CN106127188B - A kind of Handwritten Digit Recognition method based on gyroscope - Google Patents
A kind of Handwritten Digit Recognition method based on gyroscope Download PDFInfo
- Publication number
- CN106127188B CN106127188B CN201610524482.XA CN201610524482A CN106127188B CN 106127188 B CN106127188 B CN 106127188B CN 201610524482 A CN201610524482 A CN 201610524482A CN 106127188 B CN106127188 B CN 106127188B
- Authority
- CN
- China
- Prior art keywords
- section
- zero
- angular speed
- point
- gesture
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 241001269238 Data Species 0.000 claims abstract description 8
- 210000000707 wrist Anatomy 0.000 claims abstract description 6
- 230000015572 biosynthetic process Effects 0.000 claims description 11
- 238000003786 synthesis reaction Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 3
- 230000005484 gravity Effects 0.000 claims description 3
- 238000005204 segregation Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 230000008901 benefit Effects 0.000 abstract description 4
- 238000005265 energy consumption Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 210000004247 hand Anatomy 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000003340 mental effect Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/37—Writer recognition; Reading and verifying signatures based only on signature signals such as velocity or pressure, e.g. dynamic signature recognition
- G06V40/376—Acquisition
-
- 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
- G06F3/014—Hand-worn input/output arrangements, e.g. data gloves
-
- 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
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/033—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor
- G06F3/0346—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/37—Writer recognition; Reading and verifying signatures based only on signature signals such as velocity or pressure, e.g. dynamic signature recognition
- G06V40/382—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/37—Writer recognition; Reading and verifying signatures based only on signature signals such as velocity or pressure, e.g. dynamic signature recognition
- G06V40/394—Matching; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The present invention provides a kind of Handwritten Digit Recognition method based on gyroscope, acquires angular speed by the gyroscope being worn in wrist；Effective gesture data is intercepted using the method for three threshold values searched for backward forward；Selected characteristic amount simultaneously calculates；By the characteristic quantity of calculating, this 10 digital gesture datas are compared with preset 0 ~ 9, which number judge the corresponding writing of effective gesture data section is.Utilize the method for the present invention, it is only necessary in weared on wrist gyroscope, have the advantages that wearable effect it is good, it is at low cost, without other sensors, low in energy consumption.
Description
Technical field
The invention belongs to gesture identification fields, and in particular to a kind of Handwritten Digit Recognition method based on gyroscope.
Background technique
With the development of science and technology, public lives and amusement develop progressively towards intelligence, convenient and fast direction.Hand
Gesture is the operation and exchange way that people are commonly used, and using advanced sensing equipment and mode identification technology, is carried out to gesture
Exchange and remote control may be implemented in identification.Particularly with handicapped crowd, become using gesture progress remote control urgent
The demand cut.
Arabic numerals are a kind of symbols for having and determining specification stroke, are known by most people, easily operated, conveniently
Memory, and it is the control interface (such as: television set) inputted that most equipment, which is integrated with Arabic numerals,.By handwritten numeral rail
Mark is consistent as interaction gesture with the routine use of user habit and mental model.
There is also several obvious disadvantages for current gesture identification method: 1) there are mainly two types of technologies for gesture identification: being based on
The gesture identification of computer vision and sensor-based gesture identification.Gesture identification based on computer vision is by external disturbance
Larger (such as illumination condition variation), and (such as actor's dressing or carrying weight), or even activity are influenced by actor's shape
Range is also limited by camera.2) various sensors are used for gesture identification, but still be difficult solve comfort level it is poor, knowledge
Caused by other range is limited, discrimination is low, expensive, number of sensors is more the problems such as high power consumption.3) algorithm complexity is high,
It is difficult to carry out operation under conditions of hardware is limited.
Summary of the invention
The technical problem to be solved by the present invention is providing a kind of Handwritten Digit Recognition method based on gyroscope, having can
Dress effect it is good, it is at low cost, without other sensors, advantage low in energy consumption.
