a sequence set for calculating similarities and determining a position label of a to-be-detected point in the next step. The stage of calculating the unequal length sequence similarities is used for searching the sequence most approximate to the to-be-detected sequence from the candidate sequences. The result is used for searching position information from the background database. According to the wireless fingerprint matching method based on unequal length sequence similarities provided by the invention, on the premise of low cost deployment, the positioning demands suitable for the positioning precision of a parking lot are realized; and the method has relatively high practical value in application of positioning a vehicle in a large underground parking lot.

Description

Wireless fingerprint matching process based on Length discrepancy sequence similarity

[technical field]

The present invention relates in indoor positioning technologies wireless fingerprint coupling technical field, particularly to one based on The wireless fingerprint matching process of long sequence similarity.

[background technology]

Along with the development of mobile interchange technology, location-based service and location-aware computing etc. increasingly come into one's own.In outdoor Under environment, Satellite Navigation Technique is widely used to the life of people.And the location-based service demand under indoor environment is continuously increased, Such as special population monitoring, large stadium management, disaster relief, mining management, mobile marketing etc..Under these application scenarios, Satellite-signal is limited, and location technology faces many difficult problems.The most conventional indoor positioning technologies relates to three classes: one is GNSS skill Art, such as pseudo satellite technology etc.；Two is wireless location technology, as wireless communication signals, wireless radiofrequency label, ultrasound wave, light follow the tracks of, Wireless senser location etc.；Three is other technologies, such as computer vision, dead reckoning etc..Wherein, fingerprint based on matching idea Matching technique has the features such as simple, the low cost of deployment, is suitable for the location requirement under above-mentioned scene.

Wireless fingerprint coupling refers to be implied with the wireless signal of scene location information by real-time reception, with fingerprint database In information record mate, thus to measure point position judge.Under indoor environment, particularly building Or in hypogee, radio wave through propagation loss, reflect, reflect, diffraction and multipath transmisstion, touch barrier every time Will be absorbed by some energy.Finally measure point at certain and form specific radio signal characteristics (such as signal number, by force Degree, phase place etc.).If this feature and position coordinates are associated, then can characterize tested point position by this signal characteristic Put.Therefore, wireless fingerprint coupling is generally divided into two processes: one is zone of dispersion signal characteristic collection in scene, and signal is special Levy and associate with positional information；Two is tested point signature analysis, by signal characteristic calculating location information.

In wireless fingerprint position fixing process, being identified radio signal characteristics and comparing is committed step.Indoor environment Under, the change of the aspects such as the change of interior architecture general layout, facility and personnel move, radio signal source all can cause wireless signal special Levy and change, and then cause position error.The most conventional location evaluating standard includes positioning precision, complexity, autgmentability, is good for The indexs such as strong property.Consider in the common scene of indoor positioning, as visible in megastore, underground parking position the visual field of target Property, positioning precision reaches tens meters can meet requirement, and therefore algorithm vigorousness is more paid close attention in this type of location, i.e. occurs certain when signal Impact on Position location accuracy during change.

Wi-Fi Hotspot is the wireless base station device of a kind of low cost, is widely applied at present.By referring to from Wi-Fi The composition of stricture of vagina feature is set out, it is believed that wireless fingerprint coupling is the Similarity measures problem of Length discrepancy mutability sequence in fact, Jin Erti Go out a kind of wireless fingerprint matching process based on Length discrepancy sequence similarity.

[summary of the invention]

It is an object of the invention to overcome above-mentioned the deficiencies in the prior art, it is provided that a kind of based on Length discrepancy sequence similarity Wireless fingerprint matching process, its aim to solve the problem that prior art during wireless location, owing to Wi-Fi sequence is affected by environment relatively Greatly, and individual nodes less stable, sequence recognition and the bigger technical problem of Similarity measures difficulty are caused.

