CN104267439A - Unsupervised human detecting and positioning method - Google Patents

Abstract

The invention relates to an unsupervised human detecting and positioning method. The method includes the steps that a transmitter transmits signals, and the signals pass through a monitoring area and are received by a receiver; channel state information data are collected in real time, subcarrier amplitudes of all data in a sliding time window are averaged to obtain a vector containing the average of all the subcarrier amplitudes, and the vector serves as the signal characteristic; the signal characteristic of the detection signals is extracted from the sliding window, the correlation between the detection signal characteristic and the static standard signal characteristic is calculated, if the correlation received by all receiver antennas is larger than a set threshold value, it shows that the human body appears in the monitoring area, the position of the human body is positioned, or else signal detection is continued on the monitoring area; parameter information in a positioning model parameter file is read, the probabilities, at all positions capable of being positioned, of the detection signal characteristic are calculated, and the position with the largest probability correlation is the position where the human body is located most likely.

Description

A kind of method without supervising human detection and location

Technical field

The present invention relates to wireless location technology, particularly relating to a kind of method without supervising human detection and location.

Background technology

Along with developing rapidly of WLAN technology, wireless WLAN device is deployed to the place of people's frequent activity in a large number.Positioning system based on WLAN is taken advantage of a situation and is developed rapidly, in a large number reliably, stably indoor wireless locating system flood the market.Wireless WLAN positioning system comprises an important subsystem: wireless passive human body detection & localization.Wireless passive human body detection & localization refers to and utilizes wireless technology to detect, locates, follows the tracks of the activity of non-helpmate in guarded region.Non-helpmate is defined as the people not carrying Positioning Electronic Devices, not active participate position data collecting activity.Equally, non-helpmate is the existence of not knowing passive detection and positioning system under normal circumstances.Contrast active location system, passive detection is same with location has universality, has huge application potential in safety guard field, as intrusion detection, safeguarding of assets, intelligence nurse etc.

Early stage at wireless passive human body detection & localization System Development, most of system is all realized human detection and location by analyte signal intensity information (RSSI) change.Mostly early stage wireless passive human body detection & localization is to utilize wireless sensor devices to realize, and by disposing sensor node in large quantities, improving the wireless signal density of guarded region, reaching the detection and positioning occurred human body.Along with the appearance of WiFi equipment, at a distance, on a large scale radio communication become possibility, wireless passive human body detection & localization system then utilizes received signals fingerprint technology to realize human detection and location on a large scale.This type systematic comprises two parts: under off-line state, and when in training acquisition monitoring region, diverse location place human body exists, received signals fingerprint information is stored in fingerprint database; Under presence, once human body occurs, then the method for Land use models coupling or machine learning estimates position of human body to detection signal and fingerprint signal similarity.But under indoor environment, complicated signal multipath fading and time domain dynamic change problem cause that RSSI value is changeable not to be surveyed, even if the fluctuation of its value also can reach 5dB under static environment.The raising that instability limit passive human body detection & localization system accuracy of RSSI.

After ofdm system (OFDM) occurs, multiple carrier channel status information causes numerous focus of attention.Channel condition information (CSI) is a kind of fine-grained radio signal characteristics, and it contains amplitude and the phase information of signal on different frequency, can describe indoor multipath feature to a certain extent.At present under 802.11a/g/n standard, channel condition information can obtain from OFDM receiver.As a kind of signal characteristic being better than RSSI, the passive human body detection & localization based on CSI has attracted a large amount of scholar to study, and achieves some achievements.Similar to based on RSSI passive human body detection & localization system, the passive human body detection & localization system based on CSI developed at present or dependence are disposed a large amount of receiver and transmitter or need fingerprint on site collecting work in large quantities.The deployment of large number quipments limits systematic difference scope to a certain extent, such as usually only comprises a small amount of in domestic environment, even single communication link.Equally, fingerprint on site collecting work adds system undoubtedly and disposes expense in large quantities.

Summary of the invention

The object of the invention is to provide a kind of method without supervising human detection and location, reduces the deployment facility of human detection and location, effectively strengthens the application of human detection and location.

