CN106959756A - The sky mouse recognition methods acted based on electromyographic signal monitoring and device - Google Patents
The sky mouse recognition methods acted based on electromyographic signal monitoring and device Download PDFInfo
- Publication number
- CN106959756A CN106959756A CN201710193046.3A CN201710193046A CN106959756A CN 106959756 A CN106959756 A CN 106959756A CN 201710193046 A CN201710193046 A CN 201710193046A CN 106959756 A CN106959756 A CN 106959756A
- Authority
- CN
- China
- Prior art keywords
- electromyographic signal
- sky mouse
- judgment
- rule
- eigenvalue
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/015—Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
- G06F2203/01—Indexing scheme relating to G06F3/01
- G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Neurosurgery (AREA)
- General Health & Medical Sciences (AREA)
- Neurology (AREA)
- Health & Medical Sciences (AREA)
- Dermatology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The invention discloses a kind of sky mouse recognition methods acted based on electromyographic signal monitoring and device, the method comprising the steps of：Obtain the First Eigenvalue of the first limb action；When the First Eigenvalue meets default first Rule of judgment, sky mouse pattern is opened；Under sky mouse pattern, the Second Eigenvalue for the electromyographic signal that the second limb action of monitoring is produced；When Second Eigenvalue meets default second Rule of judgment, the second limb action is recognized.The present invention based on electromyographic signal monitoring act sky mouse recognition methods and device can recognize that specifically press or lift wait the second limb action so that realize simulate sky mouse function, improve Consumer's Experience.
Description
Technical field
The present invention relates to IED technical field, more particularly to IED based on electromyographic signal
The sky mouse recognition methods of monitoring action and device.
Background technology
At present, what intelligent household appliances occurred is more and more, also make it that people control Household electric appliance intelligent to require
More and more higher.And present mouse control is probably divided into three kinds of situations：1. common wireline mouse；2. common wireless mouse；3.
Sky mouse equipment.
In the prior art, still intelligent control household appliances are carried out using common wireline mouse, this is not easy to the reality of user
Border is operated, and also limit the development of intelligent household appliances.
Therefore, it is necessary to propose a kind of sky mouse recognition methods acted based on electromyographic signal monitoring and device, it is to avoid above-mentioned
The occurrence of.
The content of the invention
It is a primary object of the present invention to propose a kind of sky mouse recognition methods acted based on electromyographic signal monitoring and device,
It is intended to substitute traditional mouse action, and reaches the purpose of intelligent control electronic equipment.
To achieve the above object, a kind of sky mouse recognition methods acted based on electromyographic signal monitoring that the present invention is provided, institute
Stating method includes step：Obtain the First Eigenvalue of the first limb action；Sentence when the First Eigenvalue meets default first
During broken strip part, sky mouse pattern is opened；Under sky mouse pattern, the second feature for the electromyographic signal that the second limb action of monitoring is produced
Value；When the Second Eigenvalue meets default second Rule of judgment, second limb action is recognized.
Alternatively, before the First Eigenvalue of the first limb action of the acquisition, methods described also includes：Set first
The movement locus of limbs；Calculate the equal of movement locus acceleration magnitude at least one direction and the acceleration magnitude
Value；The acceleration average is write in matrix；According to the movement locus and the matrix, define described first and judge bar
Part.
Alternatively, after first Rule of judgment is defined, methods described also includes：Produced when gathering the second limb motion
At least one raw electromyographic signal；The electromyographic signal collected is pre-processed；Extract in pretreated electromyographic signal
Activity section；Extract the statistical characteristics in the activity section；According to the statistical characteristics, define described second and sentence
Broken strip part.
Alternatively, the statistical characteristics at least includes：The wavelet coefficient energy of single channel signal, wavelet coefficient total energy
Amount, wavelet coefficient absolute average and variance.
Alternatively, after the unlatching sky mouse pattern, methods described also includes：Detect that first limb action is produced
Displacement；The displacement is sent to operating system, to determine the distance of the displacement.
In addition, to achieve the above object, the present invention also proposes that a kind of recognized based on the sky mouse that electromyographic signal monitoring is acted is filled
Put, described device includes：Acquisition module, the First Eigenvalue for obtaining the first limb action；Opening module, for when described
When the First Eigenvalue meets default first Rule of judgment, sky mouse pattern is opened；Monitoring modular, under sky mouse pattern, supervising
Survey the Second Eigenvalue of the electromyographic signal of the second limb action generation；Identification module, for meeting pre- when the Second Eigenvalue
If the second Rule of judgment when, recognize second limb action.
Alternatively, described device also includes：Setup module, the movement locus for setting the first limbs；Computing module, is used
In the average for calculating movement locus acceleration magnitude at least one direction and the acceleration magnitude；Writing module,
For the acceleration average to be write in matrix；Definition module, for according to the movement locus and the matrix, defining institute
State the first Rule of judgment.
Alternatively, described device also includes：Acquisition module, at least one flesh produced during for gathering the second limb motion
Electric signal；Processing module, for being pre-processed to the electromyographic signal collected；Extraction module, it is pretreated for extracting
The statistical characteristics in activity section and extraction the activity section in electromyographic signal；The definition module, is also used
According to the statistical characteristics, second Rule of judgment is defined.
Alternatively, the statistical characteristics at least includes：The wavelet coefficient energy of single channel signal, wavelet coefficient total energy
Amount, wavelet coefficient absolute average and variance.
Alternatively, described device also includes：Detection module, for detecting the displacement that first limb action is produced；Hair
Module is sent, for the displacement to be sent into operating system, to determine the distance of the displacement.
The sky mouse recognition methods proposed by the present invention acted based on electromyographic signal monitoring and device, when the limbs of arm etc. first
When action meets the first Rule of judgment, sky mouse pattern is opened, and under sky mouse pattern, detects the production of the limb action of finger etc. second
When the Second Eigenvalue of raw electromyographic signal meets the second Rule of judgment, identify and second limbs such as specifically press or lift
Action, so as to realize the function of simulation sky mouse, improves Consumer's Experience.