A kind of technical solution taken by the invention to solve the above technical problem are as follows: handwritten numeral knowledge based on gyroscope
Other method, it is characterised in that: it the following steps are included:
S1, angular speed is acquired by the gyroscope being worn in wrist, wherein each sampled point corresponds to one group of angular speed group,
Every group of angular speed group includes x-axis angular speed, y-axis angular speed and z-axis angular speed, and x-axis is the direction horizontally to the right of human body, and z-axis is
Gravity direction, the direction horizontally forward of y-axis human body；Calculate separately the synthesis angular speed of each sampled point；
The interception of S2, effective gesture data: using the method for three threshold values searched for backward forward, to the value of accumulated angle speed
ω is temporally intercepted, and judges the Origin And Destination of effective gesture, intercepts the effective gesture data section of several segments；Every section of effective hand
Gesture data segment includes several continuous sampling points and its corresponding angular speed group and synthesis angular speed；
The selection of S3, characteristic quantity:
In effective gesture data section, the angular speed of each axis is made of 3 class subsegments: positive section, negative section and zero section；Positive section
Is defined as: in the data segment, the value of data is all larger than 0, and maximum value is higher than first threshold；Negative section is defined as: at this
In data segment, the value of data is respectively less than 0, and minimum value is lower than second threshold；Zero section is defined as: in the data segment, data
Value change between first threshold and second threshold；
Choose following 5 characteristic quantities: (1) the effectively number T of gesture data section sampled point1, as from the effective gesture of the section
In data segment, effective gesture starts sampled point to the total number for terminating sampled point；(2) x-axis angular velocity omegaxSegments T2, it is
The positive section of x-axis angular speed, negative section, zero section of total number；(3) z-axis angular velocity omegazSegments T3, be z-axis angular speed positive section,
Negative section, zero section of total number；(4) zero section of position T of x-axis angular speed4, the position occurred according to the zero of x-axis angular speed section is
First section, endpiece, interlude are identified respectively again without zero section；(5) zero section of position T of z-axis angular speed5, according to z-axis
The position of zero section of appearance of angular speed is first section, endpiece, interlude again without zero section, is identified respectively；
The calculating of S4, characteristic quantity:
According to the definition of each characteristic quantity, calculated separately from effective gesture data section that S2 is obtained；
The identification of S5, handwritten numeral:
By the characteristic quantity of calculating, this 10 digital gesture datas are compared with preset 0-9, judge effective gesture number
Which number what it is according to the corresponding writing of section is.
According to the above method, the S5 comprising the following steps:
The foundation of S5-1, decision-making technique:
6 kinds of decision-making techniques are established, are represented sequentially as with the feature vector sequence of calculating: W1=[T2,T4,T3,T5], W2=
[T3,T5,T2,T4], W3=[T2,T3,T5,T4], W4=[T2,T3,T4,T5], W5=[T3,T2,T4,T5], W6=[T3,T2,T5,
T4]；
The dynamic select of S5-2, decision-making technique:
By preset 0-9, this 10 digital gesture datas are divided into 4 classes, respectively S1={ 1 }, S2={ 0,6,7 }, S3=
{ 2,3,9 }, S4={ 4,5,8 }；This corresponding decision-making technique of 4 class is respectively as follows: S1Corresponding aritrary decision method, S2Corresponding W1, S3It is right
Answer W2, S4Corresponding W2；
Utilize T1Mean value, T1Median Q2、T1First quartile Q1、T1Third quartile Q3This four categories
Property, fuzzy clustering is carried out according to this four attributes, according to maximum membership grade principle, it is believed that object to be identified T1It is opposite to be under the jurisdiction of S1-
S4In which kind of, select such corresponding decision-making technique；
S5-3, according to selected decision-making technique, feature vector is successively calculated, with this 10 digital hands of preset 0-9
Gesture data are compared, which number judge the corresponding writing of effective gesture data section is.
According to the above method, in the S2: the judgement of the starting point of effective gesture: choosing 3 threshold values from small to largePass throughThe judgement that gesture motion starts is carried out, selection is greater thanFirst sampled point, adopted
Sampling point Ab；Then search is less than forwardFirst sampled point, obtain sampled point As；If AbWith AsThe two sampled points
Distance is less than distance threshold DA, then determine As for the starting point of effect gesture；Otherwise, from AbStart search forward to be less thanFirst
A sampled point obtains time point Am, by AmIt is set as the starting point of effective gesture；
The judgement of the terminal of effective gesture: 3 threshold values from small to large are chosenChoose meet with thereafter
The average value of the synthesis angular speed sum of N number of sampled point is less than threshold valueTime point Bs；Then from point BsForward search be greater than compared with
Big threshold valueFirst sampled point, obtain sampled point Bb；If BsWith BbDistance be less than distance threshold DB, then by BsIt is set as
The terminal of effective gesture；Otherwise, from BbStart search backward to be less thanFirst sampled point, obtain sampled point Bm, by BmIf
For the terminal of effective gesture.
According to the above method, zero section of position T of the x-axis angular speed4, zero section when x-axis angular speed appears in first section,
T4It is identified as 1；Zero section when x-axis angular speed appears in endpiece, T4It is identified as -1；Zero section when x-axis angular speed appears in centre
Section, T4It is identified as 0；When x-axis angular speed is without zero section, T4It is identified as 2；
Zero section of position T of the z-axis angular speed5Zero section of position T of identification method and x-axis angular speed4It is identical.