Realize the technical scheme of the object of the invention:

A kind of without supervision human detection and localization method, transmitter sends signal, and signal, through monitored area, is received by receiver, it is characterized in that:

Step 1: Real-time Collection channel condition information data, averages to the sub carrier amplitude of data all in sliding time window, obtains the vector that comprises all sub carrier amplitude averages, it can be used as signal characteristic;

Step 2: the signal characteristic extracting detection signal from moving window, calculate the correlativity of detection signal feature and static criteria signal characteristic, if the correlativity on every root receiver antenna is greater than the threshold value of setting, indicate that human body appears at monitored area, then enter step 3, position of human body is positioned, otherwise continues to carry out input to monitored area;

Step 3: read the parameter information in location model Parameter File, then calculating detection signal feature can the probability at position location place at each, and the position that maximum probability is correlated with is the place that human body is most possibly positioned at.

In step 2, said static criteria signal characteristic is that nobody exists the signal characteristic in situation in guarded region, and static criteria signal characteristic is obtained by systematic learning; For static criteria signal characteristic, the coefficient of autocorrelation of the signal characteristic in all packets in sliding time window is all less than setting threshold value.

In step 2, detection signal feature H and static criteria signal characteristic H _norcorrelativity, computing formula is as follows:

$C=\mathrm{corr}(H,H_{\mathrm{nor}})=\frac{\mathrm{\Σ}(H-\stackrel{\‾}{H_{\mathrm{nor}}})(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}{\sqrt{\mathrm{\Σ}{(H-\stackrel{\‾}{H})}^2\mathrm{\Σ}{(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}^2}}$

In step 3, parameter information in location model Parameter File is average μ and the standard deviation sigma of the Euclidean distance of detection signal and static criteria signal characteristic on each position, every root receiver antenna, and on receiver antenna, the Euclidean distance value of detection signal feature and static criteria signal characteristic meets Gaussian distribution.

In step 3, calculating detection signal feature can the probability at position location place at each, and its computing formula is as follows:

$P\left(H\right|L)=\underset{k=1}{\overset{n}{\mathrm{\Π}}}\underset{-\∞}{\overset{+\∞}{\∫}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_k}\exp (-\frac{{(x-{\mathrm{\μ}}_k)}^2}{2{\mathrm{\σ}}_k^2})\mathrm{dx}$

Wherein n is receiver antenna number.

In step 3, location model Parameter File generates by the following method,

First, utilize and comprise human body can off-line data under the situation of position location, by each position feature data into clusters, said position feature data refer to the Euclidean distance of detection signal feature and static criteria signal characteristic on all receiver antennas, and its dimension size is identical with receiver antenna number;

Then, each orientable position of position principle of distinction identification is utilized;

Finally, the average of each position feature data in each dimension and standard deviation is calculated, stored in file after it being associated with position.

Position near transmitter on the position of receiver, los path on the position of los path, los path, los path medium position can be comprised in position location.

Position principle of distinction is, when human body is positioned near los path, the signal characteristic of every root antenna and the Euclidean distance of static criteria signal characteristic less, and the signal characteristic between different antennae differs less with the Euclidean distance of static criteria signal characteristic; When human body is positioned at close transmitter one end on los path, the signal characteristic of every root antenna and the Euclidean distance of static criteria signal characteristic are comparatively large, and the signal characteristic between different antennae differs less with the Euclidean distance of static criteria signal characteristic; When human body is positioned at close receiver one end on los path, the signal characteristic of certain root antenna and the Euclidean distance of static criteria signal characteristic are comparatively large, and the signal characteristic between different antennae differs larger with the Euclidean distance of static criteria signal characteristic; When in the middle part of human body is positioned at los path, signal characteristic on each receiver antenna is little near transmitter of human body compared with the Euclidean distance of standard static signal characteristic, and the otherness between these Euclidean distances compares little near receiver of human body.

