CN107744436A - A kind of wheelchair control method and control system based on the processing of neck muscle signals - Google Patents
A kind of wheelchair control method and control system based on the processing of neck muscle signals Download PDFInfo
- Publication number
- CN107744436A CN107744436A CN201710958725.5A CN201710958725A CN107744436A CN 107744436 A CN107744436 A CN 107744436A CN 201710958725 A CN201710958725 A CN 201710958725A CN 107744436 A CN107744436 A CN 107744436A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- wheelchair
- muscle signals
- signal
- 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.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61G—TRANSPORT, PERSONAL CONVEYANCES, OR ACCOMMODATION SPECIALLY ADAPTED FOR PATIENTS OR DISABLED PERSONS; OPERATING TABLES OR CHAIRS; CHAIRS FOR DENTISTRY; FUNERAL DEVICES
- A61G5/00—Chairs or personal conveyances specially adapted for patients or disabled persons, e.g. wheelchairs
- A61G5/04—Chairs or personal conveyances specially adapted for patients or disabled persons, e.g. wheelchairs motor-driven
- A61G5/041—Chairs or personal conveyances specially adapted for patients or disabled persons, e.g. wheelchairs motor-driven having a specific drive-type
- A61G5/046—Chairs or personal conveyances specially adapted for patients or disabled persons, e.g. wheelchairs motor-driven having a specific drive-type at least three driven wheels
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61G—TRANSPORT, PERSONAL CONVEYANCES, OR ACCOMMODATION SPECIALLY ADAPTED FOR PATIENTS OR DISABLED PERSONS; OPERATING TABLES OR CHAIRS; CHAIRS FOR DENTISTRY; FUNERAL DEVICES
- A61G5/00—Chairs or personal conveyances specially adapted for patients or disabled persons, e.g. wheelchairs
- A61G5/10—Parts, details or accessories
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61G—TRANSPORT, PERSONAL CONVEYANCES, OR ACCOMMODATION SPECIALLY ADAPTED FOR PATIENTS OR DISABLED PERSONS; OPERATING TABLES OR CHAIRS; CHAIRS FOR DENTISTRY; FUNERAL DEVICES
- A61G2200/00—Information related to the kind of patient or his position
- A61G2200/10—Type of patient
- A61G2200/20—Type of patient with asymmetric abilities, e.g. hemiplegic or missing a limb
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61G—TRANSPORT, PERSONAL CONVEYANCES, OR ACCOMMODATION SPECIALLY ADAPTED FOR PATIENTS OR DISABLED PERSONS; OPERATING TABLES OR CHAIRS; CHAIRS FOR DENTISTRY; FUNERAL DEVICES
- A61G2203/00—General characteristics of devices
- A61G2203/10—General characteristics of devices characterised by specific control means, e.g. for adjustment or steering
- A61G2203/18—General characteristics of devices characterised by specific control means, e.g. for adjustment or steering by patient's head, eyes, facial muscles or voice
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Electrotherapy Devices (AREA)
Abstract
The present invention relates to a kind of wheelchair control method based on the processing of neck muscle signals and control system, control method to comprise the following steps：Neck muscle signals MMG is gathered by sensor, and inputted to training pattern；Signal Pretreatment, segmentation, feature extraction, dimension-reduction treatment and headwork command signal Classification and Identification are carried out to neck muscle signals MMG using training pattern；The headwork command signal that will identify that reaches wheelchair controller, controling wheelchair operation.Feature extraction includes wavelet package transforms coefficient power feature extraction and main diagonal slices feature extraction.Command signal includes six kinds of control models：Respectively forward, to the left, to the right, backward, accelerate and halt instruction.Compared with prior art, the present invention has the advantages that MMG sensors are easy to use, signal identification is accurate, more wheelchair control patterns, sensor collection signal are accurate and applied widely.
Description
Technical field
The present invention relates to the application of muscle signals, and neck muscle signals are based on based on neck flesh one kind more particularly, to one kind
The wheelchair control method and control system of wheelchair control method and control system the sound signal processing of processing.
Background technology
With expanding economy and the raising of social civilization level, the living condition and quality of life of the elderly and the disabled
Increasingly paid close attention to by society.Electric wheelchair (EPW) has been used for the quality of life for improving the elderly and the disabled.So
And these wheelchairs are typically to be controlled by the hand of user by control stick, which prevent those the residual of handicap even amputation
Disease people and the elderly for difficulty of starting remove the wheelchair manipulated using this hand.Although many disabled persons and old man can not be freely
Using both hands, but they can move freely through head.If the idea of these people can be obtained by headwork and is intended to
Can be that EPW and other rehabilitation auxiliary implements provide effective control signal source.