According to the above method, the T2Calculation method it is as follows:
1. finding approximate zero point: it is different first with the symbol of zero point two sides numerical value, sampled point section existing for zero point is found,
The lesser point of absolute value in two interval points in the sampled point section is taken to be used as approximate zero point, by all approximate zero points according to sampling
Time sequencing sequence, obtains approximate xero-sequence；
2. removal head and the tail zero point: being scanned for respectively from the head and the tail of approximate xero-sequence, removal is at a distance of the approximation zero less than 5
Point updates approximate xero-sequence；
3. finding zero section of non-zero section section: the section of the adjacent approximate zero point composition of each of pairing approximation xero-sequence first
It scans for, the x-axis angular speed absolute value of some point is greater than threshold value Ω if it existst, then the section is non-zero section section, enables this non-
Zero section of section is [a, b]；The angular speed absolute value of first section is continuously less than threshold value ΩtPoints be n1, the angular speed absolute value of tail end
Continuously it is less than threshold value ΩtPoints be n2If n1>n2And n1Greater than threshold value N, then it is located at first section for zero section, and waypoint is a+n1；
If n1<n2And n2Greater than threshold value N, then it is located at endpiece, waypoint b-n for zero section2；Otherwise the non-zero section section is not present zero section；
Waypoint is added in approximate xero-sequence, updates approximate xero-sequence again；
4. removing continuous zero section of segregation section: if there is continuous zero section, waypoint therebetween is removed, it is made to be combined into one section, and
Corresponding waypoint is deleted from approximate xero-sequence；
5. the number of the point in final approximate xero-sequence subtracts 1, as T2；
The T3Calculation method and T2It is identical.
The invention has the benefit that
1, utilize the method for the present invention, it is only necessary in weared on wrist gyroscope, have wearable effect it is good, it is at low cost,
Without other sensors, advantage low in energy consumption.
2, further, gesture identification is carried out by the method for decision tree, decision-making technique is established according to statistical analysis first,
Then fuzzy diagnosis is carried out according to fuzzy clustering method, selects corresponding decision-making technique to be identified, so as to as far as possible
Digital gesture is identified using time-consuming shortest method, is conducive to the raising of real-time, algorithm complexity is solved and causes
Real-time difference and the serious problem of computer dependence.
Detailed description of the invention
Fig. 1 is the operation overview flow chart of one embodiment of the invention.
Fig. 2 is the schematic diagram that effective gesture of one embodiment of the invention intercepts.
Fig. 3 is the program flow diagram that effective gesture of one embodiment of the invention intercepts.
Fig. 4 is the classification schematic diagram of effective gesture section of one embodiment of the invention.
Fig. 5 is the feature extraction algorithm schematic diagram of one embodiment of the invention.
Fig. 6 is the feature T of one embodiment of the invention2With T3Program flow diagram.
Specific embodiment
Below with reference to specific example and attached drawing, the present invention will be further described.
The present invention improves a kind of Handwritten Digit Recognition method based on gyroscope, as shown in Figure 1, comprising the following steps:
S1, angular speed is acquired by the gyroscope being worn in wrist, wherein each sampled point corresponds to one group of angular speed group,
Every group of angular speed group includes x-axis angular speed, y-axis angular speed and z-axis angular speed, and x-axis is the direction horizontally to the right of human body, and z-axis is
Gravity direction, the direction horizontally forward of y-axis human body；Calculate separately the synthesis angular speed of each sampled point.
The interception of S2, effective gesture data: using the method for three threshold values searched for backward forward, to the value of accumulated angle speed
ω is temporally intercepted, and judges the Origin And Destination of effective gesture, intercepts the effective gesture data section of several segments；Every section of effective hand
Gesture data segment includes several continuous sampling points and its corresponding angular speed group and synthesis angular speed.
Refinement, as shown in Figures 2 and 3, in the S2: the judgement of the starting point of effective gesture: choosing 3 from small to large
Threshold valuePass throughThe judgement that gesture motion starts is carried out, selection is greater thanFirst sampled point,
Obtain sampled point Ab；Then search is less than forwardFirst sampled point, obtain sampled point As；If AbWith AsThe two are adopted
The distance of sampling point is less than distance threshold DA, then determine As for the starting point of effect gesture；Otherwise, when illustrating that gesture starts, manpower is trembled
Dynamic error is larger, from AbStart search forward to be less thanFirst sampled point, obtain time point Am, by AmIt is set as effective gesture
Starting point；
The judgement of the terminal of effective gesture: 3 threshold values from small to large are chosenChoose meet with thereafter
The average value of the synthesis angular speed sum of N number of sampled point is less than threshold valueTime point Bs；Then from point BsForward search be greater than compared with
Big threshold valueFirst sampled point, obtain sampled point Bb；If BsWith BbDistance be less than distance threshold DB, then by BsIt is set as
The terminal of effective gesture；Otherwise, at the end of illustrating gesture, the jitter error of manpower is larger, from BbStart search backward to be less than
First sampled point, obtain sampled point Bm, by BmIt is set as the terminal of effective gesture.
