CN102133103A - Method for recognizing human walking gait cycle with electromyographic signal - Google Patents

Abstract

The invention discloses a method for recognizing a human walking gait cycle with an electromyographic signal and relates to a method for measuring human limbs movements. The method for recognizing the human walking gait cycle with the electromyographic signal is characterized in that a utilized device comprises a singlechip, an electromyographic signal sensor and a three-point differential input electrocardio electrode based on a segmentation integral algorithm, wherein the electromyographic signal sensor records and processes the electromyographic signal collected by the electromyographic signal sensor to obtain a positive level unstable signal, then the singlechip transfers, moves and calculates an average filter by utilizing a peak-valley segmentation integral algorithm or a simplified algorithm of a peak-valley linear interpolation segmentation integral algorithm to obtain a human walking gait cycle, therefore, the method solves the problems of easy abrasion of a signal sensor of a human movement signal and zero drift of the electromyographic signal sensor and also overcomes the defects that the traditional gait cycle recognizing method is complex and has high equipment price and larger calculation amount. The method for recognizing the human walking gait cycle with the electromyographic signal only needs a single channel signal as an information source and improves the application universality because a plurality of muscles of a person can be selected.

Description

Method with electromyographic signal identification human body walking gait cycle

Technical field

Technical scheme of the present invention relates to the method for measuring human body limb movement, specifically discerns the method for human body walking gait cycle with electromyographic signal.

Background technology

Find through literature search, mainly adopt video detection technology for the identification of gait cycle both at home and abroad at present.Pedestrian's gait cycle extraction method in the CN101477618 video, the theory of utilizing frequency domain analysis by a kind of is disclosed, the method that the gait cycle feature of pedestrian's target in the video is extracted automatically, but in fields such as artificial limb, walking aid and rehabilitation training evaluation, obtain gait cycle information and make that the price of equipment is too high, and amount of calculation is bigger by these methods.The paper that people such as He Lesheng delivered on " data acquisition and processing " magazine in 2006 " the initial time recognition method of a kind of action " based on electromyographic signal, make referrals in the literary composition by self organizing artificial neural network to discern to moving the initial moment, this method is comparatively complicated, and exists the disconnected problem of the erroneous judgement of nonspecific action.

Prior art is analyzed and researched to the human body walking gait cycle, is as information source by the human motion signal.But because there is problem easy to wear in the signal transducer of human motion signal, the long-time use can influence recognition effect.

Summary of the invention

Technical problem to be solved by this invention is: the method with electromyographic signal identification human body walking gait cycle is provided, be based on the method for the electromyographic signal identification human body walking gait cycle of subsection integral algorithm, employing is that the signal amplitude of benchmark is analyzed with the trough amplitude, therefore the inventive method overcome in the prior art signal transducer of human motion signal exist easy to wear in, overcome the null offset problem of electromyographic signal pick off again, only need single channel signal as information source, and everyone has polylith muscle as selection, improved the popularity of using, it is too high also to have overcome existing gait cycle recognition methods complexity and equipment price, and the bigger shortcoming of amount of calculation.

The present invention solves this technical problem the technical scheme that is adopted: with the method for electromyographic signal identification human body walking gait cycle, be based on the method with electromyographic signal identification human body walking gait cycle of subsection integral algorithm, concrete steps are as follows:

The first step, device therefor and installation thereof

Device therefor comprises single-chip microcomputer, electromyographic signal pick off and electrocardioelectrode, and wherein said electrocardioelectrode is the differential input electrode of bikini, constitutes by positive input electrode, negative input electrode with reference to the earth polar; Connect with lead between single-chip microcomputer, electromyographic signal pick off and the electrocardioelectrode; Electrocardioelectrode is affixed on the muscle surface of human body shank, be attached to the belly of muscle place with its positive input electrode and negative input electrode place straight line along the direction placement of the meat fiber of human body shank, then be attached to and the impartial place of the distance of positive input electrode and negative input electrode with reference to the earth polar, the positive input electrode, negative input electrode and all equal with reference to the distance of any two electrodes in the earth polar, the electromyographic signal pick off is placed on the electrocardioelectrode, link together with the electromyographic signal pick off by three electrodes of conduction button electrocardioelectrode, by lead the electromyographic signal pick off is connected with single-chip microcomputer again, and single-chip microcomputer is attached on artificial limb or the walking aid;

Second step, the collection of signal and processing

On the basis of the first step, note and carry out the preposition amplification of low noise, high-pass filtering, the filtering of 50Hz power frequency, variable gain amplification, low-pass filtering and true rms circuit and handle with the positive input electrode of electrocardioelectrode, negative input electrode with reference to the electromyographic signal that gather in the earth polar by the electromyographic signal pick off, finally obtain the positive level non-stationary signal;

The 3rd step, the algorithm computation of identification gait cycle

Carry out A/D conversion, rolling average filtering and adopt the subsection integral algorithm to calculate by the I/O mouth of the second positive level non-stationary signal that obtain of step, finally obtain the human body walking gait cycle by above-mentioned single-chip microcomputer;

Main program flow by single-chip microcomputer identification gait cycle is:

Beginning → I/O mouth initialization → algorithm process parameter initialization → intervalometer initialization → finish?: be → finish; Deny → enter low-power consumption, enable total interruption → return and finish?

