CN108903936A - The Intelligent mining helmet control method merged based on human body information and environmental information - Google Patents
The Intelligent mining helmet control method merged based on human body information and environmental information Download PDFInfo
- Publication number
- CN108903936A CN108903936A CN201810715885.1A CN201810715885A CN108903936A CN 108903936 A CN108903936 A CN 108903936A CN 201810715885 A CN201810715885 A CN 201810715885A CN 108903936 A CN108903936 A CN 108903936A
- Authority
- CN
- China
- Prior art keywords
- signal
- wavelet
- neural network
- eeg
- neuron
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
- A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
- A61B5/18—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state for vehicle drivers or machine operators
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
- A61B5/374—Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6802—Sensor mounted on worn items
- A61B5/6803—Head-worn items, e.g. helmets, masks, headphones or goggles
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6802—Sensor mounted on worn items
- A61B5/681—Wristwatch-type devices
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/74—Details of notification to user or communication with user or patient ; user input means
- A61B5/746—Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/20—Workers
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2560/00—Constructional details of operational features of apparatus; Accessories for medical measuring apparatus
- A61B2560/02—Operational features
- A61B2560/0242—Operational features adapted to measure environmental factors, e.g. temperature, pollution
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Veterinary Medicine (AREA)
- Surgery (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Psychiatry (AREA)
- Psychology (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Social Psychology (AREA)
- Educational Technology (AREA)
- Hospice & Palliative Care (AREA)
- Developmental Disabilities (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Child & Adolescent Psychology (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a kind of Intelligent mining helmet control method merged based on human body information and environmental information, the acquisition including original eeg signal；The pretreatment of original brain wave；The Wavelet decomposing and recomposing of original eeg signal；Brain wave signal feature extraction；Artificial fish school algorithm BP neural network identifies fatigue characteristic；Identification miner's fatigue strength, carries out alarm operation to dangerous situation at monitoring miner's local environment information.The present invention extracts four sub-belt energies, ApEn approximate entropy, KC complexity and C in the high-precision eeg signal obtained afterwards to original eeg signal pretreatment (filtering, denoising)0Input layer of the complexity as artificial fish school algorithm BP neural network, training artificial fish-swarm algorithm Optimized BP Neural Network identification miner's fatigue, recognition accuracy is high, and triple alarms of the wearable bracelet of the intelligence by being worn in miner's wrist to the safe condition carry out sound of different danger classes, three kinds of light, vibration modes, to improve downhole operations safety coefficient.
Description
Technical field
It is especially a kind of to be merged based on human body information and environmental information the present invention relates to miner's fatigue identification technology field
Intelligent mining helmet control method.
Background technique
China is maximum producing coal country in the whole world.It is traditional on the basis of the technical background in existing mine
The mining helmet is no longer satisfied the market demand, so needing that Digital Mine monitoring system is combined to carry out intelligence to the mining helmet
Upgrading.The mining intelligent helmet is studied, the monitoring and early warning of underground complex environment can be completed, while providing mine for monitoring room on well
Tool body information, is the exclusion of accident, and rescue and analysis on accident cause provide reliable data.Existing mining helmet monitoring
System is as follows：
1, mine alarm mode is mounted in tunnel more and carries out sound-light alarm, for deviation warning source (such as in driving one
In the environment of line excess noise, more dust) personnel for, there is situations such as receiving warning message not in time, i.e., information is real-time
Property is unable to ensure.
2, underground coal mine sensor is often monitored environmental information etc. in fixed position, when personnel pass by this position
When setting, the information etc. of itself local environment can not be just recognized, there are security risks.
3, the existing mining helmet all only completes monitoring, the voice communication etc. of miner's local environment mostly, does not have substantially
Miner itself is monitored, more not to the check and evaluation of miner's fatigue strength etc..
Summary of the invention
The invention aims to solve the deficiencies in the prior art, provide a kind of based on human body information and environment letter
Cease the Intelligent mining helmet control method of fusion.
In order to achieve the above objectives, the present invention is implemented according to following technical scheme：
A kind of Intelligent mining helmet control method merged based on human body information and environmental information, is included the following steps：
The acquisition of S1, original eeg signal：Mine is obtained by the EEG signals acquisition device being mounted on the mining helmet
The original eeg signal of work simultaneously pre-processes it, and transmission is mounted on rear side of mining ontology after obtaining original eeg signal
EEG Processing device；Meanwhile using infrared compound five in one gas sensor, DHT11 Temperature Humidity Sensor to titanium dioxide
Carbon, methane, carbon monoxide, formaldehyde, volatile organic matter concentration measure, and gas parameter is transferred to master controller；
The EEG signals acquisition device includes that forehead on the right side of brain is obtained on the inner surface for be mounted on mining helmet main body
Current potential at first electrode for encephalograms of the leaf site state of mind, the ear-lobe of acquisition and shield reference signal the second electrode for encephalograms and
The third electrode for encephalograms of the following artefact signal of brain is shielded, and is mounted on rear side of mining helmet main body for the first brain electricity electricity
The signal of pole, the second electrode for encephalograms and the acquisition of third electrode for encephalograms carries out pretreated EEG Processing device, at EEG signals
The signal output end of reason device connects with the signal input part of master controller；
S2, the denoising of original eeg signal, process are as follows：
S21, in EEG Processing device, selection with the most similar wavelet basis function of original signal, determine wavelet transformation
Decomposition scale, wavelet transformation is carried out to signal using Mallat tower process, and then obtains the high frequency system of different decomposition scale
Number component and low frequency coefficient component；
Wavelet threshold on S22, calculating jth scale, compares the Wavelet Entropy of n subband signal, it is maximum to choose small echo entropy
Subband wavelet coefficient, it is believed that the wavelet coefficient of the subband is to calculate the intermediate value of the subband wavelet coefficient as caused by noise,
As the noise variance of jth scale, so as to which the wavelet threshold of jth scale is calculated；
S23, due to the noise wavelet coefficients value on different scale it is different, with the increase of decomposition scale, the small echo of noise
Coefficient is smaller and smaller, so calculate separately the wavelet threshold of different scale by S22, and to the high frequency coefficient component of each scale into
The processing of row thresholding, obtains approximate high-frequency wavelet coefficient；
S24, the low frequency coefficient component of maximum layer wavelet decomposition and the approximation of the different scale Jing Guo threshold process are utilized
High-frequency wavelet coefficient component, composition carry out coefficient component required for signal reconstruction, carry out by the reconstruct formula of multiresolution analysis
Reconstruct, obtains pure eeg signal；
S3, characteristics of EEG extract, and process is as follows：
S31, wavelet basis is chosen to eeg signal progress Wavelet decomposing and recomposing：EEG Processing device utilizes sym5 small echo
The eeg signal that basic function obtains module acquisition to EEG signals carries out wavelet decomposition, and signal decomposition expression formula is：
Detail coefficients are in formula：It is approached in formula
Coefficient is：Formula mesoscale function is：Formula Wavelets are：Then final signal reconstruct expression formula is：
S32, six layers of decomposition are carried out to brain wave signal using sym5 wavelet function, respectively obtains signal six layers of low frequency a6, six
Layer high frequency d6, five layers of high frequency d5, four layers of high frequency d4；Six layers of low frequency a6, six layers of high frequency d6, five are found by the frequency detecting of FFT