The selection of S3, characteristic quantity:
As shown in figure 4, carrying out the acquisition of gesture data in order to analyze data.It is found through analysis, in effective gesture number
According in section, the angular speed of each axis is made of 3 class subsegments: positive section, negative section and zero section；Positive section is defined as: in the data segment
In, the value of data is all larger than 0, and maximum value is higher than first threshold；Negative section is defined as: in the data segment, the value of data is equal
Less than 0, and minimum value is lower than second threshold；Zero section is defined as: in the data segment, the value of data is in first threshold and second
Change between threshold value.
Choose following 5 characteristic quantities: (1) the effectively number T of gesture data section sampled point1, as from the effective gesture of the section
In data segment, effective gesture starts sampled point to the total number for terminating sampled point；Due to there is the regulation of limitation gesture speed,
So the number of effectively gesture data section sampled point is able to reflect out the length of gesture to a certain extent.(2) x-axis angular speed
ωxSegments T2, it is positive section, the negative section, zero section of total number of x-axis angular speed.(3) z-axis angular velocity omegazSegments T3, it is
The positive section of z-axis angular speed, negative section, zero section of total number.(4) zero section of position T of x-axis angular speed4, according to x-axis angular speed
The position of zero section of appearance is first section, endpiece, interlude again without zero section, is identified respectively.(5) zero section of z-axis angular speed
Position T5, the position occurred according to the zero of z-axis angular speed section is first section, endpiece, interlude again without zero section, is carried out respectively
Mark.
In the present embodiment, zero section of position T of the x-axis angular speed4, zero section when x-axis angular speed appears in first section,
T4It is identified as 1；Zero section when x-axis angular speed appears in endpiece, T4It is identified as -1；Zero section when x-axis angular speed appears in centre
Section, T4It is identified as 0；When x-axis angular speed is without zero section, T4It is identified as 2.
Zero section of position T of the z-axis angular speed5Zero section of position T of identification method and x-axis angular speed4It is identical.
The calculating of S4, characteristic quantity:
According to the definition of each characteristic quantity, calculated separately from effective gesture data section that S2 is obtained.
Wherein, T1Calculating can be by being obtained in the valid data after intercepting.Note obtains its sampling number from valid data section
Purpose average length of time is t1。
As shown in Figure 5 and Figure 6, the T2Calculation method it is as follows:
1. finding approximate zero point: since sampled point is discrete, so not ensuring that there are a certain sampled point be zero point.
However, it is possible to which the symbol first with zero point two sides numerical value is different, sampled point section existing for zero point is found, the sampled point section is taken
Two interval points in the lesser point of absolute value as approximate zero point, all approximate zero points are sorted according to sample time order,
Obtain approximate xero-sequence.
For example, approximate xero-sequence in Fig. 5 are as follows: { 14,23,25,46,60,62 }.For convenience of calculation, then increase
Gesture data head and the tail point enters approximate xero-sequence, obtains: { 1,14,23,25,46,60,62,63 }.
2. removing head and the tail zero point: when due to intercepting effective gesture, can inevitably have error, so to the approximate zero point of head and the tail
It tests.It is scanned for respectively from the head and the tail of approximate xero-sequence, removal updates approximation zero at a distance of the approximate zero point less than 5
Point sequence.In Fig. 5, removal point 62 and point 63, updated approximation xero-sequence is { 1,14,23,25,46,60 }.
3. finding zero section of non-zero section section: due to the certain point in the not necessarily above-mentioned sequence of zero section of waypoint, institute
It is scanned for the section that the adjacent approximate zero point of each of pairing approximation xero-sequence first forms, if it exists the x-axis angle of some point
Speed absolute value is greater than threshold value Ωt, then the section is non-zero section section, and enabling the non-zero section section is [a, b]；The angular speed of first section
Absolute value is continuously less than threshold value ΩtPoints be n1, the angular speed absolute value of tail end is continuously less than threshold value ΩtPoints be n2If
n1>n2And n1Greater than threshold value N, then it is located at first section for zero section, and waypoint is a+n1；If n1<n2And n2Greater than threshold value N, then zero
Section is located at endpiece, waypoint b-n2；Otherwise the non-zero section section is not present zero section；Waypoint is added into approximate xero-sequence
In, approximate xero-sequence is updated again.
