CN101785704B - Self-adaptive filtering device of master-slave minimally-invasive surgery robot system - Google Patents

Abstract

The invention discloses a self-adaptive filtering device of a master-slave minimally-invasive surgery robot system, comprising a trembling action self-adaptive filtering device, a master manipulator, a position collecting module, a motion control module, a slave manipulator driving module, a detection module, a slave manipulator, a feedback module and a computer control system. The trembling action filtering device is used for filtering hand trembling actions of a surgery operator and restoring a surgery operation desired signal to the maximum degree; the master manipulator driving module in the filtering device is used for driving the master manipulator; the position collecting module is used for collecting a position signal of the master manipulator; the motion control module is used for processing the collected position signal and controlling a motor to drive the slave manipulator to complete the surgery operation; finally, the feedback module is used for supplying real-time image feedback information; therefore, a closed-loop minimally-invasive surgery robot control system is formed. The invention can effectively filter the hand trembling actions, which ensure high precision and reliability of the minimally-invasive surgery.

Description

A kind of adaptive filter device of master-slave minimally-invasive surgery robot system

Technical field

The present invention is a kind of adaptive filter device of master-slave minimally-invasive surgery robot system, belongs to the renovation technique of the adaptive filter device of master-slave minimally-invasive surgery robot system.

Background technology

Along with development of science and technology, medical robotic system becomes the most active in the robot research field and invests one of maximum direction, and micro-wound surgical operation (MIS) is focus application the most among the Medical Robot.Minimally-invasive surgery robot system is the typical product that Minimally Invasive Surgery and the robotics in the medical science combines, its successful Application make Minimally Invasive Surgery degree of accuracy, reliability and handling aspect produced matter raising.Comparative maturity should be the endoscope's automatic station-keeping system that is used to assist Minimally Invasive Surgery (Aesop Aesop) and the ZEUS system of U.S. Computer Motion company development.

Wherein Aesop Aesop adopts cascaded structure, and the ZEUS system adopts the hand operating technology of principal and subordinate, and the two can imitate the function of human arm, can accurate more and control operation stably.Leonardo da Vinci (Da Vinci) the minimally invasive surgical operation robot system that develops of American I ntuitive Surgical company has obtained European CE and U.S. food and FAD (FDA) authentication in addition.This system adopts principal and subordinate's hands structure, and the doctor controls main manipulator at control station, and by come the concrete process that realizes operation from operator, the master-slave manipulator in this system can accomplish the very meticulous action that staff is difficult to accomplish.The whole surgery process just can be accomplished through very little operative incision, has reduced the risk of wound infection, has shortened the time of operation process and the rehabilitation duration of patient's postoperative, has improved surgical effect, has reduced patient's misery.It is domestic that main achievement is the neurosurgery robot that is designed by the robot of BJ University of Aeronautics & Astronautics in the minimally invasive surgical operation robot field.So far, existing more than 20 tame hospitals have adopted this surgical robot system, and have successfully accomplished the operation of 5000 many cases, have proved the reliability and the advantage that change system, have obtained good clinical and social benefit.

Although Minimally Invasive Surgery is being brought into play very big effect in improving the medical operating quality, be still waiting to improve at MIS aspect high accuracy and the real-time performance.Some need that hand is directly got involved, in the Minimally Invasive Surgery of non-remote sensing; Because the influence of trembling of people's hand various degrees; There are certain deviation in the actual input information of operator's hand and the desired input information of operator in the operation process; Having reduced as master-slave mode MIS is this needs that hand is directly got involved, the precision of the medical operating of non-remote sensing, has had influence on the quality of operation.This hand problem of trembling has caused a lot of scholars' concern, has also excited our interest to trembling and studying.

Trembling is a kind of class rectilinear oscillation signal at random that is superimposed upon on the desired signal, mainly shows people's extremity and head.Tremble and mainly be divided into two big types: physiology trembles and pathology trembles.It is that we normal person is inherent but do not influence a kind of vibration of our daily life that physiology trembles.Research shows that this trembling is to receive people's pressure, nervous system in daily life to also have some other factors to influence, and have very little amplitude, and its energy spectrum is distributed in mainly in the scope that frequency is 8Hz-12Hz.On the contrary, it is a kind of morbid state vibration that is caused by acquired disposition that pathology trembles, and affects the orthobiosis of much human, even under some occasion, makes our normal running to realize.Cause that the factor that pathology trembles has a lot, as: XIAONAO is injured, cerebral palsy, Parkinson's disease, multiple sclerosis and ataxia etc.Pathology trembles and has bigger amplitude, and its energy spectrum mainly is distributed in the such low-frequency range of 2Hz-6Hz.

Along with the robot development of high-tech, teleoperator has also obtained application at medical field.Tremble to the influence of operation though teleoperator can reduce operator's hand to a certain extent, its real-time is still waiting to improve.In view of needing hand directly to get involved., non-remote sensing MIS can make equipment cost lower; And more real-time feedback information can be provided; Let operation technique person sensation also can bring into play these advantages of surgical experience of operator better more naturally; Current in a lot of Minimally Invasive Surgerys the operation technique person relatively favor in the MIS that needs hand directly to get involved, this makes that also the research that suppresses of trembling seems to have necessity more to hand.