Layer high frequency d5, four layers of high frequency d4 are similar to the frequency of δ wave, θ wave, α wave, β wave respectively, therefore using the signal of wavelet decomposition come generation
Table δ wave, θ wave, α wave, β wave；
S33, δ (x), θ (x), α (x), β (x) subband signal that wavelet decomposition obtains different frequency bands, subband are carried out to initial data
Energy is：
Calculate the energy ratio of each signal：Eall=E (δ)+E (θ)+E (α)+E (β), En (δ)=E (δ)/Eall, En (θ)=E (θ)/
Eall, En (α)=E (α)/Eall, En (β)=E (β)/Eall, go out miner in fatigue state, waking state so as to quantitative analysis
Ratio shared by subsignal with the EEG under collected state；
S34, ApEn approximate entropy, KC complexity and C are calculated0Complexity compares fatigue state, waking state, collected state
The situation of change of these three characteristic values of lower eeg signal, and several groups fatigue characteristic value, several groups are regained consciousness characteristic value, several
Group concentrates characteristic value to distinguish list, saves, the subsequent input data as neural network；
S4, artificial fish school algorithm BP neural network identify that process is as follows to fatigue characteristic：
S41, by four sub-belt energies, ApEn approximate entropy, KC complexity and C obtained in S30Complexity is as nerve net
The input layer data of network represents tired, awake, collected state as nerve net with (1,0,0), (0,1,0), (0,0,1) respectively
The output layer of network；
S42, it sets input layer in BP neural network and has M neuron, hidden layer has J neuron, and output layer has K nerve
Member, input layer are expressed as w to the weight between hidden layerij, the threshold values of hidden layer neuron is expressed as bj, hidden layer to output layer
Between weight be expressed as vjk, the threshold values of output layer neuron is expressed as ak.The parameter to be adjusted in BP neural network is weight
wij、vjkWith threshold values bj、ak, the parameter that these to be adjusted is set as Artificial Fish state, then Artificial Fish x can be expressed as a M*J+
J+J*K+K dimensional vector：
X=(w11,…,wM1,b1,…,wiJ,…,wMJ,bJ,v11,…,vJ1,a1…,v1k,…,vJK,aK), wherein w11,…,
wM1Weight of the expression input layer to first hidden layer neuron, the threshold values of b1 expression first neuron of hidden layer,
wiJ,…wMJIndicate weight of the input layer to j-th hidden layer neuron, the valve of bJ expression hidden layer j-th neuron
Value, v11,…,vJ1Indicate weight of the hidden layer neuron to first output layer neuron, a1 expression first neuron of output layer
Threshold values, v1k,…,vJKIndicate weight of the hidden layer neuron to k-th output layer neuron, aKIndicate output layer k-th nerve
The threshold values of member, the food concentration FC of Artificial Fish are set as the inverse of BP neural network overall error E, i.e. FC=1/E, such Artificial Fish institute
The maximum point of the food concentration to be found is exactly the smallest point of BP neural network error, any two Artificial Fish x1、x2Between Europe
Distance d is obtained in several to be expressed as：
X in formula1、x2In element denounce to correspond according to dimension and subtract each other, Artificial Fish x behavior and is chased after executing foraging behavior, bunch
The state of itself will be changed after tail behavior, while carrying out the primary adjustment of initial weight and threshold values to BP neural network；
S43, master controller are according to the characteristics of BP neural network error function and the design experiences of forefathers, Binding experiment,
The neuron for choosing intermediate hidden layer is 20, and sample input is the data of 17 dimensions, and network output is 1 dimension, the i.e. knot of neural network
Structure is set as 17-20-1；
It is 10 that Artificial Fish number is selected in S44, Artificial Fish optimization algorithm, visible range 0.5, crowding factor 0.618, step
Long to be selected as 0.1, the number of iterations 50 repeats to explore number to be 50；BP neural network activation primitive 1 selects tansig function,
Activation primitive 2 selects logsig function, and training function uses traindx function, and maximum frequency of training is set as 20000 times, is learned
Habit rate is 0.05, and error requirements precision is 10-4, factor of momentum 0.9；
S45, to binary function(- 10≤x, y≤10) are fitted to test artificial fish school optimization BP
The performance of network：Taking step-length is 0.1, altogether 40000 data, wherein 35000 are done training sample, 5000 do test sample,
100 training are carried out to data；
S46,2100 groups of data are chosen from eeg data library, 600 groups of collected state data, 600 groups of waking states, 600
Group fatigue state, this 1800 groups are used for network training, remaining 300 groups are used for network test, and grouping is carried out by the way of intersecting.
The state of eeg signal is identified using the neural network of above-mentioned steps training test, is realized to personnel in the pit's fatigue strength
Monitoring；
The concentration data that S5, master controller handle various sensors judges the state that transfinites of gas, merges out current peace
Total state and danger classes, while the interface that monitoring host computer is shown is sent to by the existing communication node in underground, and pass through
The Intelligent bracelet being worn in miner's wrist that bluetooth module and master controller connect to the safe conditions of different danger classes into
Row sound, light, the triple alarms for shaking three kinds of modes.
Compared with prior art, beneficial effects of the present invention：
1, EEG signals noise-removed threshold value under strong background noise is adaptive selected using maximum wavelet entropy in the present invention, obtains one
Kind of the Weak Signal Detection Method based on Wavelet Entropy, and realize that fatigue is special in the subband signal energy information of δ wave, θ wave, α wave, β wave
The extraction of sign.
2, the present invention utilizes the function optimization BP neural network initial weight of artificial fish-swarm algorithm global optimizing, obtains people
The work shoal of fish-BP neural network sorter model.Simulation comparison discovery, artificial fish-swarm-BP neural network model compare BP neural network
The recognition correct rate of model improves 2.3%.
3, the present invention has built the mining intelligent headgear system based on brain-computer interface, completes based on environmental parameter, spirit
The safety situation evaluation algorithm of state designs；And the wearable bracelet of intelligence is introduced into underground coal mine, realize that acousto-optic vibrates triple reports
Alert and information real-time display substantially increases safe early warning level, the generation of accident can be effectively reduced, to mine safety production
It is significant.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is the pure eeg signal figure that the present invention is extracted using Wavelet Entropy：It (a) is to be buried in strong noise background
Brain wave signal figure, (b) for brain wave signal Wavelet Entropy remove dry result figure.
Fig. 3 is energy chart shared by wavelet under different conditions of the invention：(a) fatigue state, (b) waking state (c) collect
Middle state.
Fig. 4 is ApEn, KC and C under eeg signal different conditions of the invention0Value；(a) different conditions lower aprons entropy
ApEn, (b) KC value under different conditions, (c) C under different conditions0。
Fig. 5 is artificial fish school optimization BP neural network structure chart of the invention.
Fig. 6 is artificial fish school algorithm BP neural network flow chart of the invention.
Fig. 7 is the three-dimensional figure of binary function of the invention.
Fig. 8 is artificial fish-swarm BP neural network algorithm and true value and ordinary BP nerve network test of heuristics knot of the invention
Fruit comparison diagram.
Fig. 9 is the training result comparison diagram of BP neural network of the invention and AF-BP neural network：It (a) is BP nerve net
Network training error (b) is BP neural network physical training condition, (c) is AF-BP neural metwork training error, is (d) AF-BP nerve
Network training state.
Figure 10 is the Intelligent bracelet gas explosion disaster Risk-warning grade of the embodiment of the present invention.
Figure 11 is the Intelligent bracelet data flow overall data process design cycle of the embodiment of the present invention.
Figure 12 is the mash gas methane danger signal position flow chart of the embodiment of the present invention.
Figure 13 is Intelligent bracelet danger classes flag bit bubble sort of the invention.
Figure 14 is Intelligent bracelet danger classes flag bit sequence flow chart of the invention.
Figure 15 is Intelligent bracelet danger early warning operational flowchart of the invention.
Figure 16 is Intelligent bracelet danger warning operational flowchart of the invention.
Figure 17 is the circuit structure block diagram of the mining helmet of overall intelligence of the invention.
Specific embodiment
The invention will be further described combined with specific embodiments below, in the illustrative examples and explanation of the invention
For explaining the present invention, but it is not as a limitation of the invention.