For achieving the above object, the present invention proposes a kind of wireless fingerprint match party based on Length discrepancy sequence similarity Method, based on some the Wi-Fi Hotspots disposed, for the Wi-Fi detection module of detection Wi-Fi signal in real time, described Wi-Fi Detection module is arranged on mobile terminal, for storing the data base of Wi-Fi finger print information, including determine candidate sequence and Long sequence similarity calculates two stages, specifically comprises the following steps that

Step one, determine candidate sequence:

1. the signal in sequence to be measured is sorted by intensity, select N number of signal of wherein maximum intensity: due to Wi-Fi base Standing self stability and the change of propagation ducts state, the signal intensity in sequence to be measured, quantity is likely to change, mobile After the Wi-Fi detection module of terminal receives Wi-Fi signal, first it is ranked up by intensity, and takes out N number of letter of maximum intensity Number, obtain signal sequence AP to be identified₁,AP₂,...,AP_N, it is stored in the fingerprint base of mobile terminal；

The most respectively with this N number of signal above-mentioned for index, from fingerprint base, screening comprises signal AP_iFingerprint sequence, signal Index is carried out in advance under off-line state, so can quickly obtain the candidate sequence comprising these signals.Owing to Wi-Fi has office The signal that portion's property, i.e. fingerprint sequence are comprised is generally from neighbouring focus, and therefore index length can arrange shorter, it is assumed that rope Drawing a length of k, whole Index process is equivalent to once map, and its time loss may be regarded as constant；

Step 2, Length discrepancy sequence similarity calculate:

By abstract for the Wi-Fi fingerprint Similarity measures computational problem of equal value for Length discrepancy mutability sequence.If s₁=(r₁₁, r₁₂,...,r_1m), s₂=(r₂₁,r₂₂,...,r_2n) representing the intensity of hot spots in two Wi-Fi sequences respectively, m and n is focus Number, then Wi-Fi fingerprint similarity is represented by s₁And s₂Similarity.Owing to m and n is not necessarily the same, therefore use the same generic of sequence Property than and the method for property value mean square deviation Weighted Fusion calculate.

First s is constructed₁And s₂The union of middle focus, and it is rearranged for new sequence p by hotspot name or No. ID, if p Size is t.This sequence is included in s₁And s₂The middle all Wi-Fi Hotspots occurred；

2. will be original at s₁And s₂In intensity of hot spots value be tagged in sequence p, respectively formed two sequences p₁=(r '₁₁, r′₁₂,...,r′_1t), p₂=(r '₂₁,r′₂₂,...,r′_2t).Wherein, the intensity of hot spots not having in former sequence is designated as 0.r′_1iWith r′_2iRepresent No. i-th focus intensity in the two sequences respectively.In this way by Sequence Transformed for Length discrepancy for isometric sequence Row, will calculate s₁And s₂Similarity be converted into calculating p₁And p₂Similarity.

3. the mean square deviation of two sequence intensity of hot spots is calculated, as shown in formula (1):

$\mathrm{\σ}=\sqrt{\frac{1}{t}\underset{i=1}{\overset{t}{\Σ}}{(r_{2i}^{\′}-r_{1i}^{\′})}^2}---\left(1\right)$

Embody, by mean square deviation, the intensity of hot spots difference that former sequence has simultaneously, also can reflect when focus only occurs in one Impact on sequence similarity time in sequence；

Step 3, it is used for looking into from background data base by the result calculated through step 2 Length discrepancy sequence similarity Look for positional information.

As preferably, in described step one, the value of k is 10～20.

As preferably, described step one determines that the detailed process of candidate sequence is as shown in algorithm 1:

Algorithm 1

	Input:Signals (from Sensors), FingerPrintDataBase
			Output:Sequences
1	Vector Sequences=NULL
		2	Vector Signals=GetSignals () //from Sensors
3	Signals=SortByStrength (Signals)
		4	Signals=GetTheFirstNSignals (Signals, N)
5	For Each APi∈Signals Do
		6	Sequences=Index (APi, FingerPrintDataBase)
7	End For

As preferably, the detailed process that described step 2 Length discrepancy sequence similarity calculates is as shown in algorithm 2:

Algorithm 2

	Input:Vector V, Sequences
			Sequence most like with V in Output:Result//Sequences
1	Double Min=0
		2	Vector Result=NULL
3	For Each Si∈Sequences Do
		4	Vector Temp=V-Si//extension V
5	Set Temp.signals=0
		6	V=V ∪ Temp
7	Temp=Si V // extension Si
		8	Set Temp.signals=0
9	Si=Si ∪ Temp
		10	Double Sim=Simlarity (V, Si)
11	If Sim<Min Then
		12	Min=Sim
13	Result=Si
		14	End If
15	End For

Beneficial effects of the present invention: compared with prior art, the one that the present invention provides is based on Length discrepancy sequence similarity Wireless fingerprint matching process, being determined by the candidate sequence stage obtains the one group candidate sequence relevant to sequence to be identified, work Calculate similarity for next step, determine the arrangement set of tested point location tags, by Length discrepancy sequence similarity calculation stages, Achieving and find out sequence immediate with sequence to be checked in candidate sequence, this result will be used for from background data base searching position Confidence ceases.In conjunction with urban parking area problem of management, a kind of wireless fingerprint matching algorithm based on Length discrepancy sequence similarity is proposed, Preferably solve and determine wireless candidate sequence and calculate the problems such as Length discrepancy sequence similarity, in simulation Large Underground parking lot The application scenarios of vehicle location embodies its practical value, it is possible under low cost disposes premise, it is achieved be suitable for parking lot The location requirement of positioning precision.

Inventive feature and advantage will combine accompanying drawing by embodiment and be described in detail.

[accompanying drawing explanation]

Fig. 1 is the signal index schematic diagram of the present invention；

Fig. 2 is the sequence transformation schematic diagram of the present invention；

Fig. 3 is that the fingerprint of the embodiment of the present invention obtains and mates schematic diagram.

[detailed description of the invention]

For making the object, technical solutions and advantages of the present invention of greater clarity, below by accompanying drawing and embodiment, right The present invention is further elaborated.However, it should be understood that specific embodiment described herein is only in order to explain this Bright, it is not limited to the scope of the present invention.Additionally, in the following description, eliminate the description to known features and technology, with Avoid unnecessarily obscuring idea of the invention.

The embodiment of the present invention provides a kind of wireless fingerprint matching process based on Length discrepancy sequence similarity, based on some The Wi-Fi Hotspot disposed, for the Wi-Fi detection module of detection Wi-Fi signal in real time, described Wi-Fi detection module is arranged On mobile terminal, for storing the data base of Wi-Fi finger print information, including determining candidate sequence and Length discrepancy sequence similarity Calculate two stages, specifically comprise the following steps that

Step one, determine candidate sequence:

1. the signal in sequence to be measured is sorted by intensity, select N number of signal of wherein maximum intensity: due to Wi-Fi base Standing self stability and the change of propagation ducts state, the signal intensity in sequence to be measured, quantity is likely to change, mobile After the Wi-Fi detection module of terminal receives Wi-Fi signal, first it is ranked up by intensity, and takes out N number of letter of maximum intensity Number, obtain signal sequence AP to be identified₁,AP₂,...,AP_N, it is stored in the fingerprint base of mobile terminal；

The most respectively with this N number of signal above-mentioned for index, from fingerprint base, screening comprises signal AP_iFingerprint sequence, signal Index is carried out in advance under off-line state, so can quickly obtain the candidate sequence comprising these signals.Owing to Wi-Fi has office The signal that portion's property, i.e. fingerprint sequence are comprised is generally from neighbouring focus, and therefore index length can arrange shorter, it is assumed that rope Drawing a length of k, the value of k is 10～20, and whole Index process is equivalent to once map, and its time loss may be regarded as constant.

Refering to Fig. 1, signal Index process is as shown in Figure 1.

Step one determines that the detailed process of candidate sequence is as shown in algorithm 1:

Algorithm 1

	Input:Signals (from Sensors), FingerPrintDataBase
			Output:Sequences
1	Vector Sequences=NULL
		2	Vector Signals=GetSignals () //from Sensors
3	Signals=SortByStrength (Signals)
		4	Signals=GetTheFirstNSignals (Signals, N)
5	For Each APi∈Signals Do
		6	Sequences=Index (APi, FingerPrintDataBase)
7	End For

The target of algorithm 1 is to obtain the one group candidate sequence relevant to sequence to be identified, as next step calculating similarity, Determine the arrangement set of tested point location tags.