As society constantly progressive biology sensor and robot technology, many researchers are studying advanced skill
Art, such as by voice, the posture of gesture control, electroculogram (EOG), electromyogram (EMG) and electroencephalogram (EEG) etc. come control wheel
Chair, but these technologies have merits and demerits.For advantage, they are all ripe technologies, in actual life
It is applied, but they also there are many shortcomings.For the image technique based on gesture, background color, change illumination condition, sunlight
All complexity can be brought with shade etc. to gesture identification；There are some habitually gestures additionally, due to everyone, it is especially needed
Distinguish intentional or unintentional gestures detection problem.Voice command needs a single, quiet workplace to be not suitable for noise
Miscellaneous environment, because noise can cause maloperation.The poor real of brain electric control, EEG signals are extremely sensitive, or even to physiology
It is all sensitive that source includes the illusion of motion, muscle noise interference, the illusion of eye movement or blink, heartbeat.Other EEG signals
Low signal-to-noise ratio and signal mode situation about being lack of consistency also to use EEG signals controling wheelchair imperfection.Brain electric control
Using also requiring that user association effectively adjusts brain wave, this is complicated for user and has pressure, so EEG controls are logical
Often just consider in the case of other method is non-serviceable.Controlled for EOG, because the eyes of people can blink in spite of oneself
Eye, can cause maloperation, so unsafe.Greatest problem based on EOG is that system is believed dependent on the eyes of user as input
Number, so user freely can not be arbitrarily seen when travelling wheelchair to the situation that Anywhere, otherwise can cause danger.Another point
That, when user is with eyes control electric wheelchair for a long time, user can be easier to feel fatigue because eyes be more easy to it is tired
Labor.Electromyogram is also used to control EPWs, but subject matter is perspiration, electrode is worn off, electrode touches skin short circuit and makes electricity
Pole Skin Resistance change, so as to cause signal intensity to be a big problem, inevitably interference can cause mistake for these, make control
Penalty processed.
The content of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind is based on neck flesh sound
The wheelchair control method and control system of signal transacting.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of wheelchair control method based on the processing of neck muscle signals, this method comprise the following steps：
S1, by sensor gather neck muscle signals MMG, the input signal as training pattern；
S2, using training pattern to neck muscle signals MMG carry out Signal Pretreatment, segmentation, feature extraction, dimension-reduction treatment
With headwork command signal Classification and Identification；
S3, the headwork command signal that will identify that reach wheelchair controller, controling wheelchair operation.
Preferably, described feature extraction includes wavelet package transforms coefficient power feature extraction and main diagonal slices feature carries
Take.
Preferably, described main diagonal slices are characterized as w in bispectrum signal1=w2When characteristic signal, described bispectrum letter
Number formula is：
Wherein Bx(w1,w2) it is bispectrum, C3x(τ1,τ2) be random process Third-order cumulants, τ1And τ2It is lag time,
And | w1|≤π,|w2|≤π,|w1+w2|≤π, w1、w2Respectively frequency variable, x represent certain sample.
Preferably, described headwork command signal includes six kinds of control models：Respectively forward, to the left, to the right, to
Afterwards, acceleration and halt instruction.
Preferably, neck muscle signals MMG when human body head acts is gathered by sensor in step S1, as training
The input signal of model.
Preferably, the muscle signals MMG at four positions of neck, described four positions are gathered in step S1 by sensor
Musculi splenius capitis including arranged on left and right sides nutator and arranged on left and right sides.
Preferably, described Signal Pretreatment includes filtering and normalization.
Preferably, described sensor is TD-3 acceleration transducers.
A kind of wheelchair control of the wheelchair control method based on the processing of neck muscle signals using described in any one as described above
System processed, sensor, computer, wireless communication module, controller and the motor that described control system includes being sequentially connected drive
Dynamic model block.
Preferably, described sensor is TD-3 acceleration transducers, and described controller is STM32 controllers.
Compared with prior art, the present invention has advantages below：
(1) MMG sensors are easy to use：MMG makes it place sensor due to its propagation characteristic in musculature
It is insensitive in the position of skin surface, so the position of MMG sensors need not place very accurate, it is fixed on subject's
It is more convenient on body.
(2) signal identification is accurate：Identified by headwork, reduce the misrecognition under malfunction in the past, MMG is a kind of machine
Tool signal, strong antijamming capability, high s/n ratio, be pattern-recognition an excellent signal source.
(3) more wheelchair control patterns：The general wheelchair control of signal only need to realize four kinds or five kinds of control models (to
Before, to the left, to the right, backward and stop), but this research realize six kinds of control models (forward, to the left, to the right, backward,
Accelerate, stop), based on the fact that, if we only pick out five kinds of patterns from six kinds of patterns comes controling wheelchair, pattern choosing
It is wide to select scope；Other six kinds of head movement patterns can also be used for need more multi-control modes other equipment in, as game paddle,
Manipulator etc..