A kind of Intelligent mining helmet controlling party merged based on human body information and environmental information of the present invention as shown in Figure 1
Method includes the following steps：
The acquisition of S1, original eeg signal：Mine is obtained by the EEG signals acquisition device being mounted on the mining helmet
The original eeg signal of work simultaneously pre-processes it, and transmission is mounted on rear side of mining ontology after obtaining original eeg signal
EEG Processing device, EEG Processing device is powered by the rechargable power supplies being mounted in the mining helmet；Meanwhile
Using infrared compound five in one gas sensor, DHT11 Temperature Humidity Sensor to carbon dioxide, methane, carbon monoxide, formaldehyde,
Volatile organic matter concentration measures, and gas parameter is transferred to master controller；
As shown in figure 17, the EEG signals acquisition device includes obtaining greatly on the inner surface for be mounted on mining helmet main body
Second of current potential and shielding reference signal at first electrode for encephalograms of the brain right side forehead leaf site state of mind, the ear-lobe of acquisition
The third electrode for encephalograms of electrode for encephalograms and the shielding following artefact signal of brain, and be mounted on rear side of mining helmet main body for pair
The signal of first electrode for encephalograms, the second electrode for encephalograms and the acquisition of third electrode for encephalograms carries out pretreated EEG Processing device,
The signal output end of EEG Processing device connects with the signal input part of master controller；
S2, the denoising of original eeg signal, process are as follows：
It, can be on time-frequency domain since small echo entropy theory is the theory for setting up similar information entropy based on wavelet analysis method
Power distribution properties are quantitatively described.The coefficient matrix of wavelet transformation is processed into a probability distribution sequence, is calculated by it
Obtained entropy just reflects the sparse degree of this coefficient matrix.Small echo entropy theory is the sparse journey using wavelet transform matrix
Degree is to inhibit unrelated ingredient.
According to the Frame Theory of wavelet transformation, when wavelet basis function is one group of orthogonal basis, wavelet transformation is kept with energy
Permanent property, i.e.,：
Define the quadratic sum that the wavelet energy under single scale is wavelet coefficient under the scale.
By the characteristic of orthogonal wavelet transformation it is found that in sometime window, total power signal be equal to each component power it
With.
The high-frequency information amount of each decomposition scale is regarded as an individual signal source, by each layer of high frequency wavelet
Coefficient is divided into n equal minizones, calculates the Wavelet Entropy of each minizone, chooses the intermediate value of that maximum minizone of entropy
As the variance of noise, realize that the threshold adaptive based on Wavelet Entropy is chosen.
If the high-frequency wavelet coefficient of jth layer is dj(k), the wavelet coefficient on these sampled points is divided into n by sampled point N
Equal part, then the corresponding energy of the wavelet coefficient in k-th of subinterval be：
The gross energy of jth layer high-frequency wavelet coefficient is expressed as：
If the signal energy that k-th of subinterval includes probability present in gross energy on the scale is：
Then defining the corresponding signal Wavelet Entropy in k-th of subinterval is：
The most similar wavelet basis function of S21, selection and original signal, determines the decomposition scale of wavelet transformation, utilizes
Mallat tower process carries out wavelet transformation to signal, and then obtains the high frequency coefficient component and low frequency coefficient of different decomposition scale
Component；
Wavelet threshold on S22, calculating jth scale, compares the Wavelet Entropy of n subband signal, it is maximum to choose small echo entropy
Subband wavelet coefficient, it is believed that the wavelet coefficient of the subband is to calculate the intermediate value of the subband wavelet coefficient as caused by noise,
As the noise variance of jth scale, so as to which the wavelet threshold of jth scale is calculated；
S23, due to the noise wavelet coefficients value on different scale it is different, with the increase of decomposition scale, the small echo of noise
Coefficient is smaller and smaller, so calculate separately the wavelet threshold of different scale by S22, and to the high frequency coefficient component of each scale into
The processing of row thresholding, obtains approximate high-frequency wavelet coefficient；
S24, the low frequency coefficient component of maximum layer wavelet decomposition and the approximation of the different scale Jing Guo threshold process are utilized
High-frequency wavelet coefficient component, composition carry out coefficient component required for signal reconstruction, carry out by the reconstruct formula of multiresolution analysis
Reconstruct, obtains pure eeg signal；The pure eeg signal extracted using Wavelet Entropy, as shown in Figure 2.
S3, characteristics of EEG extract, and process is as follows：
S31, wavelet basis is chosen to eeg signal progress wavelet decomposition：EEG Processing device utilizes sym5 wavelet basis letter
The eeg signal that several pairs of EEG signals obtain module acquisition carries out Wavelet decomposing and recomposing, and signal decomposition expression formula is：
Detail coefficients are in formula：It is approached in formula
Coefficient is：Formula mesoscale function is：Formula Wavelets are：Then final signal reconstruct expression formula is：
S32, six layers of decomposition are carried out to brain wave signal using sym5 wavelet function, respectively obtains signal six layers of low frequency a6, six
Layer high frequency d6, five layers of high frequency d5, four layers of high frequency d4；Six layers of low frequency a6, six layers of high frequency d6, five are found by the frequency detecting of FFT
Layer high frequency d5, four layers of high frequency d4 are similar to the frequency of δ wave, θ wave, α wave, β wave respectively, therefore using the signal of wavelet decomposition come generation
Table δ wave, θ wave, α wave, β wave；
S33, δ (x), θ (x), α (x), β (x) subband signal that wavelet decomposition obtains different frequency bands are carried out to initial data,
Its sub-belt energy is：
Calculate the energy ratio of each signal：Eall=E (δ)+E (θ)+E (α)+E (β), En (δ)=E (δ)/Eall, En (θ)=E (θ)/
Eall, En (α)=E (α)/Eall, En (β)=E (β)/Eall, go out miner in fatigue state, waking state so as to quantitative analysis
Ratio shared by subsignal with the EEG under collected state；Sub-belt energy under different conditions is as shown in table 1.
The energy of each subband under 1 different conditions of table
It is as shown in Figure 3 that result is normalized in energy value in table 1.As can be seen from Figure 3 δ under three kinds of states
Wave accounts for main component.Under fatigue state, the sub-belt energy of eeg signal focuses primarily upon δ wave；The energy of δ wave under waking state
Amount decline, other wave energies rise；Under collected state, α wave energy is than more prominent.From the point of view of frequency, with the state of mind
Raising, low frequency signal energy reduce, higher frequency signal energy increase.Using wavelet decomposition can quantitative analysis go out under different conditions
Brain wave subsignal shared by ratio, the feature extracted is obvious and easily distinguishes.
S34, ApEn approximate entropy, KC complexity and C are calculated0Complexity, as shown in Figure 4.Compare fatigue state, awake shape
The situation of change of these three characteristic values of eeg signal under state, collected state, and several groups fatigue characteristic value, several groups are regained consciousness
Characteristic value, several groups concentrate characteristic value to distinguish list, save, the subsequent input data as neural network；Wherein ApEn
Approximate entropy, KC complexity and C0The calculation method of complexity is specially：
(1) ApEn approximate entropy
Need artificially to be arranged two parameters in approximate entropy calculating process, one is to embody signal sequence in different dimensions
The mode dimension m of lower complexity variation characteristic, another is similar tolerance r, is according to circumstances adjusted in programming process, still
Once obtaining ideal r value, the two parameters are kept fixed constant in next program.Approximate entropy algorithm steps are as follows：
1) it sets raw data set and is combined into { s(1),s(2),…,s(N)N number of data point in total.