Interrupt routine flow process by single-chip microcomputer identification gait cycle is:

Enter interrupt → keep the scene intact → sample, filtering → signal calculated amplitude change rate k ₁→ according to k ₁And k ₂, crest/trough? → trough then writes down this moment t _tAnd the moment at quarter signal amplitude v _tCrest then writes down this moment t _pAnd the moment at quarter signal amplitude v _pBe not → does n add 1 → n equals T* trend cycle for trend counting?: deny → then go to this rate of change of record k ₂=k ₁With this sampling period signal amplitude of record; Be → calculating trend amplitude change rate K ₁→ according to k ₁And k ₂, crest/trough?: 1. trough → then write down t _k=t _t, v ₁=v _t2. crest → then calculate t _kMoment eigenvalue S (t _k) → S (t _k)＞threshold value?: deny → go to this trend amplitude change rate of record K ₂=K ₁Be → gait cycle T _k=t _k-t _K-1Count this trend amplitude change rate of k=k+1 → record K with gait cycle ₂=K ₁3. be not → this trend amplitude change rate of record K ₂=K ₁This rate of change of → record k ₂=k ₁This sampling period signal amplitude → interruption is returned with record;

The algorithm computation method of signal is as follows:

Adopt the subsection integral algorithm, comprise the positive eigenvalue S of wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)={\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_k\right))\mathrm{dt}=\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_k\right)),$ If f is (t _K+1)＞f (t _k) and t=t _k(1)

Described t _kBeing meant the moment of each Wave crest and wave trough, is unit with the sampling period, and Wave crest and wave trough is adjacent, if i.e. t ₁Be trough, t then ₂Be crest, t ₃Be trough, t then ₄Be crest, by that analogy; Described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)=-{\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_{k+1}\right))\mathrm{dt}=0-\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_{k+1}\right)),$ If f is (t _K+1)＜f (t _k) and t=t _k(2)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S (t)=0 constantly, the eigenvalue that guarantees all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

Above-mentioned employing subsection integral algorithm is to adopt peak-paddy subsection integral algorithm or adopt peak-valley line interpolation subsection integral algorithm when implementing, and specific algorithm is as follows:

Adopting peak-paddy subsection integral algorithm, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, and signal is from trough t ₁To crest t ₂The integrated value S of amplitude ₁, gained crest t ₂Eigenvalue is with trough t ₃Be benchmark, signal is from crest t ₂To trough t ₃The integrated value S of amplitude ₂, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)={\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_k\right))\mathrm{dt}=\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_k\right)),$ If f is (t _K+1)＞f (t _k) and t=t _k(1)

Described t _kBeing meant the moment of each Wave crest and wave trough, is unit with the sampling period, and Wave crest and wave trough is adjacent, if i.e. t ₁Be trough, t then ₂Be crest, t ₃Be trough, t then ₄Be crest, by that analogy; Described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)=-{\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_{k+1}\right))\mathrm{dt}=0-\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_{k+1}\right)),$ If f is (t _K+1)＜f (t _k) and t=t _k(2)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S (t)=0 constantly, the eigenvalue that guarantees all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0;

Adopting peak-valley line interpolation subsection integral algorithm, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, at trough t ₁With crest t ₂Between linear interpolation, trough t ₁To crest t ₂The integrated value S of amplitude after the interpolation ₃, gained crest t ₂Eigenvalue is with trough t ₃Amplitude is a benchmark, at crest t ₂With trough t ₃Between linear interpolation, crest t ₂To trough t ₃The integrated value S of amplitude after the interpolation ₄, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₃, i.e. the negative eigenvalue S of tracer signal rising variation and crest location ₄, promptly tracer signal descends and changes two parts; Described S ₃Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, at t ₁To t ₂Linear interpolation constantly is from t ₁To t ₂The integrated value of interpolated signal constantly, described S ₄Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, at t ₃To t ₄Linear interpolation constantly is from t ₃To t ₄The integrated value of interpolated signal constantly, its computing formula is as follows:

$S^{\′}\left(t\right)=\frac{1}{2}(f\left(t_{k+1}\right)-f\left(t_k\right))\×(t_{k+1}-t_k)/T,$ If t=t _k(3)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S ' constantly (t)=0, the eigenvalue that so both can guarantee all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

The English of above-mentioned peak-paddy subsection integral algorithm is Peak-valley piecewise integrator, is called for short PVPI; The English of peak-valley line interpolation subsection integral algorithm is Peak-valley linear interpolation piecewise integrator, is called for short PVLI﹠amp; PI.

Above-mentioned method with electromyographic signal identification human body walking gait cycle, wherein the model of used single-chip microcomputer is that the MSP430 serial model No. is F2274, the model of electromyographic signal pick off is MyoScan-Pro, electrocardioelectrode is that these equipment all can be commercially available with the differential input electrocardioelectrode of disposable bikini of the Signa Gel that do not dry up.