So far existing a lot of scholars have done relevant research and have proposed many feasible methods the inhibition of trembling, and wherein a lot of methods realize through low pass, bandpass filtering.These methods mainly contain 2 deficiencies; At first; Aspect accuracy; Because these filtering methods much are exactly a definite value being provided with of filtering midband width, this will cause the reservation of the losing of useful information and the information of trembling to a certain extent, so can not describe the signal that trembles exactly.Secondly, aspect real-time, because the existence meeting of wave filter causes hysteresis effect on certain degree, the signal that makes the operator import can not get handling timely, has influenced the quality of operation.Jing Zhang and Fang Chu propose can realize signal in real time modeling and the prediction of trembling with three rank AR models.The adaptive notch filter that Riviere and Thakor proposed is all having good effect aspect accuracy and the real-time.This method is to adopt the linear Fourier's equalizer (WFLC) based on weights that the signal that trembles is carried out modeling from these three aspects of the frequency of the signal that trembles, amplitude and phase place; And produce amplitude with the signal that trembles but compensating signal that phase place opposite identical with frequency, just realized filtering that hand is trembled with the signal (contain and tremble) of this compensating signal and the actual input of operator's hand is superimposed again.

Fourier analysis method is not so good as wavelet transformation in the characteristic that is shown aspect some.Such as aspect the time-frequency characteristic, because the weight coefficient of Fourier analysis is the function of frequency, and the weight coefficient in the wavelet transformation is the binary function of frequency and time, and this just makes the time-frequency characteristic of wavelet transformation be better than Fourier analysis.Fourier analysis can only but can not get its corresponding time-domain information at frequency domain to the signal analysis of trembling, and this will certainly have influence on our analysis to the signal that trembles.At the localization aspect of performance, the windowing Fourier transformation is all got identical window width in time domain, and the window width of wavelet transformation then is adjustable, and it uses short window when high frequency, when low frequency, then use wide window.The adaptive time frequency window that wavelet transformation had, but and the local time-domain position of focus analysis signal and local band characteristic be that Fourier analysis is incomparable, this also is why people are referred to as wavelet transformation " school microscop ".For can be accurately to tremble signal analysis and modeling we need adopt a kind of window with self adaptation resolution characteristic, and the self-adapting window in the wavelet transformation should be worth us to go to consider.In addition, Fourier analysis is suitable for the processing of gradual change signal and live signal, but the sudden change of reflected signal sensitively; And wavelet analysis is suitable for jump signal or have processing and the processing of self-adapting signal of the function of isolated singularity.Since operator's hand tremble along with the influence of several factors; Such as: hand exercise track, pathological characters, psychological factor and environmental factors etc.; The wherein variation of any one factor sudden change that all possibly cause hand to tremble is so the behavior of trembling of hand is fit to adopt wavelet transformation to handle.In view of there is the contradiction of time domain and frequency localization in Fourier analysis, lack spatial locality, and operator's hand tremble be a kind of at random, unsettled, the time change signal.

Summary of the invention

The objective of the invention is to consider the problems referred to above and a kind of adaptive filter device that reaches the master-slave minimally-invasive surgery robot system of the filtration result that well trembles is provided.The present invention is convenient and practical.

Technical scheme of the present invention is: the adaptive filter device of master-slave minimally-invasive surgery robot system of the present invention; Include the behavior adaptive filter device that trembles, main manipulator, station acquisition module, motion-control module, from operator driver module, detection module, from operator, feedback module and computer control system; The operation signal that the operator sends obtains the quasiexpectation operation signal via the behavior adaptive filter device of trembling; And by the action of quasiexpectation operation signal driving main manipulator; The station acquisition module that links to each other with main manipulator is gathered the main manipulator spatial positional information; Send this positional information to motion-control module again; Realize processing by computer control system assisted movement control module to the main manipulator spatial positional information; And send control signal from operator, this control signal is amplified rear drive through the operator driver module of associating and is moved electric current, speed and the positional information of detection module detection motor from the operator driver module from operator; And feeding back to motion-control module to realize closed loop control from operator, computer control system realizes and the communicating by letter and monitoring between behavior adaptive filter device, motion-control module and the feedback module of trembling.

The above-mentioned behavior adaptive filter device that trembles comprises the inertia measurement module, the filtration module that trembles, s operation control module and main manipulator driver module; The locus acceleration and the joint angle velocity information of inertia measurement module detecting operation person hand; And this is transferred to the s operation control module; Convert locus acceleration signal and joint angle rate signal into locus signal and joint angles signal respectively by the s operation control module, and send this to filtration module that trembles, the filtering of trembling by the filtration module realization of trembling again; And obtain the quasiexpectation operation signal; Again this quasiexpectation operation signal is input to the s operation control module and handles, and obtain a driving signal, this driving signal drives main manipulator via the main manipulator driver module.

The inertial measurement module contains a three-dimensional acceleration sensor module is used to measure the surgical operator hand position in three-dimensional space, acceleration, that:

A three-dimensional angular velocity sensor module for measuring surgical hand operator in three-dimensional space velocity namely:

The above-mentioned filtration module that trembles adopts Fuzzy Wavelet Network sef-adapting filter filtering operator hand its desired signal of signal reduction that trembles; Said Fuzzy Wavelet Network sef-adapting filter obtains compensating signal x ', y ', z ' and θ ' identical with tremble signal amplitude and frequency but that phase place is opposite through the modeling to the signal that trembles _x, θ ' _y, θ ' _zSaid Fuzzy Wavelet Network sef-adapting filter comprises obfuscation, fuzzy rule coupling, fuzzy reasoning and the defuzzification of input quantity to the modeling process of the signal that trembles; Fuzzy Wavelet Network comprises seven layers; Input layer, obfuscation layer, fuzzy rule layer, wavelet network layer, fuzzy reasoning layer, reconstruction of layer and output layer.