2) by sequence { s(i)It is arranged successively composition m n dimensional vector n S in order(i), S(i)From S (1) to S(N-m+1), wherein：
S(i)=[s(i),s(i+1),…,s(i+m-1)], i=1~N-m+1
Vector S is calculated to each i value defined above(i)With remaining vector S(j)The distance between：
3) similar tolerance r is set, to the value of each i≤N-m+1, counts d [S(i),S(j)] it is less than the number and this number of r
With the ratio of distance sum N-m+1, it is denoted asI.e.：
4) willLogarithm is taken, then it is asked to be denoted as φ to the average value of all im(r) i.e.：
5) dimension m is added 1, repeats above-mentioned (1) and is calculated to (4) stepAnd Φm+1(r), then approximate entropy can be determined
Justice is：
ApEn (m, r)=Φm(r)-Φm+1(r)
In present invention experiment, m=2, r=0.1*Std are set, wherein Std is the standard deviation of initial data s (i).
Original E.E.G is carried out except noise processed, the sampling frequency of eeg signal first before being analyzed eeg signal
Rate is 512Hz, i.e. 512 data of acquisition per second, extracts eeg signal under three kinds of states respectively, the time for acquiring 30s is total
Total 512*30=15360 data calculate shown in approximate entropy per second such as Fig. 4 (a), in terms of above-mentioned curve graph according to above-mentioned algorithm
There is apparent difference between three kinds of states out, the approximate entropy under fatigue state is concentrated mainly between 0~0.2, approximate entropy
Very little shows that the unordered degree of human thinking reduces, and complexity reduces.Awake then higher by one with approximate entropy under collected state
As be higher than 0.55 or more, show that human thinking is very active, E.E.G fluctuation range is big, and complexity is high, by the calculating of approximate entropy compared with
The good complexity characteristics for having reflected eeg signal have embodied the inherent nonlinear characteristic of EEG signals, in human body
The accuracy of identification of state, accuracy, which increase, becomes the index for measuring fatigue state.
(2) KC complexity
In the algorithm of Kasper design, timing signal sequence is subjected to coarse processing, i.e. each original sequence first
Column point is all replaced with a bit, then entire signal sequence has actually reformed into (0, a 1) sequence, it is specific thick
Being granulated rule is：The average value for calculating a period of time sequence is set to 1 if the value in sequence is greater than this average value, otherwise with 0
To indicate.It can be obtained by one section of new time series by the processing of the method.
It is designed by algorithm, so that it may analyze the different mode number for being included of above-mentioned coarse sequence, specific algorithm mistake
Journey is as follows：
1) increase by one below in (0,1) sequence A (i.e. A takes s1 to A usually since first character) of one section of coarse
Character B；
2) judge whether B belongs to character string ABC (ABC is the result that AB subtracts last character), if B is in front
Being occurred then claiming B is a substring of ABC, does not occur new mode, this process is known as reproduction process；
3) above-mentioned B is added to the back of entire character string, continues growing the length of B, judged, if B is at this time
It does not appear in character string ABC, a new mode has occurred then may be considered, carried out insertion operation at this time, utilize
One additional character such as " * " be inserted into and B is attached to behind symbol, and two sections of character strings are divided into two sections of front and back at this time；
4) B is reconfigured, B repeats above-mentioned algorithm steps until having traversed all characters again since a character at this time
Until string, then " * " occurred in character string has reacted the mode summation of sequence.
Then complexity c (n) can use institute " * " for number of segment that all character strings are divided into indicate.Experiments have shown that mostly
The complexity c (n) of number (0,1) sequence tends to definite value：
C (n) is normalized using above-mentioned b (n), that is, it is as follows to can define complexity formula：
Ckc(n) shown in=c (n)/b (n), calculated result such as Fig. 4 (b), KC value in a state of fatigue is big as seen from the figure
It causes between 0.2~0.3, the KC value of waking state is between 0.3~0.5, between collected state 05~0.7.Brain exists
When work, nerve cell activity is opposite to be in more orderly state, and with the increase of mental fatigue degree, nerve is living to cell
Dynamic control ability weakens, and nerve cell activity excitement degree increases, and the disordering of nerve cell activity is gradually aggravated, KC value
It will will increase, and show that carrying out classification using fatigue state of the KC value to human body will have certain feasibility.
(3) C0 complexity
C0What complexity defined is that irregular ingredient proportion, algorithm main thought are to resolve into sequence in sequence
Regular ingredient and irregular ingredient carry out discrete FFT to the time series { x (n), n=0,1,2 ..., N-1 } that given length is N
Transformation.
In formula, K=0,1,2 ..., N-1, if the mean-square value of { x (n), n=0,1,2 ..., N-1 } is：
Parameter r is introduced, reservation is more than r times of mean-square value of frequency spectrum, and rest part is set to zero：
Carrying out inverse Fourier transform to above formula has：
N=0 in formula, 1,2 ..., N-1 define C0Complexity is：
With the increase of parameter r, C0Measure value is gradually increased, this shows the increase with r, and the Rule section removed is got over
Few, measure value will accordingly increase, and after r is greater than 2, measure value is stablized, so suggesting that r value range is 5~10.In view of C0It calculates
Method calculating speed is fast, it is proposed that sequence length N is greater than 2000；Shown in calculated result such as Fig. 4 (c).Use C0It is special that value extracts brain wave
Value indicative calculates the complexity of brain electric array, by sequence regular signal and means of chaotic signals parse, brain is got over
It enlivens that means of chaotic signals contained in brain wave is more, C is obtained according to the content for calculating means of chaotic signals0Value, brain
More active then C0It is worth bigger.
S4, artificial fish school algorithm BP neural network identify that process is as follows to fatigue characteristic：
S41, by four sub-belt energies, ApEn approximate entropy, KC complexity and C obtained in S30Complexity is as nerve net
The input layer data of network represents tired, awake, collected state as nerve net with (1,0,0), (0,1,0), (0,0,1) respectively
The output layer of network；
S42, most of neural metwork training are usually very slow, because the weight of network is updated based on control information,
The selection of initial value is most important to network training, and a good initial value can reduce the training time and faster obtain neural network mould
The parameter of type.Here we select the BP neural network based on artificial fish school algorithm to train, with traditional BP algorithm area
It is not not randomly selecting network initial value or initial value is assigned into 0 value, but is first found out using the algorithm of artificial fish-swarm more excellent
Initial value more.Manually fish-swarm algorithm obtains optimal network initial weight and threshold value for BP neural network prediction, after network is trained
Anticipation function output, specific flow chart are as shown in Figure 6.
The BP neural network structure chart of artificial fish school algorithm is constructed as shown in figure 5, setting input layer in BP neural network
There is M neuron, hidden layer has J neuron, and output layer has K neuron, and input layer is indicated to the weight between hidden layer
For wij, the threshold values of hidden layer neuron is expressed as bj, hidden layer is expressed as v to the weight between output layerjk, output layer neuron
Threshold values be expressed as ak.The parameter to be adjusted in BP neural network is weight wij、vjkWith threshold values bj、ak, these to be adjusted
Parameter is set as Artificial Fish state, then Artificial Fish x can be expressed as a M*J+J+J*K+K dimensional vector：
X=(w11,…,wM1,b1,…,wiJ,…,wMJ,bJ,v11,…,vJ1,a1…,v1k,…,vJK,aK), wherein
w11,…,wM1Indicate weight of the input layer to first hidden layer neuron, b1 expression first neuron of hidden layer
Threshold values, wiJ,…wMJIndicate weight of the input layer to j-th hidden layer neuron, bJ expression hidden layer j-th mind
Threshold values through member, v11,…,vJ1Hidden layer neuron is indicated to the weight of first output layer neuron, a1 indicates output layer the
The threshold values of one neuron, v1k,…,vJKIndicate weight of the hidden layer neuron to k-th output layer neuron, aKIndicate output
The threshold values of layer k-th neuron, the food concentration FC of Artificial Fish are set as the inverse of BP neural network overall error E, i.e. FC=1/
E, the maximum point of the food concentration to be found of such Artificial Fish are exactly the smallest point of BP neural network error, and any two artificial
Fish x1、x2Between Euclidean distance d be expressed as：
X in formula1、x2In element denounce to correspond according to dimension and subtract each other, Artificial Fish x is executing foraging behavior, row of bunching
For with will change itself state after the behavior of knocking into the back, while the primary tune of initial weight and threshold values is carried out to BP neural network
It is whole；
S43, neural network input layer number determine that output layer network node is by sample by the index number of learning sample
The number of this result determines, for the number of nodes of the two levels, is determined according to particular problem.According to BP neural network error letter
The design experiences of several feature and forefathers, Binding experiment, the neuron for choosing intermediate hidden layer is 20, and sample input is 17 dimensions
Data, network output for 1 dimension, i.e., the structure setting of neural network be 17-20-1；To Artificial Fish optimization algorithm and BP nerve net
The parameter of network is chosen as shown in table 2 and table 3.