The invention has the beneficial effects as follows:

(1) the present invention with the principle of the method for electromyographic signal identification human body walking gait cycle is:

Electromyographic signal is that the electricity that the central nervous system follows when arranging musculation changes.It is a kind of important method that detects musculation at the body surface noinvasive that electromyographic signal is handled.Studies show that there are necessary relation in human body walking leg speed and musculation, along with the difference of human body walking leg speed, the electromyographic signal amplitude has significant change.Because there are inverse relation in human body walking gait cycle and human body walking leg speed, so also there are necessary connection in human body walking gait cycle and lower limb muscles action.The human body walking gait cycle is analyzed and researched as information source with electromyographic signal, can not have signal transducer problem easy to wear, and it has directly reflected people's activity consciousness, expressed human body walking step state information more exactly than human motion signal.

(2) the present invention with the marked improvement of the method for electromyographic signal identification human body walking gait cycle is:

1) the inventive method by the heel contact of a tested side before the initial moment of muscle movement the human body walking gait cycle is divided, the peak-paddy subsection integral algorithm and the peak-valley line interpolation subsection integral algorithm in the initial moment of identification maneuver proposed.Judge the crest and the trough moment of extracting signal according to trend, extraction gait cycle division points combines with threshold method, obtain the human body walking gait cycle, when this algorithm has been avoided the null offset phenomenon of electromyographic signal pick off, improved the discrimination of gait cycle, only need single channel signal as information source, and different people there is polylith muscle available, increased the popularity that this method is used.

What 2) the inventive method can be easier identifies the human body walking gait cycle, and the signal source that is adopted is electromyographic signal, and it can more directly express the human motion intention.Main by single pass electromyographic signal as signal source, divide the human body walking gait cycle by initial moment of discerning muscle movement, and among the present invention designed algorithm can be more reliably, accurately the initial moment of muscle movement is identified.

3) the electromyographic signal pick off is placed on the electrocardioelectrode, link together with the electromyographic signal pick off by three electrodes of conduction button electrocardioelectrode, by lead the electromyographic signal pick off is connected with single-chip microcomputer again, and single-chip microcomputer is attached on artificial limb or the walking aid.The electromyographic signal pick off is attached on the human muscle securely like this, guaranteed the stability of signals collecting, the installation of all devices is simple, firm and portable, both overcome that there is problem easy to wear in the signal transducer of human motion signal in the prior art, it is too high to have overcome existing gait cycle recognition methods complexity and equipment price again, and uses inconvenient shortcoming.

4) the inventive method can be applied to fields such as artificial limb, walking aid and rehabilitation training evaluation.

Description of drawings

The present invention is further described below in conjunction with drawings and Examples.

Fig. 1 is that the electromyographic signal trend of the inventive method is judged sketch map.

Fig. 2 is the peak-paddy subsection integral algorithm principle sketch map of the inventive method.

Fig. 3 is the peak-valley line interpolation subsection integral algorithm principle sketch map of the inventive method.

Fig. 4 is the peak-initial moment sketch map of paddy subsection integral algorithm identified muscle movement of the inventive method.

Fig. 5 is the peak-initial moment sketch map of valley line interpolation subsection integral algorithm identified muscle movement of the inventive method.

Fig. 6 is the peak-paddy subsection integral algorithm identified human body walking gait cycle sketch map of the inventive method.

Fig. 7 is the peak-valley line interpolation subsection integral algorithm identified human body walking gait cycle sketch map of the inventive method.

Fig. 8 is the main program flow block diagram of the inventive method by single-chip microcomputer identification gait cycle.

Fig. 9 is the interrupt routine FB(flow block) of the inventive method by single-chip microcomputer identification gait cycle.

Figure 10 connects and schematic flow sheet for the inventive method device therefor.

Among the figure, 1. the positive input electrode of electrocardioelectrode, the 2. negative input electrode of electrocardioelectrode, 3. the reference earth polar of electrocardioelectrode.

The specific embodiment

Figure 10 shows that device therefor of the present invention comprises single-chip microcomputer, electromyographic signal pick off and electrocardioelectrode.Wherein used single-chip microcomputer is the MSP430F2274 single-chip microcomputer, the electromyographic signal pick off is the MyoScan-Pro electromyographic signal pick off of Thought Technology company, electrocardioelectrode is the differential input electrocardioelectrode of disposable bikini with the Signa Gel that do not dry up, this electrocardioelectrode is affixed on the muscle surface of human body shank, be attached to the belly of muscle place with positive input electrode 1 and electric 2 utmost point place straight lines of negative input along the direction placement of the meat fiber of human body shank, then 3 be attached to and the impartial place of the distance of positive input electrode and negative input electrode, positive input electrode 1 with reference to the earth polar, negative input electrode 2 and all equal with reference to the distance of any two electrodes in the earth polar 3; Electromyographic signal by above-mentioned electrocardioelectrode collection is passed to MyoScan-Pro electromyographic signal pick off, MyoScan-Pro electromyographic signal pick off carries out the preposition amplification of low noise, high-pass filtering, the filtering of 50Hz power frequency, variable gain amplification, low-pass filtering and true rms circuit to the electromyographic signal of gathering to be handled, and finally obtains the positive level non-stationary signal; The positive level non-stationary signal carries out A/D conversion, rolling average filtering and peak-paddy subsection integral algorithm algorithm (PVPI algorithm) or peak-valley line interpolation subsection integral algorithm (PVLI﹠amp by the I/O mouth of MSP430F2274 single-chip microcomputer; PI) calculate, finally obtain the human body walking gait cycle.