Beneficial effect of the present invention is:

The present invention proposes a kind of method without supervising human detection and location, utilize the signal characteristic of unsupervised learning technology identification human body when diverse location according to signal characteristic otherness between receiver antenna, recycle maximum Bayesian probability and estimate to locate position of human body.Unsupervised learning technology helps human detection and location to greatly reduce deployment expense, and human detection and orientation range expand by this invention simultaneously, can successfully locate position of human body on obstructed path.

The present invention only relies on one group of receiver and transmitter, can adapt to the human detection under different size space and location requirement.Meanwhile, the present invention also can expand to many group receivers and transmitter to adapt to human detection under larger space and location, has extendability flexibly.

Accompanying drawing explanation

Fig. 1 is that the present invention is without the method flow diagram of supervising human detection and location;

Fig. 2 is that the present invention can the schematic diagram of position location;

Fig. 3 is the schematic diagram of specific implementation human detection of the present invention and location.

Embodiment

As shown in Figure 1, the present invention comprises the following steps without supervision human detection and localization method:

In a step 101, Real-time Collection channel condition information data, the signal characteristic abstraction based on sliding time window mechanism is set up for instantaneous channel conditions inter-area traffic interarea, the sub carrier amplitude of data all in sliding time window is averaged, obtain the vector that comprises all sub carrier amplitude averages, it can be used as signal characteristic.

In wireless transmission process, radio communication can be simply modeled as:

y(t)＝h(t)·x(t)+z(t)

Here t is the time, and y is Received signal strength, and x sends signal, and h is the corresponding or channel condition information of channel, and it is a plural number under normal circumstances, representation signal amplitude and phase information, and z is Gaussian white noise.In traditional indoor environment, a signal transmission can be propagated by multipath, and causes different spread lengths, path loss, different time delays, amplitude decay and phase offset.And multi-path environment can pass through linearly filtrator h (τ) characterization, i.e. channel impulse response (Channel Impulse Response, CIR):

$h\left(\mathrm{\τ}\right)=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\left|a_i\right|\exp (-j{\mathrm{\θ}}_i)\mathrm{\δ}(\mathrm{\τ}-{\mathrm{\τ}}_i)$

Wherein, a _i, θ _iand τ _irepresent the amplitude of i-th multipath, phase place and time delay respectively, i=1,2 ..., N, N represent multipath number.

In frequency field, ofdm system provides channel frequency response (Channel Frequency Response, CFR) in OFDM subcarrier granularity:

H＝{H(1),H(2),...,H(N)}

Wherein N is subcarrier number.The CFR information of each subcarrier is a complex values, and real part is amplitude-frequency response and complex number part is phase response, and each subcarrier is defined as again:

H(f)＝|H(f)|exp(jsin(∠H(f)))

Wherein H (f) represents the amplitude-frequency response of subcarrier, and ∠ H (f) represents the phase response of subcarrier.

For given bandwidth, CIR can convert CFR to by Fast Fourier Transform (FFT):

H＝FFT(h(τ))

Although CIR and CFR is of equal value in channel corresponding model.But being more partial in human detection and positioning field uses CFR as signal characteristic.At present the firmware after upgrading can be utilized to extract the CFR with 30 subcarriers from business Intel5300 wireless network card, and User space can be submitted to the form of channel condition information and carry out routine processes.

In order to the impact that abates the noise, improve detection and positioning precision, adopt sliding time window mechanism.Collect the channel condition information in certain hour.Then the sub carrier amplitude of data all in sliding time window is averaged, obtain the vector that comprises all sub carrier amplitude averages, it can be used as signal characteristic.

In a step 102, the signal characteristic of detection signal is extracted from moving window, calculate the correlativity of detection signal feature and static criteria signal characteristic, if the correlativity on every root receiver antenna is greater than the threshold value of setting, indicate that human body appears at monitored area, then enter step 3, position of human body is positioned, otherwise continue to carry out input to monitored area;

Said static criteria signal characteristic is that nobody exists the signal characteristic in situation in guarded region, and static criteria signal characteristic is obtained by systematic learning; For static criteria signal characteristic, the coefficient of autocorrelation of the signal characteristic in all packets in sliding time window is all less than setting threshold value.