Brief description of the drawings
The flow for the sky mouse recognition methods acted based on electromyographic signal monitoring that Fig. 1 provides for first embodiment of the invention is shown
It is intended to；
The flow for the sky mouse recognition methods acted based on electromyographic signal monitoring that Fig. 2 provides for second embodiment of the invention is shown
It is intended to；
The flow for the sky mouse recognition methods acted based on electromyographic signal monitoring that Fig. 3 provides for third embodiment of the invention is shown
It is intended to；
The module for the sky mouse identifying device acted based on electromyographic signal monitoring that Fig. 4 provides for fourth embodiment of the invention is shown
It is intended to；
The module for the sky mouse identifying device acted based on electromyographic signal monitoring that Fig. 5 provides for fifth embodiment of the invention is shown
It is intended to.
The realization, functional characteristics and advantage of the object of the invention will be described further referring to the drawings in conjunction with the embodiments.
Embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
Describe to realize the intelligent terminal of each embodiment of the invention referring now to accompanying drawing.In follow-up description, use
For represent element such as " module ", " part " or " unit " suffix only for be conducive to the present invention explanation, itself
Not specific meaning.Therefore, " module " can be used mixedly with " part ".
Fig. 1 is refer to, the sky mouse recognition methods acted based on electromyographic signal monitoring provided for first embodiment of the invention,
Methods described includes step：
Step 110, the First Eigenvalue of the first limb action is obtained.
Specifically, the first limbs can refer to the arm or hand of user.First limb action can refer to rocking for arm
Action, the shaking motion can be rocked or rock to the right to the left.The First Eigenvalue at least includes the first limbs and moved
The acceleration magnitude of the movement locus of work and the movement locus.
Step 120, judge whether the First Eigenvalue meets default first Rule of judgment, if so, then entering step
130, if it is not, then flow terminates.
Specifically, experimental experience, the present invention makes as described below to the first Rule of judgment.
The acceleration rate threshold B in an x-axis is provided, when the first acceleration of the action in x-axis is X, and during X >=B, is then recognized
It is an effective action for first action, the action can be that an effective left side is got rid of or right whipping is made, that is to say, that this
The acceleration of one action is effective.If 0<X<B, then the first action is that the possibility that action was got rid of and (got rid of on the right side) on a left side is P (X), P
(X) calculated using equation below：
P (X)=X>=B1:X/B formula (1)
Formula (1) is meant, when P (X) probability of calculating is more than threshold value B, then it is assumed that this is once effectively to get rid of
It is dynamic, so the P (X) of regulation at this moment is 1, that is to say, that acceleration is effective.Conversely, to judge the probability that such case occurs, such as
The really this probability rocked represents with x/B here, more than a certain empirical value (for example：80%), then also by the acceleration of current whipping
Degree is included in sum, and usually, each whipping has two acceleration (one positive one is negative), should if its probability is less than 80%
Group rejection of data.Here 80% is a statistical value, and when this probability has reached 80%, its accuracy is higher, also may be used certainly
To be other numerical value, such as 90% and 70%, but the erroneous judgement of this when is just of a relatively high.
P (X | A)=P (X) P (A) formula (2)
Formula (2) is meant, after whipping twice, and the effective percentage of acceleration is how many.In the present embodiment, provide
P (X | A) >=0.8, you can think that the acceleration of whipping twice is effective.This value is too high, can cause the mortality started shooting below
It is higher；It is too low, easily cause false start.
Further, in order to improve the accuracy rate of start, only a regulation left side is got rid of or the right action got rid of is infeasible
, in actual use, it is necessary to which regulation user must the right side be got rid of again or a left side is got rid of respectively once.In addition, it is necessary to which explanation, gets rid of every time
Dynamic acceleration, it is not necessary to B values will be exceeded, as long as the overall possibility calculated according to probability analysis has reached 80%, just
It is considered that this group of acceleration magnitude is effective.
Further, in addition it is also necessary to which the time interval of whipping action twice is set.Because, if user is violent for the first time
After strong whipping, cross and just done second of whipping for a long time, this is clearly infeasible.So needing to define a time restriction
Condition, for example, the time restriction condition can the above 2 seconds, that is to say, that regulation had two or so big whippings to move in 2 seconds
Make, just it can be assumed that acceleration is effective.It is followed by one small to get rid of but user is after once big whipping big for the first time
It is dynamic, if after an one direction whipping success, the whipping of second of opposite direction must be followed by completed within a certain period of time,
Otherwise whipping will all it abandon twice, counter O reset.So, the accuracy rate that detection pattern is opened just is greatly reduced.More specifically
Ground, algorithm of signal-wobble elimination is described as follows：
If the successful probability P of single whipping (X)<0.8, it is a shake action to be considered as this.Why it is set to
0.8, it is because defining P (X | A) >=0.8, you can think that whipping is the effective whipping action of its acceleration twice.So single
The probability of whipping<0.8, even if another whipping probability is 1, the probability being multiplied twice is also no more than 0.8, so will shake general
It is rational that rate, which is set as 0.8,.
Step 130, sky mouse pattern is opened.
Step 140, under sky mouse pattern, the Second Eigenvalue for the electromyographic signal that the second limb action of monitoring is produced.
Specifically, the second limbs are gathered using differential electrode configuration mode of the prior art in the present embodiment and moves surface flesh
Electric signal, and by carrying out a series of processing such as noise reduction, adding window, feature extraction to surface electromyogram signal, obtain the surface flesh
The characteristic value of electric signal, is referred to as Second Eigenvalue by the characteristic value of surface electromyogram signal.
Further, the second limbs can refer to finger, more specifically refer to forefinger, middle finger etc..
Step 150, judge whether the Second Eigenvalue meets default second Rule of judgment, if so, then entering step
160, if it is not, then flow terminates.
Specifically, Second Eigenvalue is compared with the second Rule of judgment in grader, judges the Second Eigenvalue
Whether default second Rule of judgment is met, if so, then entering step 160, if it is not, then flow terminates.