(4) sensor collection signal is accurate：The frequency range of MMG signals is 0-100hz, and main signal energy is low
In the range of frequency (2-50Hz), TD-3 acceleration transducers meet frequency response characteristic.
(5) it is applied widely：The headwork pattern-recognition controling wheelchair scope of application based on neck muscle signals is wider,
For the difficult the elderly of those severe paralysis, the disabled person of four limbs amputation and quadruped locomotion, still can something or somebody to fall back on portions move
Make convenient operation wheelchair.
Brief description of the drawings
Fig. 1 is neck muscle signals (MMG) controling wheelchair flow chart；
Fig. 2 is the segmentation figure of the nodding action of the neck muscle signals (MMG) collected；
Fig. 3 is the segmentation figure of the new line action of the neck muscle signals (MMG) collected；
Fig. 4 is the segmentation figure of the left swing action of the neck muscle signals (MMG) collected；
Fig. 5 is the segmentation figure that the right swing of the neck muscle signals (MMG) collected is made；
Fig. 6 is the segmentation figure of the left-hand rotation action of the neck muscle signals (MMG) collected；
Fig. 7 is the segmentation figure of the right-hand rotation action of the neck muscle signals (MMG) collected；
Fig. 8 is the main diagonal slices figure of one-dimensional bispectrum feature of neck muscle signals (MMG) when nodding；
Fig. 9 is the main diagonal slices figure of one-dimensional bispectrum feature of neck muscle signals (MMG) when coming back；
The one-dimensional bispectrum feature of neck muscle signals (MMG) main diagonal slices figure when Figure 10 is left swing；
The one-dimensional bispectrum feature of neck muscle signals (MMG) main diagonal slices figure when Figure 11 is right pendulum；
Figure 12 is the main diagonal slices figure of one-dimensional bispectrum feature of neck muscle signals (MMG) when turning left；
Figure 13 is the main diagonal slices figure of one-dimensional bispectrum feature of neck muscle signals (MMG) when turning right；
Figure 14 is the characteristic profile after FLDA dimensionality reductions；
Figure 15 is the signal sequence diagram of muscle signals (MMG) controling wheelchair；
Figure 16 is the schematic layout pattern on wheelchair chassis.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is the part of the embodiment of the present invention, rather than whole embodiments.Based on this hair
Embodiment in bright, the every other reality that those of ordinary skill in the art are obtained on the premise of creative work is not made
Example is applied, should all belong to the scope of protection of the invention.
Embodiment
The present invention proposes a kind of wheelchair control method and control system based on the processing of neck muscle signals.Muscle signals
(Mechanomyography, abbreviation MMG) is Low-Frequency Mechanical signal, is detected in contraction of muscle.It is received by muscle
Transverse movement during contracting, under the resonant frequency of muscle caused by less swaying and the change in size of movable muscle fibre.
It can be measured by the laser range sensor of microphone, electric contact sensor, accelerometer or skin surface.MMG has
There is a series of advantage to include the flexibility that sensor is placed, the robustness and cost of Skin Resistance change are low etc., MMG by with
To study muscular states and pattern-recognition.In recent years, it has been studied as the control signal of switch and electric arm, because
This, if MMG can be used for the different headwork pattern of reliable controling wheelchair and can identify exactly.So the present invention is first
The pattern-recognition based on MMG head movements is first have studied, is then moved with this pattern-recognition come controling wheelchair.
First 10 subjects are acquired with the MMG signals of 6 kinds of headworks, the acquisition order of this 6 kinds actions in the present invention
To bow, coming back, left swing, right pendulum, left-hand rotation, finally turn right.Every kind of action repeats 90 times, every time time interval 3s between action.
In order to avoid the influence of MMG signals when muscular fatigue is to action, the time of having a rest between two kinds of different actions is about 0.5 hour.
After doing subsequent treatment by the MMG signals collected from this 10 subjects and train and obtain a model, collection is another
The MMG signals of the headwork of outer 9 subjects, are sent into the mould trained before after the MMG signals collected are handled
Headwork identification is carried out in type, then will identify that the output control wheel chair sport come, headwork acts corresponding with wheelchair
Relation is as follows：Action control wheelchair of bowing, which advances, to be accelerated to a setting speed and keeps；New line action control wheelchair is fallen back；It is left
Rotation is turned left 10 ° as controling wheelchair；Right-hand rotation action control wheelchair is turned right 10 °；Left swing action control wheelchair is slowly decelerated to
Stop；Controling wheelchair acceleration is made in right swing.