Fig. 1 shows that the algorithm of the inventive method at first identifies by the initial moment of trend judgment mode with all muscle movements, and abscissa is a time shaft, and unit is second, and vertical coordinate is the amplitude of signal, and unit is mv, and described trend judges that the moment comprises q ₁, q ₂, q ₃, q ₄, q ₅, q ₆, trend rate of change computing formula is K _N=(f (t)-f (t-T ^*))/T ^*, described K _NBe the trend rate of change of N trend judging point, f (t) sequence is that myoelectricity discrete digital signal is through filtered result, T ^*For trend is judged the cycle, the identification principle is:

If 1. a＜=K _N＜=b, then K _N=K _N-1

If 2. K _N＞0, and K _N-1＜0, then trend begins to rise;

If 3. K _N＜0, and K _N-1＞0, then trend begins to descend;

Described a is a trend rate of change lower limit, and b is the trend rate of change upper limit; According to the non-stationary property of described electromyographic signal, at recognition objective trough t ₁The time (q ₂Constantly) less sampling period T of employing cycle ^*=T ' is when identifying target trough t ₁(the initial moment t of muscle movement _Up) back (q ₂To q ₅Constantly), will the sampling period become T ^*=T " (T "＞T ') recognition objective crest t ₂, when identifying crest t ₂Back (q ₅Afterwards), will the sampling period be reduced to T again ^*=T ' recognition objective trough t ₃, so circulation identification; Described crest and trough are according to the variation k of signal _n=f (t)-f (t-T) judges that T is the electromyographic signal sampling period, and judgment principle is:

If 1. k _n＞=0, and k _N-1＜0, should be the signal trough constantly then;

If 2. k _n＜=0, and k _N-1＞0, should be signal wave crest constantly then;

Described t ₁, t ₃, t ₅Be signal f (t) trough after the filtering, t constantly ₂, t ₄Be signal f (t) crest after the filtering constantly; Trend determines q ₂During point, signal trend begins to rise, then Zhi Qian t ₁Be target trough (the initial moment of muscle movement), trend determines q ₅During point, signal trend begins to descend, then Zhi Qian t ₂Be target crest, cycle criterion successively; Trend is judged at q ₃To q ₄The moment is what rise, then crest t ₄, trough t ₅Constantly all ignore and do not consider, be considered as the interference characteristic point.

Fig. 2 shows the peak-paddy subsection integral algorithm in the inventive method, comprises the positive eigenvalue S of wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts, abscissa is the time, and unit is second, and the main vertical coordinate on the left side is the amplitude f (t) of electromyographic signal, and unit is mv, the inferior vertical coordinate on the right is the amplitude S (t) of eigenvalue; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, i.e. S among the figure ₁The area value of part, described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, i.e. S among the figure ₂The negative value of part area value.

Fig. 3 shows the peak-valley line interpolation subsection integral algorithm in the inventive method, and it is the shortcut calculation of peak-paddy subsection integral algorithm, comprises the positive eigenvalue S of wave trough position ₃, i.e. the negative eigenvalue S of tracer signal rising variation and crest location ₄, promptly tracer signal descend to change two parts, abscissa is the time, unit be second, the main vertical coordinate on the left side is the amplitude f (t) of electromyographic signal, unit is mv, the inferior vertical coordinate on the right be eigenvalue amplitude S ' (t); Described S ₃Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, at t ₁To t ₂Linear interpolation constantly is from t ₁To t ₂The integrated value of interpolated signal constantly, the i.e. S that chain-dotted line surrounded among the figure ₃The triangle area value so will be converted into sampling number the time, be beneficial to single-chip microcomputer and handle described S ₄Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, at t ₃To t ₄Linear interpolation constantly is from t ₃To t ₄The integrated value of interpolated signal constantly, the i.e. S that chain-dotted line surrounded among the figure ₄The triangle area value.

Fig. 4 shows that the peak-paddy subsection integral algorithm identified initial moment of muscle movement in the inventive method only comprises positive eigenvalue S ₁, promptly tracer signal uphill process and eigenvalue are 0 two parts, and abscissa is the time, and unit is second, and the main vertical coordinate on the left side is the amplitude f (t) of electromyographic signal, unit is mv, the inferior vertical coordinate on the right is the amplitude S of eigenvalue " (t); Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, i.e. S among the figure ₁The area value of part,

Fig. 5 shows that the peak-valley line interpolation subsection integral algorithm identified initial moment of muscle movement in the inventive method only comprises positive eigenvalue S ₃, promptly tracer signal uphill process and eigenvalue are 0 two parts, and abscissa is the time, and unit is second, and the main vertical coordinate on the left side is the amplitude f (t) of electromyographic signal, unit is mv, the inferior vertical coordinate on the right is the amplitude S ' of eigenvalue " (t); Described S ₃Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, at t ₁To t ₂Linear interpolation constantly is from t ₁To t ₂The integrated value of interpolated signal constantly, the i.e. S that chain-dotted line surrounded among the figure ₃The triangle area value so will be converted into sampling number the time, be beneficial to single-chip microcomputer and handle.