Above-mentioned s operation control module comprises AD conversion unit, D/A conversion unit, bandwidth filter, pose collecting unit, inverse kinematics computing unit and simple joint control unit; This s operation control module and computer control system two-way communication; To realize computing and control function; AD conversion unit will comprise that the hand analog-signal transitions of locus acceleration signal and joint angle rate signal is a digital signal by the collection of inertia measurement module; Repeated transmission is given bandwidth filter; With the clocking noise signal that filtering inertia measurement module produces, the pose collecting unit is locus signal and joint angles signal with the digital signal transition after bandwidth filter is handled, and sends the filtration module that trembles to and carry out Filtering Processing; The quasiexpectation operation signal that the inverse kinematics computing unit is exported the filtration module that trembles carries out inverse kinematics and calculates joint variable; By this joint variable of simple joint control unit control, and output will change analogue signal to the control signal of main manipulator into by D/A conversion unit again and directly send the main manipulator driver module to drive main manipulator the control signal of main manipulator.

Above-mentioned main manipulator driver module comprises power amplifier and Piexoelectric actuator, in order to drive main manipulator, makes it the desired track action according to the operator of operation; Comprise driver, motor and actuating device from the operator driver module, realize the DSP control module and from the driving between the operator.

Above-mentioned main manipulator directly is connected with the behavior adaptive filter device of trembling, and realizes the operator of operation and the human computer conversation between the minimally-invasive surgery robot system; Above-mentioned station acquisition module realizes the collection to the positional information of main manipulator, quantizes the movement locus of main manipulator, and directly imports the positional information of gathering into motion-control module; The DSP control module that adopts above-mentioned motion-control module realizes three closed loop controls and PWM control; The outer shroud of said three closed loop controls is the Position Control ring, and innermost ring is a current regulator, and a middle ring is the speed controlling ring, said DSP control module and computer control system realization two-way communication.

Above-mentioned detection module realizes detecting and providing the closed loop feedback signal of three closed loop controls, comprises A/D converter, current sensor, photoelectric encoder, QEP circuit and frequency measurement circuit; The pulse signal of the photoelectric encoder output on the machine shaft is transferred to QEP circuit and frequency measurement circuit; Pulse signal obtains position feed back signal through the QEP processing of circuit; And sending the Position Control ring in the motion-control module to, pulse signal is handled through frequency measurement circuit, obtains feedback speed signal; And send the rate control module in the motion-control module to; Current sensor senses motor windings electric current, and obtain its digital current signal through A/D converter, send it in the motion-control module current regulator again.

Above-mentioned is that the most key and patient get in touch a closest unit the minimally-invasive surgery robot system from operator, operating theater instruments is housed on it, and accomplishes the master-slave mode micro-wound surgical operation thus; Above-mentioned feedback module is realized the supervision of minimal invasive surgical procedures and real-time information feedback through the graphics processing unit in endoscope and monitor and the computer control system, and making whole master-slave minimally-invasive surgery robot system is a closed-loop control system.

Aforementioned calculation machine control system comprises communication unit, Operations Analysis and graphics processing unit; Said Operations Analysis and filter two-way communication realize the computings such as digital-to-analogue conversion, analog digital conversion, inverse kinematics calculating in the filter, and the simple joint control unit is monitored; Motion-control module is realized the transmission of data through communication unit and computer control system; Graphics processing unit is accepted the image information of endoscope's output in the feedback module, this is handled and sends to the monitor in the feedback module; Said Operations Analysis is all realized by communication unit with communicating by letter of other intermodules with graphics processing unit.

The present invention considers that the behavior of trembling of people's hand belongs to a kind of behavior of men owing to adopt hand the tremble modeling and the filtering of signal to introduce wavelet transformation, and the present invention adopts fuzzy language to express this behavior.The experience of operation technique person in operation process also is a very The key factor for operation quality and success or failure in addition; The present invention can portray this operation technique experience with fuzzy rule; It is incorporated in the minimal invasive surgical procedures, so adopted fuzzy control theory to describe this operation process qualitatively among the present invention.The present invention from wavelet transformation and these two angle design of fuzzy control a adaptive filter device based on Fuzzy Wavelet Network, overcome based on linear Fourier's equalizer of weights and surgical experience indifferent based on the time frequency analysis of BP neutral net wave filter in the filtering that trembles and the person that can't incorporate the operation technique.Adaptive filter device among the present invention is through the parameter in the automatic correction wave filter of study of neutral net.This method for designing of experiment proof has reached the filtration result that well trembles.The present invention is that a kind of design is ingenious, function admirable, the adaptive filter device of convenient and practical master-slave minimally-invasive surgery robot system.

Description of drawings

Fig. 1 master-slave mode minimally invasive surgical operation robot overall system block diagram;

Fig. 2 behavior adaptive filter device theory diagram that trembles;

Fig. 3 behavior adaptive filter device mathematical model figure that trembles;

Fig. 4 Fuzzy Wavelet Network adaptive-filtering figure;

Fig. 5 operation technique person hand desired operation signature tune line chart;

Fig. 6 operation technique person is trembled influences the operation signal curve chart of the actual output of hand;

Fig. 7 FWNN sef-adapting filter and BP neutral net wave filter are to the error curve diagram of the signal filtering of trembling;

Fig. 8 BP neutral net wave filter is to the reduction curve chart of the disturbed operation signal of reality.

Fig. 9 FWNN sef-adapting filter is to the reduction curve chart of the disturbed operation signal of reality.

The specific embodiment

Embodiment:

The present invention relates to the tremble adaptive filter device of behavior of a kind of filtering Minimally Invasive Surgery operator hand; Based on the approximation capability of Fuzzy Wavelet Network (FWNN) to any nonlinear function; To the tremble off-line modeling of behavior of operation operator hand; And can produce a kind of and the signal amplitude that trembles, frequency is identical but compensating signal that phase place is opposite, thereby reach the tremble purpose of signal of filtering.