2 Artificial Fish optimization algorithm parameter of table is chosen
Table 3BP neural network parameter is chosen
It is 10 that Artificial Fish number is selected in S44, Artificial Fish optimization algorithm, visible range 0.5, crowding factor 0.618, step
Long to be selected as 0.1, the number of iterations 50 repeats to explore number to be 50；BP neural network activation primitive 1 selects tansig function,
Activation primitive 2 selects logsig function, and training function uses traindx function, and maximum frequency of training is set as 20000 times, is learned
Habit rate is 0.05, and error requirements precision is 10-4, factor of momentum 0.9；
S45, to binary function(- 10≤x, y≤10) are fitted to test artificial fish school optimization BP
The performance of network is illustrated in figure 7 the three-dimensional figure of binary function, and taking step-length is 0.1, altogether 40000 data, wherein 35000
A to do training sample, 5000 do test sample, carry out 100 training to data, test results are shown in figure 8；From test result
From the point of view of, either trained number of iterations or actual test error, artificial fish-swarm-BP algorithm will be substantially better than BP neural network
Algorithm.
2100 groups of data, 600 groups of collected state numbers are chosen in the state recognition of S46, EEG signals from eeg data library
According to 600 groups of waking states, 600 groups of fatigue states are used for network training, remaining 300 groups are used for network test, and grouping is using intersection
Mode carry out, be trained respectively using BP neural network and artificial fish-swarm-BP (AF-BP) neural network, training result is such as
Shown in Fig. 9, training comparing result is as shown in table 4 below.
4 network training result of table
Test sample is divided into 5 groups, every group of 60 measured signals are verified using BP and improved AF-BP network respectively.
Shown in its test result is as follows table 5 and table 6.
Table 5BP neural network test result
6 artificial fish-swarms of table-BP neural network test result
It is obtained from the above analysis result：The average accuracy of BP neural network is 84%, artificial fish-swarm-BP neural network mould
Type average accuracy is 86.3%, is better than tradition to the identification of eeg signal using artificial fish-swarm-BP neural network model
BP neural network, accuracy rate improves 2.3%.
S47, the state of eeg signal is identified using the neural network of above-mentioned steps training test, utilizes support
Vector machine model is assisted in identifying, and realizes the monitoring to personnel in the pit's fatigue strength；
The concentration data that S5, master controller handle various sensors judges the state that transfinites of gas, merges out current peace
Total state and danger classes, while the interface that monitoring host computer is shown is sent to by the existing communication node in underground, and pass through
Be worn on the Intelligent bracelet in miner's wrist to the safe condition carry out sound of different danger classes, light, shake three kinds of modes three
It alarms again.Specifically：
Safety of coal mines situation is divided into safety, Generally Recognized as safe, more dangerous and dangerous etc. 4 situation grades.Similarly,
Gas explosion disaster warning grade is also divided into 4 grade intervals, they are respectively：Safety, Generally Recognized as safe, it is more dangerous and
It is dangerous, it is (red to correspond respectively to level Four early warning (green), three-level early warning (white), second level early warning (yellow) and level-one early warning
Color), as shown in Figure 10.
Coal Mine Disasters Situation Assessment and early warning the result is that general description to underground coal mine system risk state, risk point
Grade is to carry out hierarchical description to safety production system risk level on the basis of reducing risk.The Comment gathers of risk class have
A variety of representations, the present invention define coal mine according to the characteristics of Safety of Coal Mine Production system and common risk stratification standard
The Comment gathers K={ " safety " of disaster Situation Assessment risk class；" safer "；" more dangerous "；" dangerous " }, it is corresponding etc.
Grade vector { 4,3,2,1 }, referring to table 7.
7 mine safety situation of table divides table
Wearable Intelligent bracelet use sound, light, vibration alarming, wherein the sound of alarm module use buzzer, one
I/O mouthfuls of controls；Light alarm indicates that four I/O mouthfuls control using green light, white lamp, amber light, red light；Vibration alarming is using vibration
Motor realizes, an I/O mouthfuls of controls, and passes through the data of bluetooth module reception processor.
This bracelet system is integrally placed at alarm module on the left side of control panel, is led to by bluetooth module and processor
Letter, bracelet box right side is mini mobile charger baby.Bracelet box is autonomous Design 3D printing manufacture, buzzer and vibrating motor sealing
Inside bracelet box, display screen is exposed in bracelet box front, and in order to ensure the stability of communication, bluetooth is fixed on outside bracelet box
Portion, while installation watch chain is dressed outside bracelet box convenient for personnel.
The following table 8 is the corresponding grade alarm condition of security postures of various grades, and four kinds of states of security postures are right respectively
Should safely, the danger signal position of four kinds of ranking vectors of sensor informations such as safer, more dangerous, unsafe conditions.System is logical
The state for judging danger signal position is crossed, different alarms is taken to operate.
8 safe condition of table and alarm
When security postures are in a safe condition, ranking vector 4, the corresponding state of mind is to concentrate, at this time light warning
For green light, buzzer is not alarmed, and vibrating motor is not alarmed；When security postures are in compared with safe condition, ranking vector 3 is right
Answer the state of mind awake, light warning is Bai Dengliang at this time, and buzzer is not alarmed, and vibrating motor is not alarmed；At security postures
In compared with unsafe condition, ranking vector 2, light warning is the bright sudden strain of a muscle of red light at this time, and buzzer interval is yowled, vibrating motor interval
Long vibration；When security postures are in unsafe condition, ranking vector 1, the corresponding state of mind is fatigue, and light warning is Huang at this time
The bright sudden strain of a muscle of lamp, buzzer hurriedly pipe, vibrating motor hurriedly short vibration.
Since the acquisition display of wearable Intelligent bracelet is closely related with environment acquisition parameter, so the present invention is to Intelligent bracelet
Data flow carries out whole design.Bracelet data flow overall data process design cycle is as shown in figure 11.
Environmental parameter has temperature, humidity, methane, carbon dioxide, carbon monoxide etc..These environmental parameters both define danger
Grade mark position and the state of mind together, are normalized to four danger classes：It is safe, safer, more dangerous, dangerous, it is fixed
Their danger signal position of justice is 4,3,2,1.By taking methane as an example, the danger signal position collecting flowchart of methane, detailed process are described
Figure is as shown in figure 12.
Wherein, CCH4 is methane concentration, and DCH4Flag is methane danger signal position.Sensor acquires gas density.If small
It is equal to 4 in 0.4% DCH4Flag, otherwise continues to judge；DCH4Flag is equal to 3 if less than 0.5%, otherwise continues to judge；
DCH4Flag is equal to 2 if less than 0.9%, and otherwise DCH4Flag is equal to 1；Then methane danger signal position DCH4Flag is exported.
Behind the danger signal position that control panel acquires out each ambient parameter data and environmental parameter, data are beaten by bluetooth
Packet gives Intelligent bracelet.Intelligent bracelet carries out bubble sort for danger classes, highest danger signal position is judged, according to highest
The sequence of hazard class is alarmed, i.e., level-one is alarmed>Secondary alarm>Three-level early warning>Level Four early warning.Danger signal bit comparison mode is as schemed
Shown in 13.