Step 2, Length discrepancy sequence similarity calculate:

By abstract for the Wi-Fi fingerprint Similarity measures computational problem of equal value for Length discrepancy mutability sequence.If s₁=(r₁₁, r₁₂,...,r_1m), s₂=(r₂₁,r₂₂,...,r_2n) representing the intensity of hot spots in two Wi-Fi sequences respectively, m and n is focus Number, then Wi-Fi fingerprint similarity is represented by s₁And s₂Similarity.Owing to m and n is not necessarily the same, therefore use the same generic of sequence Property than and the method for property value mean square deviation Weighted Fusion calculate.

First s is constructed₁And s₂The union of middle focus, and it is rearranged for new sequence p by hotspot name or No. ID, if p Size is t.This sequence is included in s₁And s₂The middle all Wi-Fi Hotspots occurred；

2. will be original at s₁And s₂In intensity of hot spots value be tagged in sequence p, respectively formed two sequences p₁=(r '₁₁, r′₁₂,...,r′_1t), p₂=(r '₂₁,r′₂₂,...,r′_2t).Wherein, the intensity of hot spots not having in former sequence is designated as 0.r′_1iWith r′_2iRepresent No. i-th focus intensity in the two sequences respectively.In this way by Sequence Transformed for Length discrepancy for isometric sequence Row, will calculate s₁And s₂Similarity be converted into calculating p₁And p₂Similarity.

Refering to Fig. 2, sequence transformation schematic diagram is as shown in Figure 2.

3. the mean square deviation of two sequence intensity of hot spots is calculated, as shown in formula (1):

$\mathrm{\σ}=\sqrt{\frac{1}{t}\underset{i=1}{\overset{t}{\Σ}}{(r_{2i}^{\′}-r_{1i}^{\′})}^2}---\left(1\right)$

Embody, by mean square deviation, the intensity of hot spots difference that former sequence has simultaneously, also can reflect when focus only occurs in one Impact on sequence similarity time in sequence.

The detailed process that step 2 Length discrepancy sequence similarity calculates is as shown in algorithm 2:

Algorithm 2

	Input:Vector V, Sequences
			Sequence most like with V in Output:Result//Sequences
1	Double Min=0
		2	Vector Result=NULL
3	For Each Si∈Sequences Do
		4	Vector Temp=V-Si//extension V
5	Set Temp.signals=0
		6	V=V ∪ Temp
7	Temp=Si V // extension Si
		8	Set Temp.signals=0
9	Si=Si ∪ Temp
		10	Double Sim=Simlarity (V, Si)
11	If Sim<Min Then
		12	Min=Sim
13	Result=Si
		14	End If
15	End For

Algorithm 2 achieves finds out sequence immediate with sequence to be checked in candidate sequence, and this result will be used for from backstage Data base searches positional information.

Step 3, it is used for looking into from background data base by the result calculated through step 2 Length discrepancy sequence similarity Look for positional information.

Embodiment one,

Refering to Fig. 3, as a example by the bus location in intelligent parking lot, in figure, fingerprint unit records setting base ID, physical location mark, Wi-Fi sequence strength information；Similarity calculation unit realizes Length discrepancy sequence merger and similarity meter Calculate.Automobile obtains neighbouring Wi-Fi Hotspot information in real time, is packaged into s₁Mail to similarity calculation unit；Similarity calculation unit from Known fingerprint data base obtains Wi-Fi sequence s of reference point successively₂, it is then converted into isometric sequence p₁And p₂And calculate similar Property.Finally will there is sequence s of minimum similarity₂Corresponding positional information returns automobile.This process achieves from Wi-Fi Hotspot Sequence is to the mapping calculation of automobile position.Position during finding parking stall, reverse car seeking, indoor navigation determines and this figure class Seemingly.