Step 160, second limb action is recognized.
Specifically, second limb action can the action such as press or lift, so as under sky mouse pattern, pass through
The second limb action recognized realizes the control to intelligent household appliances.
Further, because user action is accustomed to difference, action is likely to imperfect or not accurate enough when completion.
During user operates, when user confirms that this secondary control implements correct, can be this user is dynamic
As being sampled, and obtained characteristic value will be sampled and preset characteristic value of dispatching from the factory is write in same matrix, retrieved point
The first Rule of judgment and the second Rule of judgment of class device, thus, substantially increasing the discrimination of grader.
For example, preset features value, which is 3000, is worth (A1-A3000) to train grader as characteristic value, in user most
Start N (0<N<3000) among secondary beginning operating process, whenever an action is completed, system can point out user to select oneself
Whether action is correctly recognized by intelligent appliance product (such as television set), if correct identification, then by the signal of this group of data
According to above-mentioned flow processing, and obtain in corresponding characteristic value (B1-BN), the accuracy to increase identification.Here to make
Limit, the study of action, at most can only obtain the data of fixed action, it is assumed that be N number of, when the quantity that standard operation learns surpasses
When crossing N, system can be automatically deleted the data learnt earliest, to ensure that standard operation study is no more than N.
The sky mouse recognition methods acted based on electromyographic signal monitoring of the present embodiment, when the limb action of arm etc. first is met
During the first Rule of judgment, sky mouse pattern is opened, and under sky mouse pattern, detects the myoelectricity of the limb action of finger etc. second generation
When the Second Eigenvalue of signal meets the second Rule of judgment, identify and second limb action such as specifically press or lift, from
And the function of simulation sky mouse is realized, improve Consumer's Experience.
Fig. 2 is refer to, the sky mouse recognition methods acted based on electromyographic signal monitoring provided for second embodiment of the invention,
Methods described includes step：
Step 210, the First Eigenvalue of the first limb action is obtained.
Step 220, judge whether the First Eigenvalue meets default first Rule of judgment, if so, then entering step
230, if it is not, then flow terminates.
Step 230, sky mouse pattern is opened.
Step 210- steps 230 in the present embodiment are identical with the step 110- steps 130 in first embodiment, for phase
Same content, the present embodiment is repeated no more.
Step 240, the displacement that first limb action is produced is detected.
Specifically, after sky mouse pattern is opened, the first limbs are (for example：Hand) 360 degree in selected plane
When mobile, gestures detection ring records motion track and its displacement.
Step 250, the displacement is sent to operating system, to determine the distance of the displacement.
Specifically, feed back information to operating system, operating system can corresponding mouse moving event it is determined that opening
After sky mouse pattern, using displacement transducer, to detect the action of the first limbs movement.
Step 260, under sky mouse pattern, the Second Eigenvalue for the electromyographic signal that the second limb action of monitoring is produced.
Step 270, judge whether the Second Eigenvalue meets default second Rule of judgment, if so, then entering step
280, if it is not, then flow terminates.
Step 280, second limb action is recognized.
Step 260- steps 280 in the present embodiment are identical with the step 140- steps 160 in first embodiment, for phase
Same content, the present embodiment is repeated no more.
The sky mouse recognition methods acted based on electromyographic signal monitoring of the present embodiment, is produced when the first limb action of detection
During displacement, the displacement is sent to operating system, to determine the distance of the displacement.
Fig. 3 is refer to, the sky mouse recognition methods acted based on electromyographic signal monitoring provided for third embodiment of the invention.
In the third embodiment, the sky mouse recognition methods acted based on electromyographic signal monitoring is real in first embodiment or second
Made further improvement on the basis of example is applied, is differed only in, before the First Eigenvalue of the first limb action is obtained,
Methods described also includes：
Step 310, the movement locus of the first limbs is set.
Specifically, in order to simplify algorithm, also for the accuracy rate for improving probability analysis, the first limbs are defined (for example：Hand
Portion) double swerve be a kind of special movement locus, the movement locus existed only in set plane.
In the present embodiment, the plane where can defining can only be vertical or parallel with hand, before runtime, user
It is vertical or parallel relative to arm to need selection movement locus plane.
Step 320, movement locus acceleration magnitude at least one direction and the acceleration magnitude are calculated
Average.
Specifically, acceleration magnitude of the detection in x-axis, when rocking according to set track motion, or its fortune
Dynamic rail mark is close to when presetting an equation of locus, so that it may be judged to triggering the action that bracelet is opened.Probability analysis
Result be a probable value, the value illustrates whether the first limb action is that the possibility of boot action is much, this and human brain
Pattern-recognition mode be consistent.Judgement of the mankind to certain part things, is substantially based on the judgement of probability.
In addition, being described and limiting to rocking, i.e.,：Only rocked in given plane as once conduct between -1 to 1
Successfully once rock.After rocking successfully, acceleration magnitude is extracted.
Further, successfully make as described below to rocking：
It can will rock to regard as and complete a curvilinear motion, its curvilinear equation can be understood as straight line, it is possible thereby to advise
Accurate movement locus is calibrated, is set：When movement locus standard compliant movement locus, just it is considered once successfully to open
Start and make.After the action of double swerve has been done, the acceleration magnitude of extraction both direction, two acceleration currently extracted,
Referred to as one group characteristic vector, is defined herein：When the difference between two characteristic vectors is less than or equal to 0.05, you can
Think that this two groups of characteristic vectors are similar.Herein on basis, it is assumed that current characteristic vector and all features in preset matrix to
When measuring all similarities, then it is assumed that the vector is 1 with eigenmatrix similarity.If current this group of characteristic vector and preset features square
When any one group of characteristic vector is similar in battle array, you can it is available to think this group of acceleration magnitude.
Further, whipping is all on a center of circle, to rock back and forth each time, and the whipping must be symmetrical gets rid of
It is dynamic.For example, just having to once move to 225 degree of direction after 45 degree upwards motion, the movement locus obtained with this.