Below in conjunction with accompanying drawing and instantiation the Minimally Invasive Surgery operator hand that the present invention the designed behavior adaptive filter device that trembles is carried out detailed explanation.

Fig. 1 is a master-slave mode minimally invasive surgical operation robot overall system block diagram of the present invention.The operating procedure that the operation technique person performs the operation well according to the medical image planning of patient's focus in advance in Fig. 1.Individual adaptive filter device is arranged between operation technique person and main operation person; The arithmetic element of this adaptive filter device in computer control system auxiliary down with operator's hand target signal filter that trembles; Provide a quasiexpectation signal to drive main manipulator, make it to accomplish operation process according to the desired surgical operation step of operator.The station acquisition module is gathered into discrete digital signal with main manipulator at spatial movement position, and in the motion-control module the Position Control ring of sending.The Position Control ring to from the control of operator enforcing location, and is exported the quasiexpectation rate signal of a rate signal as the speed controlling ring according to input quasiexpectation position signalling and the position feed back signal that provided by detection module.The speed controlling ring to implementing speed controlling from operator, and is exported the quasiexpectation current signal of a current signal as electric current loop according to the feedback speed signal that provides by the quasiexpectation rate signal of Position Control ring input with by detection module.Current regulator is implemented Current Control according to the quasiexpectation current signal of being imported by the speed controlling ring with by the current feedback signal that detection module provides to motor again, and output voltage control signal is as the voltage control signal of PWM control module.The PWM control module changes the voltage that offers motor driver according to the voltage control signal adjustment pulse width of input, thereby realizes the control to motor.Driver according to the Control of Voltage rotating speed of motor of PWM control module output with turn to.Wherein Position Control ring, speed controlling ring, current regulator and PWM control is all realized by dsp controller, this dsp controller directly with computer control system in communication unit implementation two-way communication.Realize by motor and actuating device from the motion of operator, operating theater instruments is installed on it just can implements operation the patient.Settle endoscope just can in operation process, send the present situation of operation in the computer control system image processing module with pictorial form in real time at patient's affected area; After Image Information Processing is intact again by computer control system with image information through the monitor person that feeds back to the operation technique, the operation technique person makes the surgical effect of corresponding adjustment to guarantee to reach best according to the image of feedback.

Fig. 2 is the behavior adaptive filter device theory diagram that trembles of the present invention.The said operation technique person's of Fig. 1 hand behavior realizes being connected through adaptive filter device and main manipulator; The main purpose of this device is the behavior of the trembling filtering with operation technique person hand in operative process, to greatest extent the operation technique of reduction expectation.The tremor behavior in adaptive filtering device first is through inertial measurement unit and three-dimensional three-dimensional acceleration sensor module were used to measure angular velocity sensing module Chucao hand in the space of three-dimensional acceleration signals

and three-dimensional angular velocity signal

and then through analog-digital conversion unit to the measured analog signals into digital signals the computer can process.Signal by analog quantity convert into use after the digital quantity frequency range as in this signal of bandwidth filter filtering of 2.5Hz-50Hz by the caused clocking noise signal of measurement module.The pose acquisition module is locus signal x from handling the pose signal of gathering operation technique person hand afterwards the information via bandwidth filter again, y, z and space anglec of rotation signal θ _x, θ _y, θ _zThe locus signal that obtains and space anglec of rotation signal carry out the modeling of off-line through the FWNN sef-adapting filter to the behavior of trembling as the input quantity of the filtration module that trembles, output tremble signal estimated value promptly: x, y, z and θ ' _x, θ ' _y, θ ' _zThis estimated value negate has just been accomplished the tremble filtering of behavior of hand as the posture information of compensating signal that trembles and the collection of pose acquisition module is superimposed.The operation technique signal of process Filtering Processing carries out inverse kinematics by the Operations Analysis in the computer control system to it and calculates joint variable λ ₁..., λ _n, and have by the simple joint controller to come its control, be the signal transition of simple joint controller output analog voltage signal v then ₁..., v _nSend power amplifier to, drive main manipulator through piezoelectric actuator at last.

Fig. 3 is the mathematical model of the said behavior adaptive filter device that trembles.The expectation of the person of operation technique in the present invention current time as shown in Figure 3 operation signal d (k), the operation technique person current time hand signal n (k) that trembles, the operation signal s (k) of the actual output of operation technique person current time hand, wherein s (k)=d (k)+n (k).The tremble estimated value of signal of Fuzzy Wavelet Network (FWNN) output current time The quasiexpectation operation signal y (k) of person's current time hand that after Filtering Processing, obtains the operation technique

Wherein That statistical module is exported is quasiexpectation operation signal power E [y ²(k)].Can know by Fig. 3; The pose signal s (k) that measures from Inertial Measurement Unit obtains its past pose signal constantly promptly after delay component is handled: s (k-1); S (k-2); These delay signals of s (k-n) are as the input quantity of Fuzzy Wavelet Network, finally obtain current time through obfuscation, fuzzy rule coupling, fuzzy reasoning and defuzzification and tremble the estimated value

of signal with the estimated value negate of this signal that trembles and the superimposed quasiexpectation that has just obtained preceding moment hand of the operation signal s (k) of the actual output of the preceding moment hand signal y (k) that performs the operation.Sef-adapting filter among the present invention is the signal power E [y that calculates the quasiexpectation signal y (k) of current time hand through statistical module ²(k)], be that the parameter that criterion is revised in the Fuzzy Wavelet Network realizes the self study process to minimize this signal power.

Fig. 4 is said Fuzzy Wavelet Network adaptive-filtering figure.This Fuzzy Wavelet Network has combined fuzzy theory, wavelet analysis and neuron computes, has brought into play it to the advantage of neutral signal modeling study at random, has reached ideal filter effect.The network structure that is designed in the present invention is seven layers.