The MMG signal acquisitions of headwork are to gather MMG signals to the musculi colli at four positions, and this four positions are
Arranged on left and right sides nutator (SternocleidomastoidMuscle, SCM) and the musculi splenius capitis of the left and right sides
(Splinius, SPL), sensor is sticked with glue at the surface skin at this four positions, the frequency range of MMG signals is 0-
100hz, and main signal energy (2-50Hz) in low-frequency range, the TD-3 acceleration transducers of Beijing Yi Song companies meet
Frequency response characteristic, therefore sensor elects TD-3 as.The output signal of sensor is by capture card (National Instruments
NI 9205) collection, sample frequency 1000Hz.The signal collected is recorded and handled by MATLAB R2012a.Whole mistake
Journey is as shown in Figure 1.
The pretreatment and segmentation of signal：
Signal Pretreatment includes the filtering and normalization of primary signal.Primary signal first passes through 2-49hz Kaiser windows filter
Ripple device filters.After filtering, signal is normalized according to formula (1).
Wherein x is the magnitude of voltage of acquired signal, and M is the average voltage of same channel, and S is standard deviation.
It is that signal is split to separate each action in next step.Used in this research a kind of based on twice-enveloping curve
Non-uniformly distributed load method.Each, which is acted, a corresponding twice-enveloping line, two smallest points difference before and after envelope
It is the beginning and end position of action.Result after segmentation is shown in Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6 and Fig. 7.
Feature extraction：
By pre-processing and after dividing processing, obtaining the segmentation waveform per individual part in four passages, then believe from segmentation
The leading diagonal slice feature in wavelet package transforms coefficient energy feature and bispectrum feature is extracted in number.Wavelet packet coefficient becomes transducing
The extraction of measure feature following formula (2), (3) represent.
F={ log10(Ei), i=1 ... 2L} (2)
Wherein F representative features, L are Decomposition orders,Be i-th layer of WAVELET PACKET DECOMPOSITION signal subspace band in j-th of coefficient,
NiIt is CiThe length of signal band, EiIt is CiWavelet package transforms coefficient the average energy value of signal band.
MMG signals carry out four layers of decomposition of wavelet packet, therefore, the wavelet package transforms coefficient of each passage using Coif4 small echos
Energy feature is 24=16 dimensions.Because one shares four passages, the energy feature one of wavelet package transforms coefficient shares 16x4
=64 dimensions.
In addition to wavelet package transforms coefficient energy feature, bispectrum feature is also extracted herein.At traditional random signal
In reason, it is to have Gaussian stationary signal that can typically assume signal, but muscle signals are not gaussian signals, nor steadily
's.In consideration of it, herein need to be by the method for high-order statistic.
Three rank spectrums (also known as bispectrum) in higher-order spectrum, exponent number is minimum, and processing method is also most simple, therefore will use herein double
Further feature extraction is done in spectrum analysis.
Higher-order spectrum is defined by accumulation flow function, thus alternatively referred to as cumulant spectrum.{ if x (t) } is the k ranks of zero-mean
Stationary process, then the k rank cumulants of the process be defined as formula (4).
Ckx(τ1,τ2,…,τk-1)=cum { x (t), x (t+ τ1),…,x(t+τk-1)} (4)
Wherein Ckx(τ1,τ2,…,τk-1) it is k rank cumulants, τ1,…,τk-1It is lag time.
Then k ranks cumulant spectrum is defined as the k-1 dimension Fourier transformations of k rank cumulants, such as formula (5).
Wherein Skx(w1,w2,…,wk-1) it is k rank cumulant spectrums, Ckx(τ1,τ2,…,τk-1) it is k rank cumulants, w1,
w2,…,wk-1It is frequency, τ1,τ2,…,τk-1It is lag time.
Power spectrum, bispectrum and three spectrums are all the special circumstances of k rank cumulant spectrums, and as k=2, second-order cumulant spectrum is
Power spectrum, as k=3, as bispectrum, as k=4, as three spectrums.So bispectrum is defined as follows formula (6).Bispectrum is complex value
Spectrum, there is 2 frequency variable w1And w2, bispectrum is in w1And w212 symmetrical regions are shared in the frequency plane of composition.
C3x(τ1,τ2)=E [x (t) x (t+ τ1)x(t+τ2)] (6)
Wherein Bx(w1,w2) it is bispectrum, S3x(w1,w2) it is that Third-order cumulants are composed, C3x(τ1,τ2) it is random process { x (t) }
Third-order cumulants, τ1And τ2All it is lag time, and | w1|≤π,|w2|≤π,|w1+w2|≤π, w1、w2For two frequency values.