The statistical characteristics in the acceleration of the multigroup motion of repetition of this equation of locus is calculated, i.e.,：Acceleration average,
Threshold value B namely in first embodiment.
Step 330, the acceleration average is write in matrix.
Step 340, according to the movement locus and the matrix, first Rule of judgment is defined.
Step 350, at least one electromyographic signal produced when gathering the second limb motion.
Specifically, the collection of surface electromyogram signal is to use differential electrode configuration mode of the prior art, and sample rate is set
2000Hz is set to, in other embodiments, the sample rate can be set as the case may be.
Step 360, the electromyographic signal collected is pre-processed.
Specifically, go out electromyographic signal using electrode test, according to the feature (20-500Hz) of electromyographic signal, lead to its needs
Conventional bandpass filter is crossed, the noise less than 20Hz and more than 500Hz is filtered, herein on basis, then passes through existing skill
Noise reduction in the small wavestrip of art, makes noise of the signal collected between 20-500Hz obtain a certain degree of suppression, by with
Signal after upper processing is to compare pure electromyographic signal for primary signal is collected.
Step 370, the activity section in pretreated electromyographic signal is extracted.
Specifically, in order to obtain surface electromyogram signal characteristic parameter on individual part, it is necessary to in step 360
Surface electromyogram signal after the noise reduction arrived carries out active segment detection.For some acting surface electromyographic signal, according to its energy
The change of value, windowing process is carried out to it.
Windowing process derivation of energy formula is：
Wherein, Qn is the total energy value of the segment signal, and tn is the starting point of a certain time-ofday signals, and x represents the segment signal, T tables
Show the signal segment length of interception in window function.
The principle of windowing process is estimated according to the signal energy value collected, according to formula (3), by whole
The calculating of signal energy, and according to the energy threshold M and N of the segment signal, active segment detection is carried out to it.Wherein, M, N are to pass through
The continuous data of one section of starting point and terminating point to multiple actions are calculated obtained from average value.Specifically, assuming from letter
Number starting point is elapsed backward, during movement, when the energy value for the point for continuous predetermined number (such as 500) occur is all high
In threshold value M, then it is assumed that be exactly the starting point of action signal here, then proceed to elapse backward, when continuous 500 (this numerical value of appearance
Can be with self-defined) energy value of individual point is when being less than N, and just explanation action signal terminates here.
Thus to the mobile windowing process of surface electromyogram signal energy, and the setting of suitable threshold parameter is combined, so that it may essence
True detects single acting surface electromyographic signal.Here the signal value obtained is seen in time domain, between threshold value M and N,
Signal is normal value, and the signal outside M and N, entirely null value, now just it is believed that after active segment detection
In signal, the continuous part of independent non-zero is the surface electromyogram signal of the second limb action.
Step 380, the statistical characteristics in the activity section is extracted.
Specifically, it is the single dynamic of N to length using sym8 small echos as basic function on the basis of active segment detection
Make surface electromyogram signal number and carry out WAVELET PACKET DECOMPOSITION, and calculate every layer of wavelet coefficient r statistical nature respectively, i.e.,：
For the surface electromyogram signal number after step 360 and step 370 processing, the wavelet coefficient of single channel signal
Energy is：
Wherein, Ej is the total energy value of wavelet coefficient, and N is the number of wavelet coefficient in the segment signal, and ri is corresponding points
Wavelet coefficient.
Correspondingly, wavelet coefficient gross energy should be：
E=E1+E2+E3+E4Formula (5)
So, the energy percentage on wavelet coefficient is then represented by：
ρj=Ej/ E formula (6)
For single pass acting surface electromyographic signal number, its wavelet coefficient absolute average and variance can be represented respectively
For：
Step 390, according to the statistical characteristics, second Rule of judgment is defined.
Specifically, the statistical characteristics of the electromyographic signal wavelet coefficient got according to step 380, as characteristic parameter,
And combination supporting vector machine realizes the pattern classification to gesture motion.Specifically：
Input parameter is mapped to high-dimensional feature space by SVMs by the non-linear relation of its kernel function, and is constructed
Corresponding optimal separating hyper plane.To sample (xi,yi) for, the optimal classification discriminant function of SVMs is represented by
Wherein, parameter alphaiIt is to need the function coefficients of optimization calculating, inner product k (x, x with bi) it is its kernel function, m is sample number
And i=1,2 ... m.
Based on two class principles of classification, SVMs is also extrapolated to multicategory classification problem, and it can pass through " a pair
One " is realized with " one-to-many " two ways, and " one-to-many " mode is used here.For K class classification problems, " one-to-many "
Mode constructs K binary classifier, i.e., by jth (j=1,2 ..., K) individual grader by the data of jth class and other class data
Make a distinction.For relative " one-to-one " mode, the classification speed of institute's employing mode is very fast.
, it is necessary to be trained using the above method to pattern classifier before dispatching from the factory, for example：The sampled value of individual part
For 3000, and this 3000 values are input in SVM pattern classifiers, it is trained, classifier result conduct is obtained
Second Rule of judgment.
Further, the second Rule of judgment is written among chip, enables to play action knowledge in follow-up use
Other effect.
The sky mouse recognition methods acted based on electromyographic signal monitoring of the present embodiment, first before being dispatched from the factory by definition judges
Condition and the second Rule of judgment, so as to accurately determine unlatching sky mouse pattern, and then rapidly identify follow-up action.
Fig. 4 is refer to, the sky mouse identifying device acted based on electromyographic signal monitoring provided for fourth embodiment of the invention,
Described device includes：
Acquisition module 410, the First Eigenvalue for obtaining the first limb action.
Specifically, the first limbs can refer to the arm or hand of user.First limb action can refer to rocking for arm
Action, the shaking motion can be rocked or rock to the right to the left.The First Eigenvalue at least includes the first limbs and moved
The acceleration magnitude of the movement locus of work and the movement locus.
Opening module 420, for when the First Eigenvalue meets default first Rule of judgment, opening sky mouse pattern.
Specifically, experimental experience, the present invention makes as described below to the first Rule of judgment.