Calculate detection signal feature H and static criteria signal characteristic H _norcorrelativity, its computing formula is as follows:

$C=\mathrm{corr}(H,H_{\mathrm{nor}})=\frac{\mathrm{\Σ}(H-\stackrel{\‾}{H_{\mathrm{nor}}})(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}{\sqrt{\mathrm{\Σ}{(H-\stackrel{\‾}{H})}^2\mathrm{\Σ}{(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}^2}}$

When the correlativity on every root antenna is all greater than the threshold value of setting, then think that in surrounding environment, nobody occurs, otherwise illustrate have people to appear in surrounding environment.Then system performs positioning function, orients the position at human body place.

Exemplary, carry out also comprising before detection and positioning appears in human body setting up static criteria signal characteristic file described.

For the process of establishing of static criteria signal characteristic file, when haplopia is unmanned near communication link, signal receiver extracts channel condition information from the 802.11n protocol signal that signal transmitter is launched, then channel condition information is sent to the equipment of the method performing human detection and location, this equipment reads signal characteristic, calculate the coefficient of autocorrelation of signal characteristic in each receiver antenna up-sampling cycle, its computing formula is as follows:

$C_k=\frac{2}{N*(N-1)}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\underset{j=i+1}{\overset{N}{\mathrm{\Σ}}}\mathrm{corr}(H_j,H_i),k\∈[1,n]$

Wherein n is receiver antenna number, and N is packet number in the employing cycle.When the coefficient of autocorrelation of the signal characteristic on every root receiver antenna is less than the threshold value of setting, then think that this environment is stable.In packet in this sampling period cumulative, the amplitude of each subcarrier asks its mean value, and its computing formula is as follows:

$H_{\mathrm{nor}}^f=\frac{{\mathrm{\Σ}}_{i=1}^N{H}_i^f}{N}$

When the equal value set of amplitude of all subcarriers of root receiver antenna every in acquisition sample, be then stored in static criteria signal characteristic file, static criteria signal characteristic file structure is as shown in table 1, occurs that detection & localization uses for online human body.

Table 1 static criteria signal characteristic file structure

In step 103, read the parameter information in location model Parameter File, then calculating detection signal feature can the probability at position location place at each, and the position that maximum probability is correlated with is the place that human body is most possibly positioned at.

Parameter information in location model Parameter File is average μ and the standard deviation sigma of the Euclidean distance of detection signal and static criteria signal characteristic on each position, every root receiver antenna, and on receiver antenna, the Euclidean distance value of detection signal feature and static criteria signal characteristic meets Gaussian distribution.

Calculating detection signal feature can the probability at position location place at each, and its computing formula is as follows:

$P\left(H\right|L)=\underset{k=1}{\overset{n}{\mathrm{\Π}}}\underset{-\∞}{\overset{+\∞}{\∫}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_k}\exp (-\frac{{(x-{\mathrm{\μ}}_k)}^2}{2{\mathrm{\σ}}_k^2})\mathrm{dx}$

Wherein n is receiver antenna number.

After calculating detection signal probability possible on each position, by contrast probability size, the place that human body most probable occurs is regarded as in the position of maximum probability being correlated with:

$L=\underset{x}{\max }P\left(H\right|L_x)$

Exemplary, described carry out human body location before also comprise and set up location model Parameter File.

First, utilize and comprise human body can off-line data under the situation of position location, by each position feature data into clusters, said position feature data refer to the Euclidean distance of detection signal feature and static criteria signal characteristic on all receiver antennas, and its dimension size is identical with receiver antenna number.