Test:

Having surveyed and drawn certain large supermarket's underground parking layout on the spot, according to actual geographic design data, Wi-Fi disposes flat Face figure, and non-homogeneous in 120 parking stalls of simulation, in the range of about 2500 square metres deploy 12 Wi-Fi Hotspots, every spacing About 10～20 meters.It is simultaneously based on android system and develops Wi-Fi detection module, for detection Wi-Fi signal in real time.At this On the basis of, have detected the Wi-Fi signal spectrogram in the range of deployment herein.Consider during finding parking stall and reverse car seeking, The feature that people are low to positioning accuracy request, it is determined that detect in units of 2 meters, 620 Wi-Fi of mapping position base altogether Point.Wherein part setting base sample data is as shown in table 1.

Table 1 part setting base sample data table (dBm)

Position ID	AP001	AP002	AP003	AP004	AP005
						1	-78	-69	-36	-90	-93
2	-75	-81	-36	-81	-87
						3	-63	-72	-66	-81	-84
4	-48	-54	-81	-81	-93
						5	-51	-45	-90	-87	-85
6	-48	-51	-90	-87	-85
						7	-48	-54	-81	-87	-84
8	-75	-87	-39	-93	-87
						9	-72	-87	-69	-84	-90
10	-66	-78	-57	-75	-90

Signal strength expression is negative.620 basic points be have collected 5～12 dimension intensity datas by actual test respectively.Table 1 Only listing the data instance of wherein 10 setting bases, each setting base signal vector have chosen 5 Wi-Fi Hotspot letters Number.

First test intensity region indexing situation.Test respectively away from focus 5 meters, 10 meters, 15 meters, 20 meters etc. several in the case of Signal intensity.Wherein the intensity data of 3 focuses is as shown in table 2.

Table 2 intensity and distance relation signal (dBm)

Focus ID	5m	10m	15m	20m
					AP001	-57	-63	-69	-81
AP002	-45	-51	-66	-75
					AP003	-42	-57	-77	-90

What table 2 reflected signal intensity and distance generally is monotonically correlated relation.But, due to focus stability factor Impact, even if the data recorded in same point also can change, as AP001, records 10 meters of distances the most respectively Two stationary values :-63 and-60, and extremum may reach-75 and-54.This species diversity is the most unfavorable to calculating.In test Focus all cut off network transmission, only power offer wireless signal with regulated power supply, reduce by information transmission, power jitter as far as possible The change in signal strength caused etc. factor.

On this basis, in the case of testing firm power and random disturbance power two kinds respectively, keep Wi-Fi Hotspot complete Open, and open at random 10 focuses, 8 focuses situation to simulate the situation of Length discrepancy sequence.Precision is the assignment test result of 2 meters As shown in table 3.

Table 3 test result

Show from (1st) of table 3 group result, under fixed transmission power premise, test zone interior-heat count 10 with Good result was embodied time upper.(2nd) group result shows, by the random feelings reducing transmitting power analog change in signal strength Condition, accuracy is decreased obviously.While it is true, but under the application background of parking position location, it is therefore an objective to make car owner in the ken In the range of it can be found that parking stall or vehicle, and owing to car owner has the visual range of at least tens meters, and remote-control key effectively seeks car Signal also can cover this scope, and therefore positioning precision be enough to solve problem within the scope of 10～20 meters.It practice, stopping Under the application scenarios such as parking lot, megastore, exhibition, it is all acceptable that positioning precision reaches ten meters of scopes.Positioning accurate herein Degree is entirely capable of meeting the location requirement of intelligent parking lot.

Realize positioning function by disposing Wi-Fi Hotspot, greatly reduce positioning system hardware expense, it is achieved that low cost Dispose.Higher correct localization it is obtained in that in the case of power invariability.Although efficiency paying no attention in the case of power variable Think, but during Practical Project, can be by increasing redundant nodes, guaranteeing that the forms such as signal stabilization ensure stablizing of signal covering Property.

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention Any amendment, equivalent or the improvement etc. made within god and principle, should be included within the scope of the present invention.