Bispectrum is two-dimensional function, if taking all bispectrum information also to need the template matches for solving two dimension to ask as feature
Topic, amount of calculation is also very huge in addition, can have an impact to real-time, so the one-dimensional characteristic of our selected bispectras, i.e. leading diagonal
Another big feature of the slice information as signal.If the bispectrum of random process { x (t) } is Bx(w1,w2), then main cutting on the cross
The feature extraction of piece is exactly to calculate w1=w2B during=wxFig. 8, Fig. 9, Figure 10, figure are shown in (w, w), the feature extraction of main diagonal slices
11st, Figure 12 and Figure 13.
Dimension-reduction treatment：
After the wavelet package transforms coefficient energy feature obtained using previous step feature extraction, the intrinsic dimensionality of wavelet packet is found
(64 dimension) are excessive, it is necessary to carry out dimensionality reduction, and the method that dimensionality reduction uses is FLDA, after dimensionality reduction with the distribution of the scatter diagram of each action come
Image shows dimensionality reduction effect.Now introduce the principle of FLDA dimensionality reductions：FLDA (Fisher linear discriminant analysis) is to be used to distinguish difference
The most efficient method of species sample.Scatter matrix is formula (7) in class：
Scatter matrix is formula (8) between class：
The species number of all samples of wherein NC, it is 6 in this research, each subclassification ciIn include niIndividual sample value.miIt is
ciThe average value of subclassification, m are the averages of all samples, and x is each sample value.By scatter matrix table between scatter matrix and class in class
The Fisher criterions function shown is formula (9)：
Wherein SBIt is scatter matrix between class, SwIt is stroll matrix in class, P is projection matrix.
The purpose of Fisher discriminant analyses is to find the P for maximizing canonical function J (P).Derivation sends as an envoy to J (P) maximumlly
Column vector P is the characteristic vector of following characteristic equation (10).
SBPi=λiSWPi (10)
Wherein λiCorrespond to ith feature vector PiCharacteristic value.
When carrying out FLDA dimensionality reductions, the calculation formula of projection properties is shown in formula (11).
yf=PTxf (11)
Wherein xfIt is original eigenmatrix, yfIt is by the eigenmatrix after FLDA dimensionality reductions.
In view of SBRank of matrix is not more than NC-1, and the number of the nonzero eigenvalue of equation (10) is less than or equal to NC-1.
In this research, NC is 6, then the dimension of the characteristic vector after FLDA dimension-reduction treatment is 6-1=5.Knot after dimension-reduction treatment
Fruit sees Figure 14.
SVM (SVMs)：
It is exactly finally the spy after processing after the pretreatment of several steps above, segmentation, feature extraction, Feature Dimension Reduction
Space input SVM training patterns are levied, go to classify to test sample again by the model trained.During this investigation it turned out, sample
Amount is considerably beyond intrinsic dimensionality, and this often leads to have between two kinds of samples more overlapping, therefore non-linear SVM compares Linear SVM
It is more suitable for.Firstly the need of using a kind of reflecting for linear problem that nonlinear problem by lower dimensional space is converted into higher dimensional space
Shooting method, and this process needs a kernel function.The kernel function that this research uses is RBF (RBF), such as formula
(10)。
Wherein K (z, xc) is RBF, and xc is the center of function, and σ is the width parameter of function, plays control radially
The function of distance, z are the points in former lower dimensional space.
Wheelchair control：
After algorithm in Matlab successfully identifies headwork, you can send corresponding actions by Matlab serial interfaces
Code to 433 wireless communication modules, 433 wireless communication modules the 433 of wheelchair are sent data to by wireless telecommunication
Wireless communication module, then sent by uart serial communications to STM32 controllers, STM32 controllers are according to the action received
Code name carries out different control to wheelchair by motor drive module.Realize the motion control of wheelchair all around.Whole stream
Journey block diagram is as shown in figure 15.
The operation principle of wheelchair is as follows：Wheelchair is independently driven using four wheels.Each wheel has corresponding motor.
When receive nodding action signal need advance when, four wheels all rotate forward wheelchairs accelerate advance to setting speed；Work as reception
When needing to retreat to new line action signal, four wheels all invert wheelchair rollback；A left side is needed when receiving left-hand rotation action signal
When turning, left side wheel reversion, the right wheel rotates forward, and wheelchair turns left；When receive right-hand rotation action signal need turn right when, the left side wheel
Son is rotated forward, and the right wheel reversion, wheelchair is turned right；When receive left swing action signal need stop when, the left and right wheel of wheelchair is slow
Slowly it is decelerated to stopping；When receiving right pendulum action signal and needing to accelerate, the wheel of wheelchair accelerates rotation.