Receiver collection comprises the data sample of the signal characteristic of human body on diverse location.This process need operating personnel stand with two ends and walk about a few minutes around link, and receiver is automatically collected signal data and sent it to the equipment performing human detection and localization method and stores.On the equipment performing human detection and localization method, system reads the channel condition information of each packet in off-line files, extract signal characteristic, then calculate the Euclidean distance of signal characteristic and static criteria signal characteristic on each root receiver antenna, its computing method are as follows:

$d_k=d(H_{k,i},H_{k,\mathrm{nor}})=\sqrt{\frac{\underset{f=1}{\overset{30}{\mathrm{\Σ}}}{(H_{k,i}^f-H_{k,\mathrm{nor}}^f)}^2}{30}},$ Wherein k is antenna serial number.

The signal of each like this sliding time window can be expressed as: w _i={ d ₁, d ₂..., d _n.Off-line data can be expressed as Euclidean distance set a: S={w ₁, w ₂..., w _t.Euclidean distance S set is the input data that algorithm is distinguished in position.The unsupervised learning technology (what adopt in the invention process process is DBSCAN technology) performing the equipment utilization density based of the method for human detection and location carries out cluster to data S set.After cluster, the dimension of each bunch is identical with receiver antenna number.Then position judgment rule is utilized to identify each bunch of relevant classification belonging to position.

Then, each orientable position of position principle of distinction identification is utilized;

In this example, so-called position distinguishing rule is: when human body is near transmitter, and the Euclidean distance of the signal characteristic on each receiver antenna and standard static signal characteristic is comparatively large, and otherness between these Euclidean distances is less; When human body is near receiver, larger and other then relatively little of the Euclidean distance of the signal characteristic on some receiver antenna and standard static signal characteristic, the otherness of these Euclidean distances is larger; When human body is positioned on the obstructed path near los path, the Euclidean distance of the signal characteristic on each receiver antenna and standard static signal characteristic is less, and the otherness of these Euclidean distances is also less; When in the middle part of human body is positioned at los path, signal characteristic on each receiver antenna is little near transmitter of human body compared with the Euclidean distance of standard static signal characteristic, and the otherness between these Euclidean distances compares smaller near receiver of human body.In order to portray position distinguishing rule with numerical value, invention introduces Two Variables: bunch centre distance D and antenna difference V.For some bunch of x, its central point can be represented as: w _x={ d _{x, 1}, d _{x, 2}..., d _x,n, then the computing formula of bunch centre distance is:

$D_x=\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left|d_{x,k}\right|}^2}$

The computing formula of antenna difference is:

$V_x=\frac{2}{n*(n-1)}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\underset{l=k}{\overset{n}{\mathrm{\Σ}}}|d_{x,k}-d_{x,l}|$

Wherein n is receiver antenna number.

Finally, the average of each position feature data in each dimension and standard deviation is calculated, stored in file after it being associated with position.

Because the Euclidean distance value of the signal characteristic on every root antenna and standard static signal characteristic meets Gaussian distribution, therefore relevant bunch can being expressed as in each position: L _x=(μ, σ), wherein μ, σ are distributed as average and the standard deviation of Gaussian distribution.Therefore system is after judging each position, extracts the gaussian distribution characteristic of each position, is stored in positional parameter configuration file, and concrete form is with reference to table 2.

Table 2 location model Parameter File

In this example, unsupervised learning technology is utilized to carry out the specific algorithm (algorithm 1 position distinguish algorithm) of position differentiation as follows.

As shown in Figure 2, can being divided into position location of the present invention's definition: on the non line of sight part of los path, los path on receiver, los path in the middle part of transmitter, los path.On los path, close Receiver And Transmitter part is the place that intrusion detection and defence system are extremely paid close attention to.When human body is near receiver or transmitter, near the distinct device that security system will be according to human body, carry out point other process.