Specific as follows:

1) input layer (ground floor)

N+1 neuron arranged in this layer, corresponding input signal be actual disturbed signal and its time delayed signal promptly: s (k), s (k-1) ..., s (k-n), they will directly be sent down to one deck, that is: $O_i^{\left(1\right)}=I_i^{\left(1\right)}$

Wherein $I_i^{\left(1\right)}=x=[s\left(k\right),s(k-1),...,s(k-n)]$ Be the input signal of ground floor,

Represent the output signal of ground floor.The neuron number of this layer is n+1 altogether.

2) obfuscation layer (second layer)

This layer is that input quantity is carried out Fuzzy processing.We adopt the Gaussian function as membership function at this, c _IjAnd σ _IjBe respectively the average and the standard deviation of the membership function of i input variable and j fuzzy set, that is:

$O_{\mathrm{ij}}^{\left(2\right)}=\exp (-\frac{{(O_i^{\left(1\right)}-c_{\mathrm{ij}})}^2}{{\left({\mathrm{\σ}}_{\mathrm{ij}}\right)}^2})$

I=1 wherein; 2 ..., n+1j=1; 2;, m,

represents the output signal of the second layer.The neuron number of this layer is (n+1) * m altogether.

3) fuzzy rule layer (the 3rd layer)

Each node of this layer has all been represented a fuzzy rule, and each node j is output as the product of these all input signals of node, that is:

$O_j^{\left(3\right)}=\underset{i}{\overset{n+1}{\mathrm{\Π}}}{O}_{\mathrm{ij}}^{\left(2\right)}$

J=1 wherein, 2 ..., N, N=m ⁽ⁿ⁺¹⁾Be the fuzzy rule number of this layer,

Represent trilaminar output signal.The neuron number of this layer is N altogether.

4) wavelet network layer (the 4th layer)

This layer is a wavelet network, adopts Mexico's straw hat wavelet function as neuronic excitation function, and input quantity is s (k), s (k-1) ..., s (k-n), output are the consequent part y of fuzzy rule _l(l＝1，2，…，N)

Mexico's straw hat wavelet function:

A wherein _jAnd b _jRepresent the expansion parameter and the translation parameters of wavelet function respectively, ψ _j(x) represent wavelet function family,

Represent wavelet mother function, x is an input variable, x=in invention [s (k), s (k-1) ..., s (k-n)].

L wavelet network is output as:

$O_l^{\left(4\right)}=\underset{k=1}{\overset{4}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{\mathrm{kl}}{\mathrm{\&psi;}}_{\mathrm{kl}}\left(z_{\mathrm{kl}}\right)=\underset{k=1}{\overset{4}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_{\mathrm{kl}}{\left|a_{\mathrm{kl}}\right|}^{-\frac{1}{2}}(1-z_{\mathrm{kl}}^2)e^{-\frac{{\mathrm{<figure-callout\; id="2"\; label="z\; kl"\; filenames="H2010100194526E00021.png"\; state="\{\{state\}\}">z</figure-callout>}}_{\mathrm{kl}}^{}}{2}}$

Z wherein _Kl=(X-b _Kl)/a _Kl,

K=1 in invention, 2 ..., on behalf of each wavelet network, v, v contain the number of wavelet function.L=1,2 ..., N, N represent the number of wavelet network, ω _Kl, Be respectively the weights and the output of wavelet network layer.

5) fuzzy reasoning layer (layer 5)

This layer through with the coupling of accomplishing fuzzy rule layer and being connected of wavelet network layer the front and back part of fuzzy rule, realize fuzzy operation between each node, i.e. combination through each fuzzy node obtains corresponding ignition intensity.Each node j is output as the product of these all input signals of node, that is:

$O_j^{\left(5\right)}=O_j^{\left(4\right)}*O_j^{\left(3\right)}$

J=1 wherein; 2 ... N,

represents the output of fuzzy reasoning layer.

6) reconstruction of layer (layer 6)

This layer only contains two neurons, and they represent fuzzy reasoning computing the normalization molecule and the denominator part of arithmetic expression afterwards respectively.Each neuron is realized the function of summation.That is:

$O_1^{\left(6\right)}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_i^{\left(5\right)}$

$O_2^{\left(6\right)}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_i^{\left(3\right)}$

The number of N delegate rules wherein,

represents the molecule and the denominator part of arithmetic expression respectively.

7) output layer (layer 7)

This layer is the output of whole network, only contains a neuron, realization be division function.

That is:

$O^{\left(7\right)}=O_1^{\left(6\right)}/O_2^{\left(6\right)}$

O wherein ⁽⁷⁾Being the output of FWNN, also is the tremble estimated value of signal of current time

Said Fuzzy Wavelet Network study mechanism is the signal power E [y that calculates the quasiexpectation operation signal y (k) of preceding hand constantly through statistical module through said ²(k)], be that the parameter that criterion is revised in the Fuzzy Wavelet Network realizes to minimize this signal power.Sef-adapting filter makes E [y ²(k)] minimum makes exactly

Minimum.And by formula

Can release and work as

Hour, E [(y (k)-d (k)) ²] also be minimum, promptly

Minimum.The parameter that need revise in the present invention has: the average c in the Gaussian membership function _IjAnd variances sigma _IjThe weights ω of wavelet network _Kl, expansion parameter a _KlWith translation parameters b _Kl(i=1,2 ..., n+1, j=1,2 ..., m, k=1,2 ..., v, l=1,2 ..., N) the renewal equation formula of these parameters is following:

${\mathrm{\&omega;}}_{\mathrm{kl}}(k+1)={\mathrm{\&omega;}}_{\mathrm{kl}}\left(k\right)-\mathrm{\&gamma;}\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&omega;}}_{\mathrm{kl}}}+\mathrm{\&lambda;}({\mathrm{\&omega;}}_{\mathrm{kl}}\left(k\right)-{\mathrm{\&omega;}}_{\mathrm{kl}}(k-1))$

$a_{\mathrm{kl}}(k+1)=a_{\mathrm{kl}}\left(k\right)-\mathrm{\&gamma;}\frac{\&PartialD;E}{\&PartialD;a_{\mathrm{kl}}}+\mathrm{\&lambda;}(a_{\mathrm{kl}}\left(k\right)-a_{\mathrm{kl}}(k-1))$

$b_{\mathrm{kl}}(k+1)=b_{\mathrm{kl}}\left(k\right)-\mathrm{\&gamma;}\frac{\&PartialD;E}{\&PartialD;b_{\mathrm{kl}}}+\mathrm{\&lambda;}(b_{\mathrm{kl}}\left(k\right)-b_{\mathrm{kl}}(k-1))$

$c_{\mathrm{ij}}(k+1)=c_{\mathrm{ij}}\left(k\right)-\mathrm{\&gamma;}\frac{\&PartialD;E}{\&PartialD;c_{\mathrm{ij}}}+\mathrm{\&lambda;}(c_{\mathrm{ij}}\left(k\right)-c_{\mathrm{ij}}(k-1))$

${\mathrm{\&sigma;}}_{\mathrm{ij}}(k+1)={\mathrm{\&sigma;}}_{\mathrm{ij}}\left(k\right)-\mathrm{\&gamma;}\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&sigma;}}_{\mathrm{ij}}}+\mathrm{\&lambda;}({\mathrm{\&sigma;}}_{\mathrm{ij}}\left(k\right)-{\mathrm{\&sigma;}}_{\mathrm{ij}}(k-1))$

Wherein γ is a learning rate, γ ∈ [0,1], and λ is a factor of momentum, λ ∈ [0,1].

In the following formula $\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&omega;}}_{\mathrm{Kl}}},\frac{\&PartialD;E}{\&PartialD;a_{\mathrm{Kl}}},\frac{\&PartialD;E}{\&PartialD;b_{\mathrm{Kl}}},\frac{\&PartialD;E}{{\&PartialD;c}_{\mathrm{Ij}}},\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&sigma;}}_{\mathrm{Ij}}}$ Provide by following equation, that is:

$\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&omega;}}_{\mathrm{kl}}}=\frac{\&PartialD;E}{\&PartialD;\hat{d}}\frac{\&PartialD;\hat{d}}{\&PartialD;\hat{n}}\frac{\&PartialD;\hat{n}}{{\&PartialD;O}_l^{\left(4\right)}}\frac{{\&PartialD;O}_l^{\left(4\right)}}{{\&PartialD;\mathrm{\&omega;}}_{\mathrm{kl}}}$

$=(d\left(k\right)-\hat{d}\left(k\right))O_l^{\left(3\right)}{\mathrm{\&psi;}}_{\mathrm{kl}}\left(z_{\mathrm{kl}}\right)/\underset{l=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_l^{\left(3\right)}$

$\frac{\&PartialD;E}{\&PartialD;a_{\mathrm{kl}}}=\frac{\&PartialD;E}{\&PartialD;\hat{d}}\frac{\&PartialD;\hat{d}}{\&PartialD;\hat{n}}\frac{\&PartialD;\hat{n}}{{\&PartialD;O}_l^{\left(4\right)}}\frac{{\&PartialD;O}_l^{\left(4\right)}}{{\&PartialD;\mathrm{\&psi;\&omega;}}_{\mathrm{kl}}}\frac{\&PartialD;{\mathrm{\&psi;}}_{\mathrm{kl}}}{\&PartialD;a_{\mathrm{kl}}}$

$=(d\left(k\right)-\hat{d}\left(k\right))O_l^{\left(3\right)}{\mathrm{\&omega;}}_{\mathrm{kl}}{e}^{-\frac{z_{\mathrm{kl}}^2}{2}}[-\frac{1}{2}(1-z_{\mathrm{kl}}^2)+a_{\mathrm{kl}}^{-2}{(X-b_{\mathrm{kl}})}^2(3-z_{\mathrm{kl}}^2)]/\left(a_{\mathrm{kl}}^{-\frac{3}{2}}\underset{l=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_l^{\left(3\right)}\right)$

$\frac{\&PartialD;E}{\&PartialD;b_{\mathrm{kl}}}=\frac{\&PartialD;E}{\&PartialD;\hat{d}}\frac{\&PartialD;\hat{d}}{\&PartialD;\hat{n}}\frac{\&PartialD;\hat{n}}{{\&PartialD;O}_l^{\left(4\right)}}\frac{{\&PartialD;O}_l^{\left(4\right)}}{{\&PartialD;\mathrm{\&psi;\&omega;}}_{\mathrm{kl}}}\frac{\&PartialD;{\mathrm{\&psi;}}_{\mathrm{kl}}}{\&PartialD;b_{\mathrm{kl}}}$

$=(d\left(k\right)-\hat{d}\left(k\right))O_l^{\left(3\right)}{\mathrm{\&omega;}}_{\mathrm{kl}}{e}^{-\frac{z_{\mathrm{kl}}^2}{2}}(x-b_{\mathrm{kl}})(2-z_{\mathrm{kl}}^2)/\left(a_{\mathrm{kl}}^{\frac{3}{2}}\underset{l=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_l^{\left(3\right)}\right)$