Fig. 1 show collection neck muscle signals (MMG), the whole flow process of identification headwork control experiment wheelchair operation
Figure, and the signal processing for combining Fig. 2 to Figure 14 does the explanation of beforehand control signal transacting：
The collection of neck muscle signals (MMG) when 10 volunteers of invitation do headwork first, harvester is the U.S.
The analog input card NI-9205 of National Instruments (NI), according to nod, come back, left swing, right pendulum, the headwork order turned left, turned right
6 kinds of actions are done, under every kind of action does 90, the time interval between every twice is 3s, and the time interval between every kind of action is half
Hour, therefore can collect acted under 90x6x10=5400 altogether；And 4 TD-3 sensors are needed altogether when acting collection
It is placed on the left and right nutator (SCM) of subject and the position of left and right musculi splenius capitis (SPL), i.e., port number is 4.
The signal collected by analog input card is stored in Matlab R2012a, and carries out Signal Pretreatment with this software,
Filter, normalize, Length discrepancy cutting, the one action design sketch after cutting is shown in Fig. 2 to Fig. 7, it can be seen that this six kinds actions
Cutting is complete display.There is obvious difference between difference action.The normalized amplitude of left swing and right pendulum be respectively less than other four
Individual action, the fluctuating range of right-hand rotation is maximum, and almost from -9 to 8, and the curve burr bowed is more, and the curve of new line is most smooth.
Feature extraction is carried out after pretreatment, time and frequency domain characteristics (wavelet package transforms coefficient average energy is extracted in this research
Measure feature) and frequency domain character (the main diagonal slices feature of bispectrum), see Fig. 8 to figure for the schematic diagram of the main diagonal slices feature of bispectrum
13, it is seen that the main diagonal slices of six kinds of actions have obvious difference：Other four actions of the Amplitude Ration of left swing and right pendulum are more dispersed,
Highest energy concentrates on center section, bows, comes back, turning left, the energy for four actions of turning right also concentrates on centre.For low
Head action, there is secondary lobe the both sides of slice map.In addition, for bow and new line act, spectrum size is identical, all close to
1.5x10-3.Energy intensity in the slice map of left-hand rotation and right-hand rotation action is very high, and both sides clutter does not almost have, the number of spectral peak
Magnitude is also consistent, is all 4x10-3。
Because the dimension (64) of wavelet package transforms coefficient average energy feature is excessive after feature extraction, so needing to this
Feature carries out FLDA dimension-reduction treatment, and the result after handling is shown in Figure 14, it can be clearly seen that the feature distribution feelings after FLDA dimensionality reductions
Condition is good, and similar sample is intensive, and inhomogeneity sample disperses, and is advantageous to identify.
The feature for the everything handled well is sent into SVM classifier afterwards, chooses the 50 of the every kind of action of each subject
Group feature is as training sample, and for remaining 40 groups of features as test sample, last recognition result is 95.92%.
Model is characterized as with the everything of this 10 subjects, then gathers neck when other 9 subject's heads act
Portion's muscle signals (MMG), this 9 subjects can be acted with random order, and each action is repeated 10 times, then by identical
Action signal processing step --- pretreatment, feature extraction, Feature Dimension Reduction, it is sent into the model trained, identifies head
After action, Figure 15 signal transduction process is carried out, corresponding 1 instruction of such as bowing, come back corresponding 2 instruction, corresponding 3 instruction of left swing, right
Corresponding 4 instruction of pendulum, corresponding 5 instruction of turning left, corresponding 6 instruction of turning right, corresponding actions are sent by Matlab serial interfaces by computer
Instruction to 433 wireless communication modules, this 433 wireless communication module wheelchair end is sent instructions to by wireless telecommunication
433 wireless communication modules, then sent by uart serial communications to STM32 controllers, STM32 controllers are dynamic according to what is received
Make code name by motor drive module to carry out different control to wheelchair, realize the motion control of wheelchair all around.This
The success rate that controls of process is 85.74%.
Figure 16 show wheelchair chassis schematic diagram in the present invention, identifies headwork, after being instructed accordingly, mainly
By the motor on 433 wireless communication modules, STM32 controllers, motor drive module and four wheels according to different heads
The rotation of action command controlled motor different directions and speed, so as to reach the result that wheelchair completes different motion state.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, various equivalent modifications can be readily occurred in or replaced
Change, these modifications or substitutions should be all included within the scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection domain be defined.
Claims (10)
- A kind of 1. wheelchair control method based on the processing of neck muscle signals, it is characterised in that this method comprises the following steps：S1, by sensor gather neck muscle signals MMG, the input signal as training pattern；S2, using training pattern Signal Pretreatment, segmentation, feature extraction, dimension-reduction treatment and head are carried out to neck muscle signals MMG Portion's action command Modulation recognition identification；S3, the headwork command signal that will identify that reach wheelchair controller, controling wheelchair operation.