As shown in Figure 3, during concrete enforcement, signal transmitter (TX) adopts the commercial wireless router (AP) that application is the most universal at present, and signal receiver (RX) is general commercial notebook computer, this computer is configured with Intel NIC 5300 network interface card, support 802.11n agreement, (SuSE) Linux OS and Linux CSI Tool instrument are installed.Linux CSI Tool comprises a driver iwlwifi for Intel NIC 5300 network interface card, it can obtain the channel response information of 30 subcarriers in ofdm system, and can submit to corresponding program and process with channel condition information (CSI) form.Signal transmitter and signal receiver are generally fixed on the position of overhead height 1.2m, can carry out detection and positioning more accurately like this to the appearance of human body.Signal receiver sends ICMP request message with given pace (being such as set as 20 packets p.s.) to signal transmitter, then signal receiver utilizes Linux CSI Tool to drive and obtain CSI information from the response packet of signal transmitter, and the CSI information of collection is sent to the device of human detection and location by udp protocol in real time.During system is disposed, first tester should be arranged systematic parameter, comprises the corresponding informations such as antenna number, ICMP packet sending speed, application server IP address, detection threshold, clustering algorithm parameter.

After deployed with devices completes, installation personnel needs to carry out two operations.Section 1 gathers static criteria signal characteristic.Installation personnel opens standard static signal characteristic acquisition function, and guarantees unmanned in monitored area appearance.Receiver can launch icmp packet with the frequency of 20Hz to transmitter, and receiver obtains channel condition information by network interface card firmware, and this channel condition information is uploaded to User space data transmission blocks.Data transmission blocks utilizes udp protocol, channel condition information is sent to detection and positioning equipment.After the data that detection & localization system data receiver module receiver sends, CSI Tool instrument is utilized to extract signal characteristic.Calculate the coefficient of autocorrelation of the signal characteristic of every root antenna in sliding time window buffer zone, moving window size is 500ms, and corresponding buffer size is 10 data package sizes.When the signal characteristic of every root antenna coefficient of autocorrelation lower than setting threshold value 0.996 after, system will stop data collection.System can calculate the amplitude average of each subcarrier in signal characteristic on this moving window each antenna interior automatically, and is stored in static criteria signal characteristic file.Section 2 collects the off-line data of relevant position.After installation personnel opens off-line data collecting function, tester freely walks in monitored area, and in the middle part of receiver, transmitter, link, near the position of los path, stops one minute on los path.Gatherer process probably needs about 5 minutes.Then off-line data is saved in journal file by system.After installation personnel enable position recognition function, off-line data is read in location identification module, and system will extract the signal characteristic of each packet in off-line data, by it stored in memory array.When after the signal characteristic extracting all packets, distinguished in algorithm by signal characteristic array input position, wherein DBSCAN algorithm parameter MinPts is that 3.5, Eps then takes adaptive strategy.System will distinguish the structure output of result according to table 2 in location model Parameter File after run location distinguishes algorithm.

After completing environmental adaptation process, system opens online detection and positioning function.Receiver sends ICMP bag with the frequency of 20Hz per second to receiver, and obtains channel information from network interface card, and utilizes udp protocol it to be sent in real time the equipment performed without supervisory detection and positioning system by data transmission blocks.When after detection & localization system acceptance to channel information, utilize CSI Tool instrument to extract the signal characteristic of packet, and be deposited in window buffer.After buffer zone is filled, system reads signal characteristics all in buffer zone, asks the amplitude average of each subcarrier on its every root receiver antenna.Then system reads the signal characteristic in static criteria file, calculates the correlativity of new signal characteristic on every root receiver antenna and standard static signal characteristic.If correlativity is greater than detection threshold 0.995, by buffer empty, continue next round and detect; Otherwise meaning has human body to appear in monitored area.After having human body to appear at monitored area, system reads the parameter information in location model Parameter File, calculates the probability of detection signal feature on the every root antenna in each position, then accumulates the probability that this place has antenna corresponding.So just can obtain detection signal feature corresponding each can the probability at position location place, and maximum probability value corresponding can position location be the place at the most possible place of human body.

Claims (8)

1., without supervision human detection and a localization method, transmitter sends signal, and signal, through monitored area, is received by receiver, it is characterized in that:

Step 1: Real-time Collection channel condition information data, averages to the sub carrier amplitude of data all in sliding time window, obtains the vector that comprises all sub carrier amplitude averages, it can be used as signal characteristic;

Step 2: the signal characteristic extracting detection signal from moving window, calculate the correlativity of detection signal feature and static criteria signal characteristic, if the correlativity on every root receiver antenna is greater than the threshold value of setting, indicate that human body appears at monitored area, then enter step 3, position of human body is positioned, otherwise continues to carry out input to monitored area;

Step 3: read the parameter information in location model Parameter File, then calculating detection signal feature can the probability at position location place at each, and the position that maximum probability is correlated with is the place that human body is most possibly positioned at.