In Fig. 5, N=5, Ω are takent=20, section [25,46] are non-zero section, and there are zero for the first half in the section
Section, zero section of waypoint are 36, are added in approximate xero-sequence, again updated approximate xero-sequence be 1,14,
23,25,36,46,60}。
4. removing continuous zero section of segregation section: due to the uncertainty of zero section of symbol, so approximate zero point sequence derived above
Column may be separated continuous zero section.If there is continuous zero section, waypoint therebetween is removed, it is made to be combined into one section, and from
Corresponding waypoint is deleted in approximate xero-sequence.
In Fig. 5, [14,23], [23,25], [25,36] are zero section, so removal waypoint 23 and 25, obtains approximation zero
Point sequence is { 1,14,36,46,60 }.
5. the number of the point in final approximate xero-sequence subtracts 1, as T2.In the present embodiment, T2For 5-1=4.
The T3Calculation method and T2It is identical.
Remember above-mentioned calculating T2With T3Each step altogether expend average length of time be respectively t2And t3, t2≈t3。
Each the section being made of the adjacent sectional point 4. obtained, detects the requirement whether each section meets zero section, i.e.,
Whether the absolute value of point is less than threshold value Ωt, zero section of position can be obtained, and be identified to it according to zero section of position.Fig. 5
In, the 2nd section is zero section.Remember from T2With T3Calculation method in, obtain T4With T5Average consuming time is long is respectively t4And t5,
In, t4≈t5。
The identification of S5, handwritten numeral:
By the characteristic quantity of calculating, this 10 digital gesture datas are compared with preset 0-9, judge effective gesture number
Which number what it is according to the corresponding writing of section is.
Preferably, in order to skipping the calculating of redundancy feature amount, make as far as possible by dynamic decision Method
Digital gesture is identified with time-consuming shortest method, the S5 comprising the following steps:
The foundation of S5-1, decision-making technique:
6 kinds of decision-making techniques are established, are represented sequentially as with the feature vector sequence of calculating: W1=[T2,T4,T3,T5], W2=
[T3,T5,T2,T4], W3=[T2,T3,T5,T4], W4=[T2,T3,T4,T5], W5=[T3,T2,T4,T5], W6=[T3,T2,T5,
T4]。
The dynamic select of S5-2, decision-making technique:
By preset 0-9, this 10 digital gesture datas are divided into 4 classes, respectively S1={ 1 }, S2={ 0,6,7 }, S3=
{ 2,3,9 }, S4={ 4,5,8 }；This corresponding decision-making technique of 4 class is respectively as follows: S1Corresponding aritrary decision method, S2Corresponding W1, S3It is right
Answer W2, S4Corresponding W2；
Utilize T1Mean value, T1Median Q2、T1First quartile Q1、T1Third quartile Q3This four categories
Property, fuzzy clustering is carried out according to this four attributes, according to maximum membership grade principle, it is believed that object to be identified T1It is opposite to be under the jurisdiction of S1-
S4In which kind of, select such corresponding decision-making technique；
S5-3, according to selected decision-making technique, feature vector is successively calculated, with this 10 digital hands of preset 0-9
Gesture data are compared, which number judge the corresponding writing of effective gesture data section is.
Above-mentioned decision-making technique to establish principle as follows:
Feature vector T=[T is constituted by above-mentioned 5 characteristic quantities1,T2,T3,T4,T5].But statistical analysis is found, and
Not all digital gesture needs this five characteristic quantities that can just identify.Note can most identify that the feature vector of the number is fastly
Tf, required time tf.Following table has counted this ten digital T2,T3,T4,T5,TfValue.
Due to T2With T4、T3With T5Dependence limitation, carry out Handwritten Digit Recognition be divided into six kinds of decision-making techniques, be denoted as Wi
(i=1~6).[T is expressed as with the feature vector sequence calculated respectively2,T4,T3,T5],[T3,T5,T2,T4],[T2,T3,T5,
T4], [T2,T3,T4,T5],[T3,T2,T4,T5],[T3,T2,T5,T4].When note identifies the cost of each number using every kind of method
Between be twi(i=1~6) remember twiWith tfDifference be Δ twi(i=1~6).
There are some numbers in identification process in every kind of method, calculates extra characteristic quantity, cause real-time poor
And energy consumption increases.So the method that can use dynamic select, in advance differentiates gesture data, approximate range is obtained
Afterwards, then the characteristic quantity of the calculating is determined.