The acceleration rate threshold B in an x-axis is provided, when the first acceleration of the action in x-axis is X, and during X >=B, is then recognized
It is an effective action for first action, the action can be that an effective left side is got rid of or right whipping is made, that is to say, that this
The acceleration of one action is effective.If 0<X<B, then the first action is that the possibility that action was got rid of and (got rid of on the right side) on a left side is P (X), P
(X) calculated using equation below：
P (X)=X>=B1:X/B formula (1)
Formula (1) is meant, when P (X) probability of calculating is more than threshold value B, then it is assumed that this is once effectively to get rid of
It is dynamic, so the P (X) of regulation at this moment is 1, that is to say, that acceleration is effective.Conversely, to judge the probability that such case occurs, such as
The really this probability rocked represents with x/B here, more than a certain empirical value (for example：80%), then also by the acceleration of current whipping
Degree is included in sum, and usually, each whipping has two acceleration (one positive one is negative), should if its probability is less than 80%
Group rejection of data.Here 80% is a statistical value, and when this probability has reached 80%, its accuracy is higher, also may be used certainly
To be other numerical value, such as 90% and 70%, but the erroneous judgement of this when is just of a relatively high.
P (X | A)=P (X) P (A) formula (2)
Formula (2) is meant, after whipping twice, and the effective percentage of acceleration is how many.In the present embodiment, provide
P (X | A) >=0.8, you can think that the acceleration of whipping twice is effective.This value is too high, can cause the mortality started shooting below
It is higher；It is too low, easily cause false start.
Further, in order to improve the accuracy rate of start, only a regulation left side is got rid of or the right action got rid of is infeasible
, in actual use, it is necessary to which regulation user must the right side be got rid of again or a left side is got rid of respectively once.In addition, it is necessary to which explanation, gets rid of every time
Dynamic acceleration, it is not necessary to B values will be exceeded, as long as the overall possibility calculated according to probability analysis has reached 80%, just
It is considered that this group of acceleration magnitude is effective.
Further, in addition it is also necessary to which the time interval of whipping action twice is set.Because, if user is violent for the first time
After strong whipping, cross and just done second of whipping for a long time, this is clearly infeasible.So needing to define a time restriction
Condition, for example, the time restriction condition can the above 2 seconds, that is to say, that regulation had two or so big whippings to move in 2 seconds
Make, just it can be assumed that acceleration is effective.It is followed by one small to get rid of but user is after once big whipping big for the first time
It is dynamic, if after an one direction whipping success, the whipping of second of opposite direction must be followed by completed within a certain period of time,
Otherwise whipping will all it abandon twice, counter O reset.So, the accuracy rate that detection pattern is opened just is greatly reduced.More specifically
Ground, algorithm of signal-wobble elimination is described as follows：
If the successful probability P of single whipping (X)<0.8, it is a shake action to be considered as this.Why it is set to
0.8, it is because defining P (X | A) >=0.8, you can think that whipping is the effective whipping action of its acceleration twice.So single
The probability of whipping<0.8, even if another whipping probability is 1, the probability being multiplied twice is also no more than 0.8, so will shake general
It is rational that rate, which is set as 0.8,.
Monitoring modular 430, under sky mouse pattern, monitoring the second feature for the electromyographic signal that the second limb action is produced
Value.
Specifically, in the present embodiment, monitoring modular 430 is using differential electrode configuration mode of the prior art collection second
Limbs move surface electromyogram signal, and by carrying out a series of processing such as noise reduction, adding window, feature extraction to surface electromyogram signal, obtain
To the characteristic value of the surface electromyogram signal, the characteristic value of surface electromyogram signal is referred to as Second Eigenvalue.
Further, the second limbs can refer to finger, more specifically refer to forefinger, middle finger etc..
Identification module 440, for when Second Eigenvalue meets default second Rule of judgment, recognizing second limbs
Action.
Specifically, Second Eigenvalue is compared with the second Rule of judgment in grader, when Second Eigenvalue is met
During default second Rule of judgment, the specific action of the second limb action is recognized.
Further, second limb action can the action such as press or lift, so as under sky mouse pattern, lead to
Cross control of the second limb action realization recognized to intelligent household appliances.
Further, because user action is accustomed to difference, action is likely to imperfect or not accurate enough when completion.
During user operates, when user confirms that this secondary control implements correct, can be this user is dynamic
As being sampled, and obtained characteristic value will be sampled and preset characteristic value of dispatching from the factory is write in same matrix, retrieved point
The first Rule of judgment and the second Rule of judgment of class device, thus, substantially increasing the discrimination of grader.
For example, preset features value, which is 3000, is worth (A1-A3000) to train grader as characteristic value, in user most
Start N (0<N<3000) among secondary beginning operating process, whenever an action is completed, system can point out user to select oneself
Whether action is correctly recognized by intelligent appliance product (such as television set), if correct identification, then by the signal of this group of data
According to above-mentioned flow processing, and obtain in corresponding characteristic value (B1-BN), the accuracy to increase identification.Here to make
Limit, the study of action, at most can only obtain the data of fixed action, it is assumed that be N number of, when the quantity that standard operation learns surpasses
When crossing N, system can be automatically deleted the data learnt earliest, to ensure that standard operation study is no more than N.
Further, the sky mouse identifying device acted based on electromyographic signal monitoring in the present embodiment can also include：
Detection module 450, for detecting the displacement that first limb action is produced.
Specifically, after sky mouse pattern is opened, the first limbs are (for example：Hand) 360 degree in selected plane
When mobile, detection module 450 records motion track and its displacement.
Sending module 460, for the displacement to be sent into operating system, to determine the distance of the displacement.
Specifically, sending module 460 feeds back information to operating system, operating system can corresponding mouse moving event exist
It is determined that after opening sky mouse pattern, using displacement transducer, to detect the action of the first limbs movement.