$\frac{\&PartialD;E}{\&PartialD;c_{\mathrm{ij}}}=\underset{j}{\mathrm{\&Sigma;}}\frac{\&PartialD;E}{\&PartialD;\hat{d}}\frac{\&PartialD;\hat{d}}{\&PartialD;\hat{n}}\frac{\&PartialD;\hat{n}}{{\&PartialD;O}_j^{\left(3\right)}}\frac{{\&PartialD;O}_j^{\left(3\right)}}{{\&PartialD;c}_{\mathrm{ij}}}$

$=\left\{\begin{array}{c}{2O}_j^{\left(3\right)}(d\left(k\right)-\hat{d}\left(k\right))(O_j^{\left(4\right)}-\hat{n})(x_i-c_{\mathrm{ij}})/\left({\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="H2010100194526E00021.png"\; state="\{\{state\}\}">ij</figure-callout>}}^2\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_j^{\left(3\right)}\right)\\ 0\end{array}\right.$

$\frac{\&PartialD;E}{\&PartialD;{\mathrm{\&sigma;}}_{\mathrm{ij}}}=\underset{j}{\mathrm{\&Sigma;}}\frac{\&PartialD;E}{\&PartialD;\hat{d}}\frac{\&PartialD;\hat{d}}{\&PartialD;\hat{n}}\frac{\&PartialD;\hat{n}}{{\&PartialD;O}_j^{\left(3\right)}}\frac{{\&PartialD;O}_j^{\left(3\right)}}{{\&PartialD;c}_{\mathrm{ij}}}$

$=\left\{\begin{array}{c}{2O}_j^{\left(3\right)}(d\left(k\right)-\hat{d}\left(k\right))(O_j^{\left(4\right)}-\hat{n}){(x_i-c_{\mathrm{ij}})}^2/\left({\mathrm{\&sigma;}}_{\mathrm{ij}}^3\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{O}_j^{\left(3\right)}\right)\\ 0\end{array}\right.$

The equation of enumerating has above been accomplished parameter c _Ij, σ _Ij, ω _Kl, a _KlAnd b _KlStudy.

So far realized that adaptive filter device of the present invention is to the tremble filtering of behavior of operation operator hand.Below we verify the performance of the behavior of trembling adaptive filter device proposed by the invention with emulation experiment.

The present invention comes the observation experiment result through MATLAB emulation.The purpose of the present invention's experiment is the adaptive filter device of a kind of master-slave minimally-invasive surgery robot system more proposed by the invention and traditional performance based on BP neutral net adaptive filter device, verifies the filtering performance of the behavior adaptive filter device that trembles according to the invention.We adopt d=3sin (15 π t)+2cos (3 π t)+5t in this experiment ⁴-0.8t ³-2t is as operation technique person hand desired signal, with the tremble signal of n=0.3sin (0.04 π t)+0.1sin (0.0312 π t)+0.6sin (0.1 π) as operator's hand.Get sampling period T=0.01s, the sampling time is 10s, takes 1000 samples altogether, and wherein 500 samples are used for training, and 500 are used for test.We are provided with partial parameters in this experiment in the FWNN sef-adapting filter as follows: n=2, m=3, v=4, N=27.

Fig. 5 and Fig. 6 are respectively the operation technique signal of expectation and the practical operation signal that is trembled and influence.As can be seen from Figure 6 trembling of operator's hand badly influenced the operation of operation, must this carry out high accuracy and the reliability of Filtering Processing to guarantee Minimally Invasive Surgery.

Fig. 7 can clearly compare the error that adaptive filter device proposed by the invention and traditional BP neutral net wave filter approach trembling.Wherein on behalf of BP neutral net sef-adapting filter, black dotted lines follow the tracks of to tremble the error of signal, and solid black lines is represented the tremble error of signal of FWNN filter tracks.

Fig. 8 and Fig. 9 contrast the recovering signal characteristic of filter proposed by the invention and traditional BP neutral net sef-adapting filter.Can find out that signal that signal that the FWNN sef-adapting filter restores restores than BP neutral net sef-adapting filter is Paint Gloss and near desired signal.

Claims (8)

1. the adaptive filter device of a master-slave minimally-invasive surgery robot system; It is characterized in that including tremble behavior adaptive filter device (1), main manipulator (2), station acquisition module (3), motion-control module (4), from operator driver module (5), detection module (6), from operator (7), feedback module (8) and computer control system (9); The operation signal that the operator sends obtains the quasiexpectation operation signal via the behavior adaptive filter device (1) that trembles; And by quasiexpectation operation signal driving main manipulator (2) action; The station acquisition module (3) that links to each other with main manipulator (2) is gathered the main manipulator spatial positional information; Send this positional information to motion-control module (4) again; Realize processing by computer control system (9) assisted movement control module (4) to the main manipulator spatial positional information; And send from the control signal of operator (7), this control signal is amplified rear drive from operator (7) action through the operator driver module (5) of associating, and detection module (6) detects electric current, speed and the positional information of motor from operator driver module (5); And feed back to motion-control module (4) realizing closed loop control from operator (7), computer control system (9) realize with the behavior adaptive filter device (1) that trembles, motion-control module (4) and feedback module (8) between communicate by letter and monitoring;

The above-mentioned behavior adaptive filter device (1) that trembles comprises inertia measurement module (11), the filtration module that trembles (12), s operation control module (13) and main manipulator driver module (14); The locus acceleration and the joint angle velocity information of inertia measurement module (11) detecting operation person hand; And this is transferred to s operation control module (13); Convert locus acceleration signal and joint angle rate signal into locus signal and joint angles signal respectively by s operation control module (13), and send this to filtration module that trembles (12), the filtering of trembling by the filtration module that trembles (12) realization again; And obtain the quasiexpectation operation signal; Again this quasiexpectation operation signal is input to s operation control module (13) and handles, and obtain a driving signal, this driving signal drives main manipulator via main manipulator driver module (14);