- A kind of 2. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that institute The feature extraction stated includes wavelet package transforms coefficient power feature extraction and main diagonal slices feature extraction.
- A kind of 3. wheelchair control method based on the processing of neck muscle signals according to claim 2, it is characterised in that institute The main diagonal slices stated are characterized as w in bispectrum signal1=w2When characteristic signal, described bispectrum signal formula is：<mrow> <msub> <mi>B</mi> <mi>x</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>w</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>w</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <msub> <mi>&tau;</mi> <mn>1</mn> </msub> <mo>=</mo> <mo>-</mo> <mi>&infin;</mi> </mrow> <mrow> <mo>+</mo> <mi>&infin;</mi> </mrow> </munderover> <munderover> <mo>&Sigma;</mo> <mrow> <msub> <mi>&tau;</mi> <mn>2</mn> </msub> <mo>=</mo> <mo>-</mo> <mi>&infin;</mi> </mrow> <mrow> <mo>+</mo> <mi>&infin;</mi> </mrow> </munderover> <msub> <mi>C</mi> <mrow> <mn>3</mn> <mi>x</mi> </mrow> </msub> <mrow> <mo>(</mo> <msub> <mi>&tau;</mi> <mrow> <mn>1</mn> <mo>,</mo> </mrow> </msub> <msub> <mi>&tau;</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> <mi>exp</mi> <mo>{</mo> <mo>-</mo> <mi>j</mi> <mrow> <mo>(</mo> <msub> <mi>w</mi> <mn>1</mn> </msub> <msub> <mi>&tau;</mi> <mn>1</mn> </msub> <mo>+</mo> <msub> <mi>w</mi> <mn>2</mn> </msub> <msub> <mi>&tau;</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> <mo>}</mo> </mrow>Wherein Bx(w1,w2) it is bispectrum, C3x(τ1,τ2) be random process Third-order cumulants, τ1And τ2It is lag time, and | w1 |≤π,|w2|≤π,|w1+w2|≤π, w1、w2Respectively frequency variable, x represent certain sample.
- A kind of 4. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that institute The headwork command signal stated includes six kinds of control models：Respectively forward, to the left, to the right, backward, accelerate and stopping refer to Order.
- A kind of 5. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that step Neck muscle signals MMG when human body head acts, the input signal as training pattern are gathered by sensor in rapid S1.
- A kind of 6. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that step The muscle signals MMG at four positions of neck is gathered in rapid S1 by sensor, described four positions are locked including arranged on left and right sides chest The musculi splenius capitis of papillary muscle and arranged on left and right sides.
- A kind of 7. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that institute The Signal Pretreatment stated includes filtering and normalization.
- A kind of 8. wheelchair control method based on the processing of neck muscle signals according to claim 1, it is characterised in that institute The sensor stated is TD-3 acceleration transducers.
- 9. a kind of wheelchair control method based on the processing of neck muscle signals using as described in any one of claim 1~8 Wheelchair control system, it is characterised in that described control system includes sensor, computer, the wireless telecommunications mould being sequentially connected Block, controller and motor drive module.
- 10. a kind of wheelchair control system according to claim 9, it is characterised in that described sensor accelerates for TD-3 Sensor is spent, described controller is STM32 controllers.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710958725.5A CN107744436A (en) | 2017-10-16 | 2017-10-16 | A kind of wheelchair control method and control system based on the processing of neck muscle signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710958725.5A CN107744436A (en) | 2017-10-16 | 2017-10-16 | A kind of wheelchair control method and control system based on the processing of neck muscle signals |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107744436A true CN107744436A (en) | 2018-03-02 |
Family
ID=61253946
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710958725.5A Pending CN107744436A (en) | 2017-10-16 | 2017-10-16 | A kind of wheelchair control method and control system based on the processing of neck muscle signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107744436A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491792A (en) * | 2018-03-21 | 2018-09-04 | 安徽大学 | Office scene human-computer interaction Activity recognition method based on electro-ocular signal |
| CN109171124A (en) * | 2018-09-11 | 2019-01-11 | 华东理工大学 | A kind of muscle signals wireless collection bracelet for Sign Language Recognition |
| CN117084872A (en) * | 2023-09-07 | 2023-11-21 | 中国科学院苏州生物医学工程技术研究所 | Walking aid control method, system and medium based on neck myoelectricity and walking aid |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102013016A (en) * | 2010-11-23 | 2011-04-13 | 华东理工大学 | Muscle sound signal-based hand motion mode identification method for prosthetic hand control |