Utilize T1Mean value, median Q2, first quartile Q1, third quartile Q3This four attributes, according to this four
A attribute carries out fuzzy clustering.10 digital gesture datas are divided into 4 classes, respectively { 1 }, { 0,6,7 }, { 2,3,9 }, 4,
5,8 } S is used respectively1~S4It indicates.It can be by T1By calculating affiliated all kinds of degree of membership, the class of done gesture is faintly judged
Not.According to maximum membership grade principle, it is believed that object to be identified T1It is opposite to be under the jurisdiction of class Sk, select such decision-making technique Wk.So
Afterwards, the decision-making technique of every class is determined.With Δ t digital in every classw1With minimum principle, S1Any one can be selected, S2It selects
W1, S3Select W2,S4Select W2.It, can be as much as possible using time-consuming shortest method logarithm by being dynamically selected decision-making technique
Word gesture is identified, the raising of real-time is conducive to.
Specifically, Fig. 5 is the data graphs of gesture " 4 ".The data length T of gesture1It is 64, according to maximum membership degree original
Then, which is assigned into S4Class, then by S4The characteristic quantity computation sequence W of class2=[T3,T5,T2,T4] calculated.According to this
Method identifies gesture " 4 ", it is only necessary to calculate characteristic quantity T3And T5, save and calculate T2And T4Elapsed time.
Above embodiments are merely to illustrate design philosophy and feature of the invention, and its object is to make technology in the art
Personnel can understand the content of the present invention and implement it accordingly, and protection scope of the present invention is not limited to the above embodiments.So it is all according to
It is within the scope of the present invention according to equivalent variations made by disclosed principle, mentality of designing or modification.
Claims (4)
1. a kind of Handwritten Digit Recognition method based on gyroscope, it is characterised in that: it the following steps are included:
S1, angular speed is acquired by the gyroscope being worn in wrist, wherein each sampled point one group of angular speed group of correspondence, every group
Angular speed group includes x-axis angular speed, y-axis angular speed and z-axis angular speed, and x-axis is the direction horizontally to the right of human body, and z-axis is gravity
Direction, the direction horizontally forward of y-axis human body；Calculate separately the synthesis angular speed of each sampled point；
The interception of S2, effective gesture data: using the method for three threshold values searched for backward forward, the value ω of accumulated angle speed is pressed
Time is intercepted, and judges the Origin And Destination of effective gesture, intercepts the effective gesture data section of several segments；Every section of effective gesture number
It include several continuous sampling points and its corresponding angular speed group and synthesis angular speed according to section；
In the S2: the judgement of the starting point of effective gesture: choosing 3 threshold values from small to largePass through
The judgement that gesture motion starts is carried out, the value ω for choosing synthesis angular speed is greater thanFirst sampled point, obtain sampled point
Ab；Then the value ω of search synthesis angular speed is less than Ω forwards AFirst sampled point, obtain sampled point As；If AbWith AsThis
The distance of two sampled points is less than distance threshold DA, then determine AsFor the starting point for imitating gesture；Otherwise, from AbStart search forward to close
It is less than at the value ω of angular speedFirst sampled point, obtain time point Am, by AmIt is set as the starting point of effective gesture；
The judgement of the terminal of effective gesture: 3 threshold values from small to large are chosenIt chooses and meets and n thereafter
The average value of the synthesis angular speed sum of sampled point is less than threshold value Ωs BTime point Bs；Then from point BsSearch is greater than larger forward
Threshold valueFirst sampled point, obtain sampled point Bb；If BsWith BbDistance be less than distance threshold DB, then by BsIt has been set as
Imitate the terminal of gesture；Otherwise, from BbStart search backward to be less thanFirst sampled point, obtain sampled point Bm, by BmIt is set as
The terminal of effective gesture；
The selection of S3, characteristic quantity:
In effective gesture data section, the angular speed of each axis is made of 3 class subsegments: positive section, negative section and zero section；Positive section is determined
Justice are as follows: in the data segment, the value of data is all larger than 0, and maximum value is higher than first threshold；Negative section is defined as: in the data
Duan Zhong, the value of data is respectively less than 0, and minimum value is lower than second threshold；Zero section is defined as: in the data segment, the value of data
Change between first threshold and second threshold；
Choose following 5 characteristic quantities: (1) the effectively number T of gesture data section sampled point1, as from the effective gesture data section of the section
In, effective gesture starts sampled point to the total number for terminating sampled point；(2) x-axis angular velocity omegaxSegments T2, it is x-axis angle
The positive section of speed, negative section, zero section of total number；(3) z-axis angular velocity omegazSegments T3, be the positive section of z-axis angular speed, negative section,
Zero section of total number；(4) zero section of position T of x-axis angular speed4, according to the zero of x-axis angular speed section occur position be first section,
Endpiece, interlude are identified respectively again without zero section；(5) zero section of position T of z-axis angular speed5, according to z-axis angular speed
Zero section appearance position be first section, endpiece, interlude again without zero section, be identified respectively；
The calculating of S4, characteristic quantity:
According to the definition of each characteristic quantity, calculated separately from effective gesture data section that S2 is obtained；
The identification of S5, handwritten numeral:
By the characteristic quantity of calculating, this 10 digital gesture datas are compared with preset 0-9, judge effective gesture data section
Which number corresponding writing is.