Peak-paddy subsection integral algorithm that Fig. 6 shows the inventive method is chosen appropriate threshold in the identification initial moment of muscle movement to its eigenvalue, and the initial moment of the muscle movement that eigenvalue is bigger in each cycle is identified; Described gait cycle division points t _UpBe the bigger trough moment t of eigenvalue ₁, t ₄, t ₅, t ₆, the time difference between then adjacent two gait cycle division points is described gait cycle, makes t _Upk=t ₄, t _Upk+1=t ₅, gait cycle T then _Cyclek=t _Upk+1-t _UpkDescribed crest t ₇Cause for the muscle that occurs in support phase to the shaking peroid transitional processes moves by a small margin, its amplitude is less, it can be ignored by choosing appropriate threshold; Described interference characteristic point trough is t constantly ₈With crest moment t ₉Because the uncertain action of muscle causes among the human walking procedure, because crest t ₉Amplitude is less, and trough t ₈And the time between the crest t9 is shorter, so t ₈Eigenvalue is less constantly, this interference can be ignored by choosing appropriate threshold.

Peak-valley line interpolation subsection integral algorithm that Fig. 7 shows the inventive method is chosen appropriate threshold in the identification initial moment of muscle movement to its eigenvalue, and the initial moment of the muscle movement that eigenvalue is bigger in each cycle is identified; Described gait cycle division points t _UpBe the bigger trough moment t of eigenvalue ₁, t ₄, t ₅, t ₆, the time difference between then adjacent two gait cycle division points is described gait cycle, makes t _Upk=t ₄, t _Upk+1=t ₅, gait cycle T then _Cyclek=t _Upk+1-t _UpkDescribed crest t ₇Cause for the muscle that occurs in support phase to the shaking peroid transitional processes moves by a small margin, its amplitude is less, it can be ignored by choosing appropriate threshold; Described interference characteristic point trough is t constantly ₈With crest moment t ₉Because the uncertain action of muscle causes among the human walking procedure, because crest t ₉Amplitude is less, and trough t ₈And the time between the crest t9 is shorter, so t ₈Eigenvalue is less constantly, this interference can be ignored by choosing appropriate threshold.

Reference Fig. 6 and 7 as seen, the peak of the inventive method-paddy subsection integral algorithm and peak-valley line interpolation subsection integral algorithm, the methods in these the two kinds identification initial moment of muscle movement, it comes to the same thing, different is described trough is the eigenvalue size constantly; The t of peak-paddy subsection integral algorithm identified ₆, t ₈Eigenvalue is smaller than the result of peak-valley line interpolation subsection integral algorithm identified constantly, but the range that threshold value is chosen is little, does not therefore influence recognition result; The threshold value of described eigenvalue is by the mass data The result of statistics, and the selection of threshold value does not possess uniqueness, but all desirable in a scope, makes the reliability of recognition result increase.

In the above-mentioned subsection integral algorithm, described gait cycle division points obtains by eigenvalue is chosen appropriate threshold.

Fig. 8 shows that the inventive method by the main program flow of single-chip microcomputer identification gait cycle is:

Beginning → I/O mouth initialization → algorithm process parameter initialization → intervalometer initialization → finish?: be → finish; Deny → enter low-power consumption, enable total interruption → return and finish?

The initialization of described I/O mouth comprises input pattern initialization, the initialization of AD transition enabled of setting the I/O mouth; Described algorithm process parameter initialization comprise peak-paddy subsection integral algorithm algorithm or peak-valley line interpolation subsection integral algorithm when processing signals designed to filtering cycle size, signal amplitude rate of change threshold value, signal trend judgement cycle choose, trend rate of change threshold value is chosen; Described intervalometer initialization is used for regularly sampling, thereby can guarantee the concordance in sampling period; Describedly entering low-power consumption, enabling total the interruption is an endless loop program, is used for waiting for regularly interrupting taking place, and does not enter the state that is in low-power consumption before the interruption.

Fig. 9 shows that the inventive method by the interrupt routine flow process of single-chip microcomputer identification gait cycle is:

Enter interrupt → keep the scene intact → sample, filtering → signal calculated amplitude change rate k ₁→ according to k ₁And k ₂, crest/trough? → trough then writes down this moment t _tAnd the moment at quarter signal amplitude v _tCrest then writes down this moment t _pAnd the moment at quarter signal amplitude v _pBe not → trend counting n adds 1 → n and whether equals T trend cycle ^*: deny → then go to this rate of change of record k ₂=k ₁With this sampling period signal amplitude of record; Be → calculating trend amplitude change rate K ₁→ according to k ₁And k ₂, crest/trough?: 1. trough → then write down t _k=t _t, v ₁=v _t2. crest → then calculate t _kMoment eigenvalue S (t _k) → S (t _k)＞threshold value?: deny → go to this trend amplitude change rate of record K ₂=K ₁Be → gait cycle T _k=t _k-t _K-1Count this trend amplitude change rate of k=k+1 → record K with gait cycle ₂=K ₁3. be not → this trend amplitude change rate of record K ₂=K ₁This rate of change of → record k ₂=k ₁This sampling period signal amplitude → interruption is returned with record;