Claims (4)

1. a wireless fingerprint matching process based on Length discrepancy sequence similarity, based on some the Wi-Fi Hotspots disposed, uses In the Wi-Fi detection module of detection Wi-Fi signal in real time, described Wi-Fi detection module is arranged on mobile terminal, is used for depositing The data base of storage Wi-Fi finger print information, it is characterised in that: include determining that candidate sequence and Length discrepancy sequence similarity calculate two In the stage, specifically comprise the following steps that

Step one, determine candidate sequence:

1. the signal in sequence to be measured is sorted by intensity, select N number of signal of wherein maximum intensity: due to Wi-Fi base station certainly Body stability and propagation ducts state change, and the signal intensity in sequence to be measured, quantity are likely to change, mobile terminal Wi-Fi detection module receive Wi-Fi signal after, be first ranked up by intensity, and take out N number of signal of maximum intensity, To signal sequence AP to be identified₁,AP₂,...,AP_N, it is stored in the fingerprint base of mobile terminal；

The most respectively with this N number of signal above-mentioned for index, from fingerprint base, screening comprises signal AP_iFingerprint sequence, signal index thing First carry out under off-line state, so can quickly obtain the candidate sequence comprising these signals.Owing to Wi-Fi has locality, The signal that i.e. fingerprint sequence is comprised is generally from neighbouring focus, and therefore index length can arrange shorter, it is assumed that index is long Degree is k, and whole Index process is equivalent to once map, and its time loss may be regarded as constant；

Step 2, Length discrepancy sequence similarity calculate:

By abstract for the Wi-Fi fingerprint Similarity measures computational problem of equal value for Length discrepancy mutability sequence.If s₁=(r₁₁,r₁₂,..., r_1m), s₂=(r₂₁,r₂₂,...,r_2n) representing the intensity of hot spots in two Wi-Fi sequences respectively, m and n is focus number, then Wi- Fi fingerprint similarity is represented by s₁And s₂Similarity.Owing to m and n is not necessarily the same, therefore use sequence kind attributes ratio and genus The method of property value mean square deviation Weighted Fusion calculates.

First s is constructed₁And s₂The union of middle focus, and it is rearranged for new sequence p by hotspot name or No. ID, if the size of p For t.This sequence is included in s₁And s₂The middle all Wi-Fi Hotspots occurred；

2. will be original at s₁And s₂In intensity of hot spots value be tagged in sequence p, respectively formed two sequences p₁=(r '₁₁,r ′₁₂,...,r′_1t), p₂=(r '₂₁,r′₂₂,...,r′_2t).Wherein, the intensity of hot spots not having in former sequence is designated as 0.r′_1iWith r '_2i Represent No. i-th focus intensity in the two sequences respectively.In this way by Sequence Transformed for Length discrepancy for isometric sequence, general Calculate s₁And s₂Similarity be converted into calculating p₁And p₂Similarity.

3. the mean square deviation of two sequence intensity of hot spots is calculated, as shown in formula (1):

$\mathrm{\σ}=\sqrt{\frac{1}{t}\underset{i=1}{\overset{t}{\Σ}}{(r_{2i}^{\′}-r_{1i}^{\′})}^2}---\left(1\right)$

Embody, by mean square deviation, the intensity of hot spots difference that former sequence has simultaneously, also can reflect when focus only occurs in a sequence Impact on sequence similarity time middle；

Step 3, the result calculated through step 2 Length discrepancy sequence similarity is used for from background data base search position Confidence ceases.

A kind of wireless fingerprint matching process based on Length discrepancy sequence similarity, it is characterised in that: In described step one, the value of k is 10～20.

A kind of wireless fingerprint matching process based on Length discrepancy sequence similarity, it is characterised in that: Described step one determines that the detailed process of candidate sequence is as shown in algorithm 1:

Algorithm 1

A kind of wireless fingerprint matching process based on Length discrepancy sequence similarity, it is characterised in that: The detailed process that described step 2 Length discrepancy sequence similarity calculates is as shown in algorithm 2:

Algorithm 2