| CN102614061A (en) * | 2012-03-01 | 2012-08-01 | 上海理工大学 | Human body upper limb functional rehabilitation training implement method based on muscle tone signals |
| CN103488995A (en) * | 2013-08-31 | 2014-01-01 | 中山大学 | Method for identifying rotation of neck |
-
2017
- 2017-10-16 CN CN201710958725.5A patent/CN107744436A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102013016A (en) * | 2010-11-23 | 2011-04-13 | 华东理工大学 | Muscle sound signal-based hand motion mode identification method for prosthetic hand control |
| CN102614061A (en) * | 2012-03-01 | 2012-08-01 | 上海理工大学 | Human body upper limb functional rehabilitation training implement method based on muscle tone signals |
| CN103488995A (en) * | 2013-08-31 | 2014-01-01 | 中山大学 | Method for identifying rotation of neck |
Non-Patent Citations (2)
| Title |
|---|
| HIROAKI SEKI 等: ""A Powered Wheelchair Controlled by EMG Signals from Neck Muscles"", 《HUMAN FRIENDLY MECHATRONICS(ICMA 2000)》 * |
| 邱青菊: ""表面肌电信号的特征提取与模式分类研究"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491792A (en) * | 2018-03-21 | 2018-09-04 | 安徽大学 | Office scene human-computer interaction Activity recognition method based on electro-ocular signal |
| CN108491792B (en) * | 2018-03-21 | 2022-07-12 | 安徽大学 | Office scene human-computer interaction behavior recognition method based on electro-oculogram signals |
| CN109171124A (en) * | 2018-09-11 | 2019-01-11 | 华东理工大学 | A kind of muscle signals wireless collection bracelet for Sign Language Recognition |
| CN117084872A (en) * | 2023-09-07 | 2023-11-21 | 中国科学院苏州生物医学工程技术研究所 | Walking aid control method, system and medium based on neck myoelectricity and walking aid |
| CN117084872B (en) * | 2023-09-07 | 2024-05-03 | 中国科学院苏州生物医学工程技术研究所 | Walking aid control method, system and medium based on neck myoelectricity and walking aid |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lee et al. | A convolution neural networks scheme for classification of motor imagery EEG based on wavelet time-frequecy image | |
| CN110238863B (en) | Lower limb rehabilitation robot control method and system based on electroencephalogram-electromyogram signals | |
| Ahsan et al. | Hand motion detection from EMG signals by using ANN based classifier for human computer interaction | |
| CN102521505A (en) | Brain electric and eye electric signal decision fusion method for identifying control intention | |
| Zhang et al. | PCA and LDA for EMG-based control of bionic mechanical hand | |
| Bhagwat et al. | Electromyogram (EMG) based fingers movement recognition using sparse filtering of wavelet packet coefficients | |
| CN107744436A (en) | A kind of wheelchair control method and control system based on the processing of neck muscle signals | |
| Hettiarachchi et al. | Motor imagery data classification for BCI application using wavelet packet feature extraction | |
| CN110390272B (en) | EEG signal feature dimension reduction method based on weighted principal component analysis | |
| Moctezuma et al. | EEG-based Subjects Identification based on Biometrics of Imagined Speech using EMD | |
| Wang et al. | Classification of EEG signal using convolutional neural networks | |
| Poorna et al. | EEG based control-a study using wavelet features | |
| Hajibabazadeh et al. | Brain-robot interface: Distinguishing left and right hand EEG signals through SVM | |
| Wang et al. | Optimized preprocessing and tiny ml for attention state classification | |
| Chai et al. | Selecting optimal EEG channels for mental tasks classification: An approach using ICA | |
| Ahmadi et al. | Light-weight single trial EEG signal processing algorithms: computational profiling for low power design | |
| Ebrahimpour et al. | EEG-based motor imagery classification using wavelet coefficients and ensemble classifiers | |
| CN107423668A (en) | Eeg signal classification System and method for based on wavelet transformation and sparse expression | |
| Abdulwahab et al. | EEG motor-imagery BCI system based on maximum overlap discrete wavelet transform (MODWT) and cubic SVM | |
| CN116756643A (en) | Fisher score-based electroencephalogram channel selection method | |
| CN105824321A (en) | Four-axis aircraft control system and method based on surface electromyogram signals | |
| CN110604578A (en) | Human hand and hand motion recognition method based on SEMG | |
| Zhu et al. | An approach for semg-based gesture recognition using continuous wavelet transform and alexnet convolutional neural network | |
| Wang et al. | EMG-based hand gesture recognition by deep time-frequency learning for assisted living & rehabilitation | |
| Trad et al. | Nonlinear and nonstationary framework for feature extraction and classification of motor imagery |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180302 |
|
| RJ01 | Rejection of invention patent application after publication |