Described keeping the scene intact is that depositor or variable to interrupting in the future preface stored, and change occurs in order to avoid EOI returns the back; Described sampling, filtering is meant by the I/O mouth signal gathered, and carries out the A/D conversion, and analog quantity is converted to digital quantity for microprocessor processes, and the digital signal of microprocessor after to conversion carries out rolling average filtering; Described signal calculated amplitude change rate k ₁Be according to formula k _n=f (t)-f (t-T) calculates and to get, and as previously described method according to k ₁, a last sampled value amplitude change rate k ₂Judge that this is crest, trough or the non-Wave crest and wave trough moment constantly, be the crest or the trough moment constantly when determining this, if trough, then should moment signal amplitude v _tAnd this moment time counting value t _tNote, if crest, then should moment signal amplitude v _pAnd this moment time counting value t _pNote, constantly then directly carry out next step if not crest or trough; Above computational process is that signal amplitude changes judgement.It is to be used for sampling number is carried out accumulation calculating that described trend counting n adds 1, thereby the trend that can guarantee to be separated by is judged period T ^*The place carries out trend and judges; Described trend counting n reaches T ^*The time, promptly can carry out trend and judge described calculating trend amplitude change rate K ₁According to formula K _N=(f (t)-f (t-T ^*))/T ^*Calculate, be used for writing down the trend amplitude change rate that this trend is judged, and as previously described method according to K ₁And last one trend cycle amplitude change rate K ₂Judge that this is trend crest, trough or the non-Wave crest and wave trough moment constantly, if trough is constantly, then should moment time counting value t _tAssignment is given trough register value t _k, write down this moment signal amplitude v ₁=v _t, if crest constantly, then according to foregoing peak-paddy subsection integral algorithm algorithm or peak-valley line interpolation subsection integral algorithm computation trough moment t _kEigenvalue S (t _k) and compare with threshold value, if greater than threshold value then calculate gait cycle, i.e. gait cycle T by the difference of adjacent twice trough register value _k=t _k-t _K-1, and gait cycle count value k added 1, then directly carry out next step if not trend crest or trough; This trend amplitude change rate of described record K ₂=K ₁Be to judge for next trend to prepare; More than judge flow process for signal trend.This rate of change of described record k ₂=k ₁Be for judging that next sampled value is that the crest or the trough of signal prepared, this sampling period signal amplitude of described record is to prepare for next sampling period signal calculated amplitude change rate; Finally finish above computational process, then interrupt returning.

Embodiment 1

The first step, device therefor and installation thereof

By shown in Figure 10, to be affixed on the muscle surface of human body shank with the differential input electrocardioelectrode of disposable bikini of the Signa Gel that do not dry up, be attached to the belly of muscle place with positive input electrode 1 and electric 2 utmost point place straight lines of negative input along the direction placement of the meat fiber of human body shank, then 3 be attached to and the impartial place of distance of positive input electrode and negative input electrode positive input electrode 1, negative input electrode 2 and all equate with reference to the distance of any two electrodes in the earth polar 3 with reference to the earth polar; The MyoScan-Pro electromyographic signal pick off of Thought Technology company is placed on the above-mentioned electrocardioelectrode, link together with MyoScan-Pro electromyographic signal pick off by three electrodes of conduction button above-mentioned electrocardioelectrode, by lead MyoScan-Pro electromyographic signal pick off is connected with the MSP430F2274 single-chip microcomputer again, and the MSP430F2274 single-chip microcomputer is attached on artificial limb or the walking aid.

Second step, the collection of signal and processing

After the test beginning, the electromyographic signal of being gathered by above-mentioned electrocardioelectrode is delivered to MyoScan-Pro electromyographic signal pick off, MyoScan-Pro electromyographic signal pick off carries out the preposition amplification of low noise, high-pass filtering, the filtering of 50Hz power frequency, variable gain amplification, low-pass filtering and true rms circuit to the electromyographic signal of gathering to be handled, and finally obtains the positive level non-stationary signal;

The 3rd step, the algorithm computation of identification gait cycle

Carry out A/D conversion, rolling average filtering and peak-paddy subsection integral algorithm algorithm (PVPI algorithm) by the I/O mouth of the second positive level non-stationary signal that obtain of step by the MSP430F2274 single-chip microcomputer, finally obtain the human body walking gait cycle.

Main program flow by single-chip microcomputer identification gait cycle is:

Beginning → I/O mouth initialization → algorithm process parameter initialization → intervalometer initialization → finish? → be to finish; Not, enter low-power consumption, enable total interruption → finish? → be to finish; Not, enter low-power consumption, enable total interruption → finish?

Interrupt routine flow process by single-chip microcomputer identification gait cycle is:

Enter interrupt → keep the scene intact → sample, filtering → signal calculated amplitude change rate k ₁→ according to k ₁And k ₂, crest/trough? → trough then writes down this moment t _tAnd the moment at quarter signal amplitude v _tCrest then writes down this moment t _pAnd the moment at quarter signal amplitude v _pBe not → trend counting n adds 1 → n and whether equals T trend cycle ^*→ not, then go to this rate of change of record k ₂=k ₁With this sampling period signal amplitude of record; Be to calculate trend amplitude change rate K ₁→ according to k ₁And k ₂, crest/trough? → 1. trough then writes down t _k=t _t, v ₁=v _t2. crest then calculates t _kMoment eigenvalue S (t _k) → S (t _k)＞threshold value → not, go to this trend amplitude change rate of record K ₂=K ₁Be gait cycle T _k=t _k-t _K-1Counting k=k+1 with gait cycle 3. is not → this trend amplitude change rate of record K ₂=K ₁This rate of change of → record k ₂=k ₁This sampling period signal amplitude → interruption is returned with record;