The flow chart specifically to sort is as shown in figure 14.
The first step：The danger signal position information that one second transmits is inputted, i=0, j=0 are defined, it is interim intermediate
Variable temp=0.
Second step：Judge whether i<Element total number -1 is then to enter third portion, is otherwise stored in final result most high-risk
Dangerous flag bit simultaneously terminates.
Third step：Continue to determine whether j<Element total number -1 is then to enter the 4th step, and otherwise i adds 1 certainly and returns to second
Step.
4th step：Judge whether j-th of element is greater than+1 element of jth, be ,+1 element value of jth is assigned to temp,
The value of j-th of element is assigned to+1 element of jth, and the value of temp is assigned to j-th of element, and then j returns to third step from after adding 1.It is no
Then j returns to third step from after adding 1.
It is possible thereby to judge highest danger signal position D.
After judging highest flag bit D on bracelet, start to carry out alarm operation.As D=4 and D=3, bracelet is shown
And warning algorithm is as shown in figure 15.
The first step：It initializes first, obtains highest danger signal position, next judge highest danger signal position D, if D >=
Or D≤4 are vacation, then continue to judge D >=1 or D≤2, carrying out danger warning operation simultaneously terminates, if true, entering in next step.
Second step：Judge whether D is equal to 4, if false, entering in next step, if true, green light, white Huang red light goes out, display 4
Grade alarm and environmental parameter page1, delay display 500ms show alarm level and spiritual parameter page2, subsequently into next
Step.
Third step：D=3, white lamp is bright at this time, and greenish-yellow red light goes out, and shows 3 grades of alarms and environmental parameter page1, delay
500ms shows alarm level and spiritual parameter page2, and be delayed 500ms, terminates.
As D=2 and D=1, bracelet is shown and alarm algorithm is as shown in figure 16.
The first step：Initially enter initialization, receive highest danger signal position, judge highest danger signal position D, if D >=3 or
D≤4 are very that system belongs to safe work state, operate into safe early warning, if false, continuing to judge D >=1 or D≤2, go forward side by side
The operation of row danger warning.
Second step：Into danger warning operate, further judge highest danger signal position, if D=2 be vacation, directly into
Enter in next step, if true, being operated into secondary alarm.2 grades of alarms are shown on bracelet interface, green white red light goes out, and amber light is with certain
Frequency slow flash, the long vibration of vibrating motor, buzzer yowl, and then further which parameter danger warning judgement is.With gas alarming
For, if gas danger signal position DCH4Flag=2 is that very, bracelet then shows gas alarming page, delay display
500ms, after display is completed or condition is if false, further judge；If carbon monoxide danger signal position DCOFlag=2 is
Very, then refresh display carbonic oxide alarming page, bracelet delay display 500ms, after having shown or condition is if false, again into one
Step judgement；If fatigue danger signal position Dco2Flag=2 is very, to refresh display fatigue warning page, delay display 500ms,
After having shown or condition is if false, further judge；If temperature danger signal position DtempFlag=2 is very, to refresh aobvious
Temperature displaying function alarm page, delay display 500ms, after having shown or condition is if false, further judge；If carbon dioxide is endangered
Dangerous flag bit DCO2Flag=2 be it is true, then refresh display carbon dioxide alarm page, delay display 500ms, show later or
Condition is if false, further judge.Since interface is limited, therefore methane, carbon monoxide and fatigue are only shown on flow chart,
His judgement similarly, terminates after the completion of judgement.
Third step：D=1 then shows 1 grade of alarm, and white amber light green at this time goes out, red light quick flashing, the short vibration of vibrating motor, buzzer
Short ring, then further which parameter alarm judgement is.Since danger signal position can all sound an alarm for 2 or 1, so here not
Only whether judgement symbol position is 1, the case where also promising 2, by taking gas as an example.If gas danger signal position DCH4Flag≤2 be it is true,
Then show gas alarming page, delay display 500ms, after having shown or condition is if false, further judgement；If carbon monoxide
Danger signal position DCOFlag≤2 be it is true, then show carbonic oxide alarming page, delay display 500ms, show later or item
Part is if false, further judgement；If tired danger signal position Dco2Flag≤2 are very, to show fatigue warning page, delay is aobvious
Show 500ms, after having shown or condition is if false, further judgement；Temperature danger signal position DtempFlag etc. is successively judged again.
The same judgement process for only showing methane, carbon monoxide and fatigue on the diagram, other danger signal positions judge process, and its is identical,
Terminate after the completion.
The limitation that technical solution of the present invention is not limited to the above specific embodiments, it is all to do according to the technique and scheme of the present invention
Technology deformation out, falls within the scope of protection of the present invention.
Claims (1)
1. the Intelligent mining helmet control method merged based on human body information and environmental information, which is characterized in that including following step
Suddenly：
The acquisition of S1, original eeg signal：Acquire miner's by the EEG signals acquisition device being mounted on the mining helmet
Original eeg signal simultaneously pre-processes it, is sent to after the eeg signal that obtains that treated and is mounted on mining helmet sheet
EEG Processing device on rear side of body；Meanwhile using infrared compound five in one gas sensor, DHT11 Temperature Humidity Sensor pair
Carbon dioxide, methane, carbon monoxide, formaldehyde, volatile organic matter concentration measure, and gas parameter is transferred to main control
Device；
The EEG signals acquisition device includes that forehead leaf portion on the right side of brain is obtained on the inner surface for be mounted on mining helmet main body
The second electrode for encephalograms and the shielding of current potential and shielding reference signal at first electrode for encephalograms of the position state of mind, the ear-lobe of acquisition
The third electrode for encephalograms of the following artefact signal of brain, and be mounted on rear side of mining helmet main body for the first electrode for encephalograms,
Second electrode for encephalograms and the signal of third electrode for encephalograms acquisition carry out pretreated EEG Processing device, EEG Processing device
Signal output end connect with the signal input part of master controller；
S2, the denoising of original eeg signal, process are as follows：
S21, in EEG Processing device, selection with the most similar wavelet basis function of original signal, determine wavelet transformation point
Scale is solved, wavelet transformation is carried out to signal using Mallat tower process, and then obtain the high frequency coefficient point of different decomposition scale
Amount and low frequency coefficient component；
Wavelet threshold on S22, calculating jth scale, compares the Wavelet Entropy of n subband signal, chooses the maximum son of small echo entropy
The wavelet coefficient of band, it is believed that the wavelet coefficient of the subband is to calculate the intermediate value of the subband wavelet coefficient as caused by noise, as
The noise variance of jth scale, so as to which the wavelet threshold of jth scale is calculated；
S23, due to the noise wavelet coefficients value on different scale it is different, with the increase of decomposition scale, the wavelet coefficient of noise
It is smaller and smaller, so calculating separately the wavelet threshold of different scale by S22, and threshold is carried out to the high frequency coefficient component of each scale
Value processing, obtains approximate high-frequency wavelet coefficient；
S24, the approximate high frequency of the low frequency coefficient component of maximum layer wavelet decomposition and the different scale Jing Guo threshold process is utilized
Wavelet coefficient component, composition carry out coefficient component required for signal reconstruction, are reconstructed by the reconstruct formula of multiresolution analysis,
Obtain pure eeg signal；
S3, characteristics of EEG extract, and process is as follows：
S31, wavelet basis is chosen to eeg signal progress wavelet decomposition：EEG Processing device utilizes sym5 wavelet basis function pair
The eeg signal that EEG signals obtain module acquisition carries out wavelet decomposition and reconstruct, signal decomposition expression formula are：
Detail coefficients are in formula：Coefficients of Approximation is in formula：Formula mesoscale function is：Formula Wavelets are：
Then final signal reconstruct expression formula is：
S32, six layers of decomposition are carried out to brain wave signal using sym5 wavelet function, respectively obtains six layers of low frequency a6 of signal, six layers of height
Frequency d6, five layers of high frequency d5, four layers of high frequency d4；Six layers of low frequency a6, six layers of high frequency d6, five layers of height are found by the frequency detecting of FFT
Frequency d5, four layers of high frequency d4 are similar to the frequency of δ wave, θ wave, α wave, β wave respectively, therefore represent δ using the signal of wavelet decomposition
Wave, θ wave, α wave, β wave；Frontal lobe area brain wave f (n) obtains F (k) by Fast Fourier Transform (FFT).