2. according to claim 1ly it is characterized in that: in step 2 without supervision human detection and localization method, said static criteria signal characteristic is that in guarded region, nobody exists the signal characteristic in situation, and static criteria signal characteristic is obtained by systematic learning; For static criteria signal characteristic, the coefficient of autocorrelation of the signal characteristic in all packets in sliding time window is all less than setting threshold value.

3. according to claim 2 without supervision human detection and localization method, it is characterized in that: in step 2, detection signal feature H and static criteria signal characteristic H _norcorrelativity, computing formula is as follows:

$C=\mathrm{corr}(H,H_{\mathrm{nor}})=\frac{\mathrm{\Σ}(H-\stackrel{\‾}{H_{\mathrm{nor}}})(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}{\sqrt{\mathrm{\Σ}{(H-\stackrel{\‾}{H})}^2\mathrm{\Σ}{(H_{\mathrm{nor}}-\stackrel{\‾}{H_{\mathrm{nor}}})}^2}}$

4. according to claim 3 without supervision human detection and localization method, it is characterized in that: in step 3, parameter information in location model Parameter File is average μ and the standard deviation sigma of the Euclidean distance of detection signal and static criteria signal characteristic on each position, every root receiver antenna, and on receiver antenna, the Euclidean distance value of detection signal feature and static criteria signal characteristic meets Gaussian distribution.

5. according to claim 4ly it is characterized in that: in step 3 without supervision human detection and localization method, calculating detection signal feature can the probability at position location place at each, and its computing formula is as follows:

$P\left(H\right|L)=\underset{k=1}{\overset{n}{\mathrm{\Π}}}\underset{-\∞}{\overset{+\∞}{\∫}}\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_k}\exp (-\frac{{(x-{\mathrm{\μ}}_k)}^2}{2{\mathrm{\σ}}_k^2})\mathrm{dx}$

Wherein n is receiver antenna number.

It is 6. according to claim 5 that without supervision human detection and localization method, it is characterized in that: in step 3, location model Parameter File generates by the following method,

First, utilize and comprise human body can off-line data under the situation of position location, by each position feature data into clusters, said position feature data refer to the Euclidean distance of detection signal feature and static criteria signal characteristic on all receiver antennas, and its dimension size is identical with receiver antenna number;

Then, each orientable position of position principle of distinction identification is utilized;

Finally, the average of each position feature data in each dimension and standard deviation is calculated, stored in file after it being associated with position.

7. according to claim 6 without supervision human detection and localization method, it is characterized in that: can to comprise on the position, los path of los path on the position, los path of receiver near the position of transmitter, los path medium position position location.

8. according to claim 7 without supervision human detection and localization method, it is characterized in that: position principle of distinction is, when human body is positioned near los path, the signal characteristic of every root antenna and the Euclidean distance of static criteria signal characteristic less, and the signal characteristic between different antennae differs less with the Euclidean distance of static criteria signal characteristic; When human body is positioned at close transmitter one end on los path, the signal characteristic of every root antenna and the Euclidean distance of static criteria signal characteristic are comparatively large, and the signal characteristic between different antennae differs less with the Euclidean distance of static criteria signal characteristic; When human body is positioned at close receiver one end on los path, the signal characteristic of certain root antenna and the Euclidean distance of static criteria signal characteristic are comparatively large, and the signal characteristic between different antennae differs larger with the Euclidean distance of static criteria signal characteristic; When in the middle part of human body is positioned at los path, signal characteristic on each receiver antenna is little near transmitter of human body compared with the Euclidean distance of standard static signal characteristic, and the otherness between these Euclidean distances compares little near receiver of human body.