The algorithm computation method of signal is as follows:

Adopting peak-paddy subsection integral algorithm, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, and signal is from trough t ₁To crest t ₂The integrated value S of amplitude ₁, gained crest t ₂Eigenvalue is with trough t ₃Be benchmark, signal is from crest t ₂To trough t ₃The integrated value S of amplitude ₂, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)={\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_k\right))\mathrm{dt}=\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_k\right)),$ If f is (t _K+1)＞f (t _k) and t=t _k(1)

Described t _kBeing meant the moment of each Wave crest and wave trough, is unit with the sampling period, and Wave crest and wave trough is adjacent, if i.e. t ₁Be trough, t then ₂Be crest, t ₃Be trough, t then ₄Be crest, by that analogy; Described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, i.e. S among Fig. 2 ₂The negative value of part area value, its computing formula is as follows:

$S\left(t\right)=-{\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_{k+1}\right))\mathrm{dt}=0-\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_{k+1}\right)),$ If f is (t _K+1)＜f (t _k) and t=t _k(2)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S (t)=0 constantly, the eigenvalue that guarantees all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0;

Embodiment 2

Computational methods peak-valley line interpolation subsection integral algorithm (PVLI﹠amp except signal; PI) calculate outside, other step method are all with embodiment 1.

Peak-valley line interpolation subsection integral algorithm is stated in employing, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, at trough t ₁With crest t ₂Between linear interpolation, trough t ₁To crest t ₂The integrated value S of amplitude after the interpolation ₃, gained crest t ₂Eigenvalue is with trough t ₃Amplitude is a benchmark, at crest t ₂With trough t ₃Between linear interpolation, crest t ₂To trough t ₃The integrated value S of amplitude after the interpolation ₄, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₃, i.e. the negative eigenvalue S of tracer signal rising variation and crest location ₄, promptly tracer signal descends and changes two parts; Described S ₃Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, at t ₁To t ₂Linear interpolation constantly is from t ₁To t ₂The integrated value of interpolated signal constantly, i.e. S among Fig. 3 ₃The triangle area value will be converted into sampling number the time, be beneficial to single-chip microcomputer and handle described S ₄Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, at t ₃To t ₄Linear interpolation constantly is from t ₃To t ₄The integrated value of interpolated signal constantly, i.e. S among the figure ₄The triangle area value will be converted into sampling number the time, be beneficial to single-chip microcomputer and handle, and its computing formula is as follows:

$S^{\′}\left(t\right)=\frac{1}{2}(f\left(t_{k+1}\right)-f\left(t_k\right))\×(t_{k+1}-t_k)/T,$ If t=t _k(3)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S ' constantly (t)=0, the eigenvalue that so both can guarantee all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

In the foregoing description, the model of used single-chip microcomputer is that the MSP430 serial model No. is F2274, the model of electromyographic signal pick off is MyoScan-Pro, and electrocardioelectrode is that these equipment all can be commercially available with the differential input electrocardioelectrode of disposable bikini of the Signa Gel that do not dry up.

Claims (2)

1. with the method for electromyographic signal identification human body walking gait cycle, it is characterized in that: be based on the method with electromyographic signal identification human body walking gait cycle of subsection integral algorithm, concrete steps are as follows:

The first step, device therefor and installation thereof

Device therefor comprises single-chip microcomputer, electromyographic signal pick off and electrocardioelectrode, and wherein said electrocardioelectrode is the differential input electrode of bikini, constitutes by positive input electrode, negative input electrode with reference to the earth polar; Connect with lead between single-chip microcomputer, electromyographic signal pick off and the electrocardioelectrode; Electrocardioelectrode is affixed on the muscle surface of human body shank, be attached to the belly of muscle place with its positive input electrode and negative input electrode place straight line along the direction placement of the meat fiber of human body shank, then be attached to and the impartial place of the distance of positive input electrode and negative input electrode with reference to the earth polar, the positive input electrode, negative input electrode and all equal with reference to the distance of any two electrodes in the earth polar, the electromyographic signal pick off is placed on the electrocardioelectrode, link together with the electromyographic signal pick off by three electrodes of conduction button electrocardioelectrode, by lead the electromyographic signal pick off is connected with single-chip microcomputer again, and single-chip microcomputer is attached on artificial limb or the walking aid;

Second step, the collection of signal and processing

On the basis of the first step, note and carry out the preposition amplification of low noise, high-pass filtering, the filtering of 50Hz power frequency, variable gain amplification, low-pass filtering and true rms circuit and handle with the positive input electrode of electrocardioelectrode, negative input electrode with reference to the electromyographic signal that gather in the earth polar by the electromyographic signal pick off, finally obtain the positive level non-stationary signal;

The 3rd step, the algorithm computation of identification gait cycle

Carry out A/D conversion, rolling average filtering and adopt the subsection integral algorithm to calculate by the I/O mouth of the second positive level non-stationary signal that obtain of step, finally obtain the human body walking gait cycle by above-mentioned single-chip microcomputer;

Main program flow by single-chip microcomputer identification gait cycle is:

Beginning → I/O mouth initialization → algorithm process parameter initialization → intervalometer initialization → finish?: be → finish; Deny → enter low-power consumption, enable total interruption → return and finish?