2. the Handwritten Digit Recognition method according to claim 1 based on gyroscope, it is characterised in that: the S5 is specific
It comprises the steps of:
The foundation of S5-1, decision-making technique:
6 kinds of decision-making techniques are established, are represented sequentially as with the feature vector sequence of calculating: W1=[T2,T4,T3,T5], W2=[T3,T5,
T2,T4], W3=[T2,T3,T5,T4], W4=[T2,T3,T4,T5], W5=[T3,T2,T4,T5], W6=[T3,T2,T5,T4]；
The dynamic select of S5-2, decision-making technique:
By preset 0-9, this 10 digital gesture datas are divided into 4 classes, respectively S1={ 1 }, S2={ 0,6,7 }, S3=2,3,
9 }, S4={ 4,5,8 }；This corresponding decision-making technique of 4 class is respectively as follows: S1Corresponding aritrary decision method, S2Corresponding W1, S3Corresponding W2,
S4Corresponding W2；
Utilize T1Mean value, T1Median Q2、T1First quartile Q1、T1Third quartile Q3This four attributes, according to
Fuzzy clustering is carried out according to this four attributes, according to maximum membership grade principle, it is believed that object to be identified T1It is opposite to be under the jurisdiction of S1-S4In
Which kind of, selects such corresponding decision-making technique；
S5-3, according to selected decision-making technique, feature vector is successively calculated, with this 10 digital gesture numbers of preset 0-9
According to being compared, which number judge the corresponding writing of effective gesture data section is.
3. the Handwritten Digit Recognition method according to claim 1 or 2 based on gyroscope, it is characterised in that: the x-axis
Zero section of position T of angular speed4, zero section when x-axis angular speed appears in first section, T4It is identified as 1；Zero section when x-axis angular speed goes out
Present endpiece, T4It is identified as -1；Zero section when x-axis angular speed appears in interlude, T4It is identified as 0；When x-axis angular speed is without zero
Section, T4It is identified as 2；
Zero section of position T of the z-axis angular speed5Zero section of position T of identification method and x-axis angular speed4It is identical.
4. the Handwritten Digit Recognition method according to claim 1 or 2 based on gyroscope, it is characterised in that: the T2's
Calculation method is as follows:
1. finding approximate zero point: it is different first with the symbol of zero point two sides numerical value, sampled point section existing for zero point is found, this is taken
The lesser point of absolute value is as approximate zero point in two interval points in sampled point section, by all approximate zero points according to the sampling time
Sequence sorts, and obtains approximate xero-sequence；
2. removal head and the tail zero point: it scans for, removes at a distance of the approximate zero point less than 5 from the head and the tail of approximate xero-sequence respectively,
Update approximate xero-sequence；
3. finding zero section of non-zero section section: the section of the adjacent approximate zero point composition of each of pairing approximation xero-sequence first carries out
Search, the x-axis angular speed absolute value of some point is greater than threshold value Ω if it existst, then the section is non-zero section section, enables the non-zero section
Section is [a, b]；The angular speed absolute value of first section is continuously less than threshold value ΩtPoints be n1, the angular speed absolute value of tail end is continuous
Less than threshold value ΩtPoints be n2If n1>n2And n1Greater than threshold value N, then it is located at first section for zero section, and waypoint is a+n1；If n1
<n2And n2Greater than threshold value N, then it is located at endpiece, waypoint b-n for zero section2；Otherwise the non-zero section section is not present zero section；It will divide
Section point adds in approximate xero-sequence, updates approximate xero-sequence again；
4. removing continuous zero section of segregation section: if there is continuous zero section, removing waypoint therebetween, it is made to be combined into one section, and from close
Like deleting corresponding waypoint in xero-sequence；
5. the number of the point in final approximate xero-sequence subtracts 1, as T2；
The T3Calculation method and T2It is identical.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610524482.XA CN106127188B (en) | 2016-07-05 | 2016-07-05 | A kind of Handwritten Digit Recognition method based on gyroscope |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610524482.XA CN106127188B (en) | 2016-07-05 | 2016-07-05 | A kind of Handwritten Digit Recognition method based on gyroscope |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106127188A CN106127188A (en) | 2016-11-16 |