The said filtration module that trembles (12) adopts Fuzzy Wavelet Network sef-adapting filter filtering operator hand its desired signal of signal reduction that trembles; Said Fuzzy Wavelet Network sef-adapting filter obtains compensating signal x ', y ', z ' and θ ' identical with tremble signal amplitude and frequency but that phase place is opposite based on Fuzzy Wavelet Network through the modeling to the signal that trembles _x, θ ' _y, θ ' _z, said Fuzzy Wavelet Network sef-adapting filter comprises obfuscation, fuzzy rule coupling, fuzzy reasoning and the defuzzification of input quantity to the modeling process of the signal that trembles; Fuzzy Wavelet Network comprises seven layers: input layer, obfuscation layer, fuzzy rule layer, wavelet network layer, fuzzy reasoning layer, reconstruction of layer and output layer.

2. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 1 is characterized in that above-mentioned inertia measurement module (11) comprises operator's hand that a three-dimensional acceleration sensing module is used to measure operation at spatial three-dimensional position acceleration promptly: operator's hand that

three dimensional angular velocity pick-up module is used to measure operation in spatial three dimensional angular speed promptly:

3. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 1 is characterized in that above-mentioned s operation control module (13) comprises AD conversion unit (131), D/A conversion unit (132), bandwidth filter (133), pose collecting unit (134), inverse kinematics computing unit (135) and simple joint control unit (136); This s operation control module (13) and computer control system (9) two-way communication; To realize computing and control function; AD conversion unit (131) will comprise that the hand analog-signal transitions of locus acceleration signal and joint angle rate signal is a digital signal by inertia measurement module (11) collection; Repeated transmission is given bandwidth filter (133); Clocking noise signal with filtering inertia measurement module (11) generation; Pose collecting unit (134) will be locus signal and joint angles signal through the digital signal transition after bandwidth filter (133) is handled; And send the filtration module that trembles (12) to and carry out Filtering Processing, inverse kinematics computing unit (135) carries out inverse kinematics to the quasiexpectation operation signal of the filtration module that trembles (12) output and calculates joint variable, controls this joint variable by simple joint control unit (136); And output will change analogue signal to the control signal of main manipulator into by D/A conversion unit (132) again and directly send main manipulator driver module (14) to drive main manipulator the control signal of main manipulator.

4. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 1; It is characterized in that above-mentioned main manipulator driver module (14) comprises power amplifier (141) and Piexoelectric actuator (142); In order to drive main manipulator, make it desired track action according to the operator of operation; Comprise driver (51), motor (52) and actuating device (53) from operator driver module (5), realize the DSP control module and from the driving between the operator.

5. according to the adaptive filter device of each described master-slave minimally-invasive surgery robot system of claim 1 to 4; It is characterized in that above-mentioned main manipulator (2) directly is connected with the behavior adaptive filter device (1) that trembles, realize the operator of operation and the human computer conversation between the minimally-invasive surgery robot system; Above-mentioned station acquisition module (3) realizes the collection to the spatial positional information of main manipulator, quantizes the movement locus of main manipulator, and directly imports the spatial positional information of the main manipulator of gathering into motion-control module; The DSP control module that adopts above-mentioned motion-control module (4) realizes three closed loop controls and PWM control; The outer shroud of said three closed loop controls is the Position Control ring, and innermost ring is a current regulator, and a middle ring is the speed controlling ring, said DSP control module and computer control system realization two-way communication.

6. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 5; It is characterized in that above-mentioned detection module (6) realization detects and provide the closed loop feedback signal of three closed loop controls, comprises A/D converter (61), current sensor (62), photoelectric encoder (63), QEP circuit (64) and frequency measurement circuit (65); The pulse signal of the photoelectric encoder on the machine shaft (63) output is transferred to QEP circuit (64) and frequency measurement circuit (65); Pulse signal is handled through QEP circuit (64) and is obtained position feed back signal; And sending the Position Control ring in the motion-control module (4) to, pulse signal is handled through frequency measurement circuit, obtains feedback speed signal; And send the rate control module in the motion-control module (4) to; Current sensor (62) detects the motor windings electric current, and obtains its digital current signal through A/D converter (61), sends it in the motion-control module (4) current regulator again.

7. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 6; It is characterized in that above-mentionedly getting in touch a closest unit for the most key the minimally-invasive surgery robot system with the patient from operator (7); Operating theater instruments is housed on it, and accomplishes the master-slave mode micro-wound surgical operation thus; Above-mentioned feedback module (8) is realized the supervision of minimal invasive surgical procedures and real-time information feedback through the graphics processing unit in endoscope and monitor and the computer control system, and making whole master-slave minimally-invasive surgery robot system is a closed-loop control system.

8. the adaptive filter device of master-slave minimally-invasive surgery robot system according to claim 3 is characterized in that aforementioned calculation machine control system (9) comprises communication unit (91), Operations Analysis (92) and graphics processing unit (93); The said Operations Analysis (92) and behavior adaptive filter device (1) two-way communication of trembling, digital-to-analogue conversion, analog digital conversion, inverse kinematics calculation operations in the behavior adaptive filter device (1) of realizing trembling, and the simple joint control unit monitored; Motion-control module (4) is realized the transmission of data through communication unit (91) and computer control system (9); Graphics processing unit (93) is accepted the image information of endoscope (81) output in the feedback module (8), this is handled and sends to the monitor (82) in the feedback module (8); Said Operations Analysis (92) is all realized by communication unit (91) with communicating by letter of other intermodules with graphics processing unit (93).

Images