Interrupt routine flow process by single-chip microcomputer identification gait cycle is:

Enter interrupt → keep the scene intact → sample, filtering → signal calculated amplitude change rate k ₁→ according to k ₁And k ₂, crest/trough? → trough then writes down this moment t _tAnd the moment at quarter signal amplitude v _tCrest then writes down this moment t _pAnd the moment at quarter signal amplitude v _pBe not → does n add 1 → n equals T* trend cycle for trend counting?: deny → then go to this rate of change of record k ₂=k ₁With this sampling period signal amplitude of record; Be → calculating trend amplitude change rate K ₁→ according to k ₁And k ₂, crest/trough?: 1. trough → then write down t _k=t _t, v ₁=v _t2. crest → then calculate t _kMoment eigenvalue S (t _k) → S (t _k)＞threshold value?: deny → go to this trend amplitude change rate of record K ₂=K ₁Be → gait cycle T _k=t _k-t _K-1Count this trend amplitude change rate of k=k+1 → record K with gait cycle ₂=K ₁3. be not → this trend amplitude change rate of record K ₂=K ₁This rate of change of → record k ₂=k ₁This sampling period signal amplitude → interruption is returned with record;

The computational methods of signal are as follows:

Adopt the subsection integral algorithm, comprise the positive eigenvalue S of wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)={\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_k\right))\mathrm{dt}=\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_k\right)),$ If f is (t _K+1)＞f (t _k) and t=t _k(1)

Described t _kBeing meant the moment of each Wave crest and wave trough, is unit with the sampling period, and Wave crest and wave trough is adjacent, if i.e. t ₁Be trough, t then ₂Be crest, t ₃Be trough, t then ₄Be crest, by that analogy; Described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)=-{\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_{k+1}\right))\mathrm{dt}=0-\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_{k+1}\right)),$ If f is (t _K+1)＜f (t _k) and t=t _k(2)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S (t)=0 constantly, the eigenvalue that guarantees all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

2. the described method of claim 1 with electromyographic signal identification human body walking gait cycle, it is characterized in that: above-mentioned employing subsection integral algorithm, be to adopt peak-paddy subsection integral algorithm or adopt peak-valley line interpolation subsection integral algorithm when implementing, specific algorithm is as follows:

Adopting peak-paddy subsection integral algorithm, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, and signal is from trough t ₁To crest t ₂The integrated value S of amplitude ₁, gained crest t ₂Eigenvalue is with trough t ₃Be benchmark, signal is from crest t ₂To trough t ₃The integrated value S of amplitude ₂, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₁, i.e. the negative eigenvalue S of tracer signal uphill process and crest location ₂, i.e. tracer signal decline process two parts; Described S ₁Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, from t ₁To t ₂The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)={\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_k\right))\mathrm{dt}=\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_k\right)),$ If f is (t _K+1)＞f (t _k) and t=t _k(1)

Described t _kBeing meant the moment of each Wave crest and wave trough, is unit with the sampling period, and Wave crest and wave trough is adjacent, if i.e. t ₁Be trough, t then ₂Be crest, t ₃Be trough, t then ₄Be crest, by that analogy; Described S ₂Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, from t ₂To t ₃The integrated value of moment signal, its computing formula is as follows:

$S\left(t\right)=-{\∫}_{t_k}^{t_{k+1}}(f\left(t\right)-f\left(t_{k+1}\right))\mathrm{dt}=0-\underset{t=t_k}{\overset{t_{k+1}}{\mathrm{\Σ}}}(f\left(t\right)-f\left(t_{k+1}\right)),$ If f is (t _K+1)＜f (t _k) and t=t _k(2)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S (t)=0 constantly, the eigenvalue that guarantees all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

Adopting peak-valley line interpolation subsection integral algorithm, is with gained trough t ₁Eigenvalue is with trough t ₁Amplitude is a benchmark, at trough t ₁With crest t ₂Between linear interpolation, trough t ₁To crest t ₂The integrated value S of amplitude after the interpolation ₃, gained crest t ₂Eigenvalue is with trough t ₃Amplitude is a benchmark, at crest t ₂With trough t ₃Between linear interpolation, crest t ₂To trough t ₃The integrated value S of amplitude after the interpolation ₄, being negative value, detailed process is:

The positive eigenvalue S that comprises wave trough position ₃, i.e. the negative eigenvalue S of tracer signal rising variation and crest location ₄, promptly tracer signal descends and changes two parts; Described S ₃Be with trough t ₁Signal amplitude f (t constantly ₁) be benchmark, at t ₁To t ₂Linear interpolation constantly is from t ₁To t ₂The integrated value of interpolated signal constantly, described S ₄Be with trough t ₃Signal amplitude f (t constantly ₃) be benchmark, at t ₃To t ₄Linear interpolation constantly is from t ₃To t ₄The integrated value of interpolated signal constantly, its computing formula is as follows:

$S^{\′}\left(t\right)=\frac{1}{2}(f\left(t_{k+1}\right)-f\left(t_k\right))\×(t_{k+1}-t_k)/T,$ If t=t _k(3)

Described t ≠ t _kNon-Wave crest and wave trough eigenvalue S ' constantly (t)=0, the eigenvalue that so both can guarantee all wave trough position is all greater than 0, and the eigenvalue of crest location is all less than 0.

Images