The sky mouse identifying device acted based on electromyographic signal monitoring of the present embodiment, the hand got when acquisition module 410
When the limb action of arm etc. first meets the first Rule of judgment, opening module 420 opens sky mouse pattern, and under sky mouse pattern, prison
Survey module 430 and monitor that the Second Eigenvalue for the electromyographic signal that the limb action of finger etc. second is produced meets the second Rule of judgment
When, identification module 440, which is identified, second limb action such as specifically press or lift, so that the function of simulation sky mouse is realized,
Improve Consumer's Experience.
Fig. 5 is refer to, fifth embodiment of the invention further provides for a kind of sky mouse knowledge acted based on electromyographic signal monitoring
Other device.In the 5th embodiment, the sky mouse identifying device acted based on electromyographic signal monitoring is in fourth embodiment
On the basis of the further improvement made, differ only in, described device also includes：
Setup module 510, the movement locus for setting the first limbs.
Specifically, in order to simplify algorithm, also for the accuracy rate for improving probability analysis, the first limbs are defined (for example：Hand
Portion) double swerve be a kind of special movement locus, the movement locus existed only in set plane.
In the present embodiment, the plane where can defining can only be vertical or parallel with hand, before runtime, user
It is vertical or parallel relative to arm to need selection movement locus plane.
Computing module 520, for calculate movement locus acceleration magnitude at least one direction and it is described plus
The average of velocity amplitude.
Specifically, acceleration magnitude of the detection in x-axis, when rocking according to set track motion, or its fortune
Dynamic rail mark is close to when presetting an equation of locus, so that it may be judged to triggering the action that bracelet is opened.Probability analysis
Result be a probable value, the value illustrates whether the first limb action is that the possibility of boot action is much, this and human brain
Pattern-recognition mode be consistent.Judgement of the mankind to certain part things, is substantially based on the judgement of probability.
In addition, being described and limiting to rocking, i.e.,：Only rocked in given plane as once conduct between -1 to 1
Successfully once rock.After rocking successfully, acceleration magnitude is extracted.
Further, successfully make as described below to rocking：
It can will rock to regard as and complete a curvilinear motion, its curvilinear equation can be understood as straight line, it is possible thereby to advise
Accurate movement locus is calibrated, is set：When movement locus standard compliant movement locus, just it is considered once successfully to open
Start and make.After the action of double swerve has been done, the acceleration magnitude of extraction both direction, two acceleration currently extracted,
Referred to as one group characteristic vector, is defined herein：When the difference between two characteristic vectors is less than or equal to 0.05, you can
Think that this two groups of characteristic vectors are similar.Herein on basis, it is assumed that current characteristic vector and all features in preset matrix to
When measuring all similarities, then it is assumed that the vector is 1 with eigenmatrix similarity.If current this group of characteristic vector and preset features square
When any one group of characteristic vector is similar in battle array, you can it is available to think this group of acceleration magnitude.
Further, whipping is all on a center of circle, to rock back and forth each time, and the whipping must be symmetrical gets rid of
It is dynamic.For example, just having to once move to 225 degree of direction after 45 degree upwards motion, the movement locus obtained with this.
The statistical characteristics in the acceleration of the multigroup motion of repetition of this equation of locus is calculated, i.e.,：Acceleration average,
Threshold value B namely in first embodiment.
Writing module 530, for the acceleration average to be write in matrix.
Definition module 540, for according to the movement locus and the matrix, defining first Rule of judgment.
Acquisition module 550, at least one electromyographic signal produced during for gathering the second limb motion.
Specifically, collection of the acquisition module 550 to surface electromyogram signal is using differential electrode of the prior art configuration
Mode, sample rate is set to 2000Hz, in other embodiments, and the sample rate can be set as the case may be.
Processing module 560, for being pre-processed to the electromyographic signal that acquisition module 550 is collected.
Specifically, go out electromyographic signal using electrode test, according to the feature (20-500Hz) of electromyographic signal, lead to its needs
Conventional bandpass filter is crossed, processing module 560 filters the noise less than 20Hz and more than 500Hz, herein on basis, then
By noise reduction in the small wavestrip of prior art, noise of the signal collected between 20-500Hz is set to have obtained a certain degree of suppression
System, the signal after being handled more than is to compare pure electromyographic signal for primary signal is collected.
Extraction module 570, for extracting the section of the activity in pretreated electromyographic signal.
Specifically, in order to obtain surface electromyogram signal characteristic parameter on individual part, it is necessary to in step 360
Surface electromyogram signal after the noise reduction arrived carries out active segment detection.For some acting surface electromyographic signal, according to its energy
The change of value, windowing process is carried out to it.
Windowing process derivation of energy formula is：
Wherein, Qn is the total energy value of the segment signal, and tn is the starting point of a certain time-ofday signals, and x represents the segment signal, T tables
Show the signal segment length of interception in window function.
The principle of windowing process is estimated according to the signal energy value collected, according to formula (3), by whole
The calculating of signal energy, and according to the energy threshold M and N of the segment signal, active segment detection is carried out to it.Wherein, M, N are to pass through
The continuous data of one section of starting point and terminating point to multiple actions are calculated obtained from average value.Specifically, assuming from letter
Number starting point is elapsed backward, during movement, when the energy value for the point for continuous predetermined number (such as 500) occur is all high
In threshold value M, then it is assumed that be exactly the starting point of action signal here, then proceed to elapse backward, when continuous 500 (this numerical value of appearance
Can be with self-defined) energy value of individual point is when being less than N, and just explanation action signal terminates here.
Thus to the mobile windowing process of surface electromyogram signal energy, and the setting of suitable threshold parameter is combined, so that it may essence
True detects single acting surface electromyographic signal.Here the signal value obtained is seen in time domain, between threshold value M and N,
Signal is normal value, and the signal outside M and N, entirely null value, now just it is believed that after active segment detection
In signal, the continuous part of independent non-zero is the surface electromyogram signal of the second limb action.
Extraction module 570, is additionally operable to extract the statistical characteristics in the activity section.