In formula, f (n) is the frontal lobe area E.E.G discrete signal of E.E.G acquisition module acquisition, and n is the serial number of sampled point, and N is sampled point
Sum, k is integer.
S33, δ (x), θ (x), α (x), β (x) subband signal that wavelet decomposition obtains different frequency bands, son are carried out to initial data
It is with energy：
Calculate the energy ratio of each signal：Eall=E (δ)+E (θ)+E (α)+E (β), En (δ)=E (δ)/Eall, En (θ)=E (θ)/
Eall, En (α)=E (α)/Eall, En (β)=E (β)/Eall, go out miner in fatigue state, waking state so as to quantitative analysis
Ratio shared by subsignal with the EEG under collected state；
S34, ApEn approximate entropy, KC complexity and C are calculated0Complexity compares fatigue state, waking state, collected state hypencephalon electricity
The situation of change of these three characteristic values of wave signal, and several groups fatigue characteristic value, the awake characteristic value of several groups, several groups are concentrated
Characteristic value distinguishes list, saves, the subsequent input data as neural network；
S4, artificial fish school algorithm BP neural network identify that process is as follows to fatigue characteristic：
S41, by four sub-belt energies, ApEn approximate entropy, KC complexity and C obtained in S30Complexity is as the defeated of neural network
Enter layer data, represents tired, awake, collected state as the defeated of neural network with (1,0,0), (0,1,0), (0,0,1) respectively
Layer out；
S42, it setting input layer in BP neural network and has M neuron, hidden layer has J neuron, and output layer has K neuron,
Input layer is expressed as w to the weight between hidden layerij, the threshold values of hidden layer neuron is expressed as bj, hidden layer to output layer it
Between weight be expressed as vjk, the threshold values of output layer neuron is expressed as ak.The parameter to be adjusted in BP neural network is weight wij、
vjkWith threshold values bj、ak, the parameter that these to be adjusted is set as Artificial Fish state, then Artificial Fish x can be expressed as a M*J+J+
J*K+K dimensional vector：
X=(w11,…,wM1,b1,…,wiJ,…,wMJ,bJ,v11,…,vJ1,a1…,v1k,…,vJK,aK), wherein w11,…,wM1
Weight of the expression input layer to first hidden layer neuron, the threshold values of b1 expression first neuron of hidden layer,
wiJ,…wMJIndicate weight of the input layer to j-th hidden layer neuron, the valve of bJ expression hidden layer j-th neuron
Value, v11,…,vJ1Indicate weight of the hidden layer neuron to first output layer neuron, a1 expression first nerve of output layer
The threshold values of member, v1k,…,vJKIndicate weight of the hidden layer neuron to k-th output layer neuron, aKIndicate output layer k-th
The threshold values of neuron, the food concentration FC of Artificial Fish is set as the inverse of BP neural network overall error E, i.e. FC=1/E, artificial in this way
The maximum point of the food concentration to be found of fish is exactly the smallest point of BP neural network error, any two Artificial Fish x1、x2Between
Euclidean distance d be expressed as：
X in formula1、x2In element denounce to correspond according to dimension and subtract each other, Artificial Fish x behavior and is chased after executing foraging behavior, bunch
The state of itself will be changed after tail behavior, while carrying out the primary adjustment of initial weight and threshold values to BP neural network；
S43, it is implied according to the characteristics of BP neural network error function and design experiences of forefathers, Binding experiment, selection centre
The neuron of layer is 20, and sample input is the data of 17 dimensions, and network output is 1 dimension, i.e. the structure setting of neural network is 17-
20-1；
It is 10 that Artificial Fish number is selected in S44, Artificial Fish optimization algorithm, visible range 0.5, crowding factor 0.618, step-length choosing
It is selected as 0.1, the number of iterations 50 repeats to explore number to be 50；BP neural network activation primitive 1 selects tansig function, activation
Function 2 selects logsig function, and training function uses traindx function, and maximum frequency of training is set as 20000 times, learning rate
It is 0.05, error requirements precision is 10-4, factor of momentum 0.9；
S45, to binary functionIt is fitted to test artificial fish school optimization BP net
The performance of network：Taking step-length is 0.1, altogether 40000 data, wherein 35000 are done training sample, 5000 do test sample, right
Data carry out 100 training；
S46,2100 groups of data are chosen from eeg data library, 600 groups of collected state data, 600 groups of waking states, 600 groups tired
Labor state, this 1800 groups are used for network training, remaining 300 groups are used for network test, and grouping is carried out by the way of intersecting.It utilizes
The neural network of above-mentioned steps training test identifies the state of eeg signal, realizes the prison to personnel in the pit's fatigue strength
It surveys；
The concentration data that S5, master controller handle various sensors judges the state that transfinites of gas, merges out current safe shape
State and danger classes, while the interface that monitoring host computer is shown is sent to by the existing communication node in underground, and pass through bluetooth
The Intelligent bracelet being worn in miner's wrist that module and master controller connect to the safe condition carry out sound of different danger classes,
Light, the triple alarms for shaking three kinds of modes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810715885.1A CN108903936B (en) | 2018-07-03 | 2018-07-03 | Intelligent mining helmet control method based on human body information and environment information fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810715885.1A CN108903936B (en) | 2018-07-03 | 2018-07-03 | Intelligent mining helmet control method based on human body information and environment information fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108903936A true CN108903936A (en) | 2018-11-30 |
| CN108903936B CN108903936B (en) | 2021-01-26 |
Family
ID=64425296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810715885.1A Active CN108903936B (en) | 2018-07-03 | 2018-07-03 | Intelligent mining helmet control method based on human body information and environment information fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108903936B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614932A (en) * | 2018-12-12 | 2019-04-12 | 西安科技大学 | Identify Environment, the mining helmet and cloud platform based on brain electric field change mechanism |
| CN109784361A (en) * | 2018-12-05 | 2019-05-21 | 鲁东大学 | Seashell products classifying identification method and device |
| CN109884465A (en) * | 2019-03-01 | 2019-06-14 | 辽宁工业大学 | A kind of one-way earth fault localization method based on signal injection method |
| CN110840452A (en) * | 2019-12-10 | 2020-02-28 | 广西师范大学 | Filtering device and method for brain wave signals |
| CN112336355A (en) * | 2020-11-06 | 2021-02-09 | 广东电网有限责任公司电力科学研究院 | Safety supervision system, device and equipment based on electroencephalogram signal operating personnel |
| CN112842359A (en) * | 2021-01-25 | 2021-05-28 | 国网江苏省电力有限公司电力科学研究院 | Pressure and fatigue information monitoring method for intelligent safety helmet |
| FR3103373A1 (en) * | 2019-11-27 | 2021-05-28 | Thales | ENVIRONMENTAL MONITORING ASSISTANCE SYSTEM AND ASSOCIATED PROCESS |
| CN113017184A (en) * | 2021-03-29 | 2021-06-25 | 广东电网有限责任公司 | Intelligent safety helmet and intelligent monitoring system |