| CN106127188B true CN106127188B (en) | 2019-05-17 |
Family
ID=57469500
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610524482.XA Expired - Fee Related CN106127188B (en) | 2016-07-05 | 2016-07-05 | A kind of Handwritten Digit Recognition method based on gyroscope |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106127188B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106950927B (en) * | 2017-02-17 | 2019-05-17 | 深圳大学 | A kind of method and intelligent wearable device controlling smart home |
| CN107153780B (en) * | 2017-05-05 | 2020-06-09 | 西安交通大学苏州研究院 | Writing behavior feature authentication method based on wrist wearable electronic equipment |
| CN110569800B (en) * | 2019-09-10 | 2022-04-12 | 武汉大学 | Detection method of handwriting signal |
| CN117075742B (en) * | 2023-10-17 | 2024-01-30 | 深圳市魔样科技有限公司 | Intelligent finger ring control method based on gesture recognition |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102945362A (en) * | 2012-10-18 | 2013-02-27 | 中国科学院计算技术研究所 | Isomerous data fusion based coordinated gesture recognition method and system of sensor |
| CN103513788A (en) * | 2013-09-13 | 2014-01-15 | 广东欧珀移动通信有限公司 | Gesture recognition method and system based on gyroscope sensor and mobile terminal |
| US9129400B1 (en) * | 2011-09-23 | 2015-09-08 | Amazon Technologies, Inc. | Movement prediction for image capture |
| CN105138133A (en) * | 2015-09-14 | 2015-12-09 | 李玮琛 | Biological signal gesture recognition device and method |
-
2016
- 2016-07-05 CN CN201610524482.XA patent/CN106127188B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9129400B1 (en) * | 2011-09-23 | 2015-09-08 | Amazon Technologies, Inc. | Movement prediction for image capture |
| CN102945362A (en) * | 2012-10-18 | 2013-02-27 | 中国科学院计算技术研究所 | Isomerous data fusion based coordinated gesture recognition method and system of sensor |
| CN103513788A (en) * | 2013-09-13 | 2014-01-15 | 广东欧珀移动通信有限公司 | Gesture recognition method and system based on gyroscope sensor and mobile terminal |
| CN105138133A (en) * | 2015-09-14 | 2015-12-09 | 李玮琛 | Biological signal gesture recognition device and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106127188A (en) | 2016-11-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106127188B (en) | A kind of Handwritten Digit Recognition method based on gyroscope | |
| CN109146921B (en) | Pedestrian target tracking method based on deep learning | |
| CN107168527B (en) | The first visual angle gesture identification and exchange method based on region convolutional neural networks | |
| CN103246891B (en) | A kind of Chinese Sign Language recognition methods based on Kinect | |
| CN109858406B (en) | Key frame extraction method based on joint point information | |
| CN105718878B (en) | The aerial hand-written and aerial exchange method in the first visual angle based on concatenated convolutional neural network | |
| RU2711029C2 (en) | Touch classification | |
| CN103226388B (en) | A kind of handwriting sckeme based on Kinect | |
| CN111709310B (en) | Gesture tracking and recognition method based on deep learning | |
| CN106096662B (en) | Human motion state identification based on acceleration transducer | |
| CN104915351B (en) | Picture sort method and terminal | |
| CN103824051B (en) | Local region matching-based face search method | |
| CN110119689A (en) | A kind of face beauty prediction technique based on multitask transfer learning | |
| CN107909042B (en) | continuous gesture segmentation recognition method | |
| CN106778456A (en) | A kind of optimization method and device of handwriting input | |
| CN109558855B (en) | A kind of space gesture recognition methods combined based on palm contour feature with stencil matching method | |
| CN109101864A (en) | The upper half of human body action identification method returned based on key frame and random forest | |
| Kalsh et al. | Sign language recognition system | |
| CN113378891A (en) | Urban area relation visual analysis method based on track distribution representation | |
| CN105335711B (en) | Fingertip Detection under a kind of complex environment | |
| CN109766873A (en) | A kind of pedestrian mixing deformable convolution recognition methods again | |
| CN110070106A (en) | Smog detection method, device and electronic equipment | |
| CN110189362A (en) | The efficient method for tracking target of network is fought from coding based on multiple-limb | |
| CN109614990A (en) | A kind of object detecting device | |
| CN107220634B (en) | Based on the gesture identification method for improving D-P algorithm and multi-template matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190517 |