Specifically, it is the single dynamic of N to length using sym8 small echos as basic function on the basis of active segment detection
Make surface electromyogram signal number and carry out WAVELET PACKET DECOMPOSITION, and calculate every layer of wavelet coefficient r statistical nature respectively, i.e.,：
Surface electromyogram signal number after for being handled by processing module 560, the wavelet coefficient energy of single channel signal
For：
Wherein, Ej is the total energy value of wavelet coefficient, and N is the number of wavelet coefficient in the segment signal, and ri is corresponding points
Wavelet coefficient.
Correspondingly, wavelet coefficient gross energy should be：
E=E1+E2+E3+E4Formula (5)
So, the energy percentage on wavelet coefficient is then represented by：
ρj=Ej/ E formula (6)
For single pass acting surface electromyographic signal number, its wavelet coefficient absolute average and variance can be represented respectively
For：
Definition module 580, for according to the statistical characteristics, defining second Rule of judgment.
Specifically, the statistical characteristics of the electromyographic signal wavelet coefficient got according to extraction module 570, joins as feature
Number, and combination supporting vector machine realizes the pattern classification to gesture motion.Specifically：
Input parameter is mapped to high-dimensional feature space by SVMs by the non-linear relation of its kernel function, and is constructed
Corresponding optimal separating hyper plane.To sample (xi,yi) for, the optimal classification discriminant function of SVMs is represented by
Wherein, parameter alphaiIt is to need the function coefficients of optimization calculating, inner product k (x, x with bi) it is its kernel function, m is sample number
And i=1,2 ... m.
Based on two class principles of classification, SVMs is also extrapolated to multicategory classification problem, and it can pass through " a pair
One " is realized with " one-to-many " two ways, and " one-to-many " mode is used here.For K class classification problems, " one-to-many "
Mode constructs K binary classifier, i.e., by jth (j=1,2 ..., K) individual grader by the data of jth class and other class data
Make a distinction.For relative " one-to-one " mode, the classification speed of institute's employing mode is very fast.
, it is necessary to be trained using the above method to pattern classifier before dispatching from the factory, for example：The sampled value of individual part
For 3000, and this 3000 values are input in SVM pattern classifiers, it is trained, classifier result conduct is obtained
Second Rule of judgment.
Further, the second Rule of judgment is written among chip, enables to play action knowledge in follow-up use
Other effect.
The sky mouse identifying device acted based on electromyographic signal monitoring of the present embodiment, is defined out by definition module 540
The first Rule of judgment and the second Rule of judgment before factory, so as to accurately determine unlatching sky mouse pattern, and then are rapidly identified
Follow-up action.
It should be noted that herein, term " comprising ", "comprising" or its any other variant are intended to non-row
His property is included, so that process, method, article or device including a series of key elements not only include those key elements, and
And also including other key elements being not expressly set out, or also include for this process, method, article or device institute inherently
Key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that including this
Also there is other identical element in process, method, article or the device of key element.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Understood based on such, technical scheme is substantially done to prior art in other words
Going out the part of contribution can be embodied in the form of software product, and the computer software product is stored in a storage medium
In (such as ROM/RAM, magnetic disc, CD), including some instructions are make it that a station terminal equipment (can be mobile phone, computer, clothes
It is engaged in device, air-conditioner, or network equipment etc.) perform method described in each embodiment of the invention.
The preferred embodiments of the present invention are these are only, are not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of sky mouse recognition methods acted based on electromyographic signal monitoring, it is characterised in that methods described includes step：
Obtain the First Eigenvalue of the first limb action；
When the First Eigenvalue meets default first Rule of judgment, sky mouse pattern is opened；
Under sky mouse pattern, the Second Eigenvalue for the electromyographic signal that the second limb action of monitoring is produced；
When the Second Eigenvalue meets default second Rule of judgment, second limb action is recognized.
2. the sky mouse recognition methods according to claim 1 acted based on electromyographic signal monitoring, it is characterised in that described
Before the First Eigenvalue for obtaining the first limb action, methods described also includes：
The movement locus of first limbs is set；
Calculate the average of movement locus acceleration magnitude at least one direction and the acceleration magnitude；
The acceleration average is write in matrix；
According to the movement locus and the matrix, first Rule of judgment is defined.
3. the sky mouse recognition methods according to claim 2 acted based on electromyographic signal monitoring, it is characterised in that in definition
After first Rule of judgment, methods described also includes：
Gather at least one electromyographic signal produced during the second limb motion；
The electromyographic signal collected is pre-processed；
Extract the activity section in pretreated electromyographic signal；
Extract the statistical characteristics in the activity section；
According to the statistical characteristics, second Rule of judgment is defined.
4. the sky mouse recognition methods according to claim 3 acted based on electromyographic signal monitoring, it is characterised in that the system
Meter characteristic value at least includes：The wavelet coefficient energy of single channel signal, wavelet coefficient gross energy, wavelet coefficient absolute average and
Variance.
5. the sky mouse recognition methods according to claim 1 acted based on electromyographic signal monitoring, it is characterised in that described
Open after sky mouse pattern, methods described also includes：
Detect the displacement that first limb action is produced；
The displacement is sent to operating system, to determine the distance of the displacement.
6. a kind of sky mouse identifying device acted based on electromyographic signal monitoring, it is characterised in that described device includes：
Acquisition module, the First Eigenvalue for obtaining the first limb action；
Opening module, for when the First Eigenvalue meets default first Rule of judgment, opening sky mouse pattern；
Monitoring modular, under sky mouse pattern, monitoring the Second Eigenvalue for the electromyographic signal that the second limb action is produced；
Identification module, for when the Second Eigenvalue meets default second Rule of judgment, recognizing that second limbs are moved
Make.
7. the sky mouse identifying device according to claim 6 acted based on electromyographic signal monitoring, it is characterised in that the dress
Putting also includes：
Setup module, the movement locus for setting the first limbs；
Computing module, for calculating movement locus acceleration magnitude at least one direction and the acceleration magnitude
Average；
Writing module, for the acceleration average to be write in matrix；
Definition module, for according to the movement locus and the matrix, defining first Rule of judgment.
8. the sky mouse identifying device according to claim 7 acted based on electromyographic signal monitoring, it is characterised in that the dress
Putting also includes：
Acquisition module, at least one electromyographic signal produced during for gathering the second limb motion；
Processing module, for being pre-processed to the electromyographic signal collected；
Extraction module, for extracting the section of the activity in pretreated electromyographic signal and extracting the activity section
In statistical characteristics；
The definition module, is additionally operable to, according to the statistical characteristics, define second Rule of judgment.
9. the sky mouse identifying device according to claim 8 acted based on electromyographic signal monitoring, it is characterised in that the system
Meter characteristic value at least includes：The wavelet coefficient energy of single channel signal, wavelet coefficient gross energy, wavelet coefficient absolute average and
Variance.
10. the sky mouse identifying device according to claim 6 acted based on electromyographic signal monitoring, it is characterised in that described
Device also includes：
Detection module, for detecting the displacement that first limb action is produced；
Sending module, for the displacement to be sent into operating system, to determine the distance of the displacement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710193046.3A CN106959756A (en) | 2017-03-28 | 2017-03-28 | The sky mouse recognition methods acted based on electromyographic signal monitoring and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710193046.3A CN106959756A (en) | 2017-03-28 | 2017-03-28 | The sky mouse recognition methods acted based on electromyographic signal monitoring and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106959756A true CN106959756A (en) | 2017-07-18 |
Family
ID=59471639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710193046.3A Pending CN106959756A (en) | 2017-03-28 | 2017-03-28 | The sky mouse recognition methods acted based on electromyographic signal monitoring and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106959756A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107608524A (en) * | 2017-10-13 | 2018-01-19 | 中国科学院软件研究所 | A kind of multi objective control command generating device based on myoelectricity |
| CN114442798A (en) * | 2020-11-06 | 2022-05-06 | 复旦大学附属妇产科医院 | Portable control system and control method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103677289A (en) * | 2013-12-09 | 2014-03-26 | 中国科学院深圳先进技术研究院 | Intelligent interactive glove and interactive method |
| CN105373214A (en) * | 2014-08-25 | 2016-03-02 | 深圳Tcl新技术有限公司 | Terminal control method and system |
| CN105807640A (en) * | 2014-12-30 | 2016-07-27 | 深圳Tcl新技术有限公司 | Terminal control method and terminal control system |
| CN106527716A (en) * | 2016-11-09 | 2017-03-22 | 努比亚技术有限公司 | Wearable equipment based on electromyographic signals and interactive method between wearable equipment and terminal |
-
2017
- 2017-03-28 CN CN201710193046.3A patent/CN106959756A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103677289A (en) * | 2013-12-09 | 2014-03-26 | 中国科学院深圳先进技术研究院 | Intelligent interactive glove and interactive method |
| CN105373214A (en) * | 2014-08-25 | 2016-03-02 | 深圳Tcl新技术有限公司 | Terminal control method and system |
| CN105807640A (en) * | 2014-12-30 | 2016-07-27 | 深圳Tcl新技术有限公司 | Terminal control method and terminal control system |
| CN106527716A (en) * | 2016-11-09 | 2017-03-22 | 努比亚技术有限公司 | Wearable equipment based on electromyographic signals and interactive method between wearable equipment and terminal |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107608524A (en) * | 2017-10-13 | 2018-01-19 | 中国科学院软件研究所 | A kind of multi objective control command generating device based on myoelectricity |
| CN114442798A (en) * | 2020-11-06 | 2022-05-06 | 复旦大学附属妇产科医院 | Portable control system and control method |
| CN114442798B (en) * | 2020-11-06 | 2024-05-07 | 复旦大学附属妇产科医院 | Portable control system and control method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106383579A (en) | EMG and FSR-based refined gesture recognition system and method | |
| Uddin et al. | Random forests based recognition of human activities and postural transitions on smartphone | |
| Liu et al. | Research on human activity recognition based on active learning | |
| CN103294199B (en) | A kind of unvoiced information identifying system based on face's muscle signals | |
| CN106910314B (en) | A kind of personalized fall detection method based on the bodily form | |
| CN108446733A (en) | A kind of human body behavior monitoring and intelligent identification Method based on multi-sensor data | |
| CN103886341A (en) | Gait behavior recognition method based on feature combination | |
| Carfi et al. | Online human gesture recognition using recurrent neural networks and wearable sensors | |
| CN107437074A (en) | A kind of identity identifying method and device | |
| CN106527716A (en) | Wearable equipment based on electromyographic signals and interactive method between wearable equipment and terminal | |
| CN106919958A (en) | A kind of human finger action identification method based on intelligent watch | |
| CN104517097A (en) | Kinect-based moving human body posture recognition method | |
| CN104269025B (en) | Wearable single node feature and the position choosing method of monitoring is fallen down towards open air | |
| CN104049752A (en) | Interaction method based on human body and interaction device based on human body | |
| CN104484644A (en) | Gesture identification method and device | |
| De Falco et al. | Evaluation of artificial intelligence techniques for the classification of different activities of daily living and falls | |
| CN110298263A (en) | Real-time accurate and contactless gesture identification method and system based on RFID system | |
| CN110929242B (en) | Method and system for carrying out attitude-independent continuous user authentication based on wireless signals | |
| CN106959756A (en) | The sky mouse recognition methods acted based on electromyographic signal monitoring and device | |
| Guelta et al. | An Improved Behavioral Biometric System based on Gait and ECG signals. | |
| Zheng et al. | L-sign: Large-vocabulary sign gestures recognition system | |
| CN114384999B (en) | User-independent myoelectric gesture recognition system based on self-adaptive learning | |
| Kulchyk et al. | Activity recognition with wearable accelerometers using deep convolutional neural network and the effect of sensor placement | |
| Moiz et al. | A comparative study of classification methods for gesture recognition using a 3-axis accelerometer | |
| CN109308133A (en) | Intelligent interaction projects interaction technique |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170718 |
|
| RJ01 | Rejection of invention patent application after publication |