| CN113506246A (en) * | 2021-06-15 | 2021-10-15 | 西安建筑科技大学 | Concrete 3D printing component fine detection method based on machine vision |
| CN114983080A (en) * | 2022-05-23 | 2022-09-02 | 北京市政建设集团有限责任公司 | Intelligent safety helmet |
| CN117649197A (en) * | 2023-11-29 | 2024-03-05 | 江苏猎人安防科技有限公司 | Visual intelligent monitoring data security control substation system and process thereof |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030055355A1 (en) * | 2000-10-16 | 2003-03-20 | Viertio-Oja Hanna Elina | Method and apparatus for determining the cerebral state of a patient with fast response |
| CN103989485A (en) * | 2014-05-07 | 2014-08-20 | 朱晓斐 | Human body fatigue evaluation method based on brain waves |
| WO2014167460A1 (en) * | 2013-04-12 | 2014-10-16 | Commissariat A L'energie Atomique Et Aux Energies Alternatives | Methods to monitor consciousness |
| US20160361041A1 (en) * | 2015-06-15 | 2016-12-15 | The Research Foundation For The State University Of New York | System and method for infrasonic cardiac monitoring |
| CN106388780A (en) * | 2016-09-21 | 2017-02-15 | 广州视源电子科技股份有限公司 | Sleep state detection method and system based on fusion of two classifiers and detector |
| CN108229401A (en) * | 2018-01-05 | 2018-06-29 | 杭州电子科技大学 | A kind of multi-modal Modulation recognition method based on AFSA-SVM |
-
2018
- 2018-07-03 CN CN201810715885.1A patent/CN108903936B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030055355A1 (en) * | 2000-10-16 | 2003-03-20 | Viertio-Oja Hanna Elina | Method and apparatus for determining the cerebral state of a patient with fast response |
| WO2014167460A1 (en) * | 2013-04-12 | 2014-10-16 | Commissariat A L'energie Atomique Et Aux Energies Alternatives | Methods to monitor consciousness |
| CN103989485A (en) * | 2014-05-07 | 2014-08-20 | 朱晓斐 | Human body fatigue evaluation method based on brain waves |
| US20160361041A1 (en) * | 2015-06-15 | 2016-12-15 | The Research Foundation For The State University Of New York | System and method for infrasonic cardiac monitoring |
| CN106388780A (en) * | 2016-09-21 | 2017-02-15 | 广州视源电子科技股份有限公司 | Sleep state detection method and system based on fusion of two classifiers and detector |
| CN108229401A (en) * | 2018-01-05 | 2018-06-29 | 杭州电子科技大学 | A kind of multi-modal Modulation recognition method based on AFSA-SVM |
Non-Patent Citations (4)
| Title |
|---|
| MEI WANG等: "Anxiety Level Detection Using BCI of Miner’s Smart Helmet", 《MOBILE NETW APPL》 * |
| MU ZHENDONG 等: "Mobile Healthcare System for Driver Based on Drowsy Detection Using EEG Signal Analysis", 《METALLURGICAL AND MINING INDUSTRY》 * |
| 江朝晖 等: "脑电近似熵分析的思维分类识别", 《应用科学学报》 * |
| 王湃 等: "一种基于智能手机与脑－ 机接口技术的环境控制系统", 《西安科技大学学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784361A (en) * | 2018-12-05 | 2019-05-21 | 鲁东大学 | Seashell products classifying identification method and device |
| CN109614932A (en) * | 2018-12-12 | 2019-04-12 | 西安科技大学 | Identify Environment, the mining helmet and cloud platform based on brain electric field change mechanism |
| CN109884465A (en) * | 2019-03-01 | 2019-06-14 | 辽宁工业大学 | A kind of one-way earth fault localization method based on signal injection method |
| CN109884465B (en) * | 2019-03-01 | 2023-09-29 | 辽宁工业大学 | Unidirectional ground fault positioning method based on signal injection method |
| FR3103373A1 (en) * | 2019-11-27 | 2021-05-28 | Thales | ENVIRONMENTAL MONITORING ASSISTANCE SYSTEM AND ASSOCIATED PROCESS |
| WO2021105351A1 (en) * | 2019-11-27 | 2021-06-03 | Thales | System for assisting in the monitoring of an environment and associated method |
| CN110840452A (en) * | 2019-12-10 | 2020-02-28 | 广西师范大学 | Filtering device and method for brain wave signals |
| CN112336355A (en) * | 2020-11-06 | 2021-02-09 | 广东电网有限责任公司电力科学研究院 | Safety supervision system, device and equipment based on electroencephalogram signal operating personnel |
| CN112842359B (en) * | 2021-01-25 | 2022-09-16 | 国网江苏省电力有限公司电力科学研究院 | Pressure and fatigue information monitoring method for intelligent safety helmet |
| CN112842359A (en) * | 2021-01-25 | 2021-05-28 | 国网江苏省电力有限公司电力科学研究院 | Pressure and fatigue information monitoring method for intelligent safety helmet |
| CN113017184A (en) * | 2021-03-29 | 2021-06-25 | 广东电网有限责任公司 | Intelligent safety helmet and intelligent monitoring system |
| CN113506246B (en) * | 2021-06-15 | 2022-11-25 | 西安建筑科技大学 | Concrete 3D printing component fine detection method based on machine vision |
| CN113506246A (en) * | 2021-06-15 | 2021-10-15 | 西安建筑科技大学 | Concrete 3D printing component fine detection method based on machine vision |
| CN114983080A (en) * | 2022-05-23 | 2022-09-02 | 北京市政建设集团有限责任公司 | Intelligent safety helmet |
| CN117649197A (en) * | 2023-11-29 | 2024-03-05 | 江苏猎人安防科技有限公司 | Visual intelligent monitoring data security control substation system and process thereof |
| CN117649197B (en) * | 2023-11-29 | 2024-08-20 | 北京希安科电子系统工程有限责任公司 | Visual intelligent monitoring data security control substation system and process thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108903936B (en) | 2021-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108903936A (en) | The Intelligent mining helmet control method merged based on human body information and environmental information | |
| CN110133610B (en) | Ultra-wideband radar action identification method based on time-varying distance-Doppler diagram | |
| CN109620211A (en) | A kind of intelligent abnormal electrocardiogram aided diagnosis method based on deep learning | |
| West et al. | Complex webs: anticipating the improbable | |
| CN106846729A (en) | A kind of fall detection method and system based on convolutional neural networks | |
| CN107822623A (en) | A kind of driver fatigue and Expression and Action method based on multi-source physiologic information | |
| CN111053549A (en) | Intelligent biological signal abnormality detection method and system | |
| CN110367967B (en) | Portable lightweight human brain state detection method based on data fusion | |
| CN105877766A (en) | Mental state detection system and method based on multiple physiological signal fusion | |
| CN109498041A (en) | Driver road anger state identification method based on brain electricity and pulse information | |
| CN110502988A (en) | Group positioning and anomaly detection method in video | |
| CN106919956A (en) | Brain wave age forecasting system based on random forest | |
| CN105809124A (en) | DWT- and Parametric t-SNE-based characteristic extracting method of motor imagery EEG(Electroencephalogram) signals | |
| CN103989485A (en) | Human body fatigue evaluation method based on brain waves | |
| CN108523907B (en) | Fatigue state identification method and system based on deep shrinkage sparse self-coding network | |
| CN106725376B (en) | Sign detection method and device | |
| CN114999237B (en) | Intelligent education interactive teaching method | |
| CN110363090A (en) | Intelligent heart disease detection method, device and computer readable storage medium | |
| Xu et al. | Intelligent emotion detection method based on deep learning in medical and health data | |
| CN107468260A (en) | A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state | |
| CN107714037A (en) | A kind of miner's fatigue identification method based on the mining helmet of brain-computer interface | |
| An et al. | Electroencephalogram emotion recognition based on 3D feature fusion and convolutional autoencoder | |
| CN107174209A (en) | Sleep stage based on nonlinear kinetics method by stages | |
| CN106236027A (en) | Depressed crowd's decision method that a kind of brain electricity combines with temperature | |
| CN109875584A (en) | The detection method and its warning system of physiological driver's fatigue |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |