Summary of the invention
In order to solve the problems of the technologies described above, the object of this invention is to provide a kind of fuzzy sign Language Recognition Method that can improve a kind of data glove of the accuracy rate of gesture identification.
The technical solution adopted in the present invention is:
A fuzzy sign Language Recognition Method for data glove, comprises the following steps:
A, obtain hand motion data Fuzzy Processing is carried out to it, obtain gesture frame sequence;
B, according to sign language database and probability database, identifying processing is carried out to the gesture frame sequence obtained, obtain gesture frame sequence recognition result.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, described steps A comprises:
The angle of bend of A1, each finger obtained in hand motion data, and according to the bending membership function preset, draw the case of bending of each corresponding finger;
A2, the palm angle of pitch obtained in hand motion data, and calculate according to the pitching membership function preset, what value of obtaining a result was maximum is corresponding pitch attitude;
A3, the palm pitch angle obtained in hand motion data, and calculate according to the inclination membership function preset, what value of obtaining a result was maximum is corresponding heeling condition;
A4, the palm crab angle obtained in hand motion data, and calculate according to the driftage membership function preset, what value of obtaining a result was maximum is corresponding driftage state;
A5, according to the pitch attitude, heeling condition and the driftage state that calculate, combine the rule preset, draw the palm of correspondence towards;
A6, according to palm towards with each case of bending pointed, draw gesture frame, and and then draw gesture frame sequence.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, it is characterized in that: described step B comprises:
B1, acquisition gesture frame sequence, extract gesture frame according to order from the beginning to the end;
B2, the gesture frame of extraction put into respectively successively corresponding node;
B3, from sign language database, extract all words corresponding to each gesture frame successively, and added in corresponding node, until all gesture frames all complete the retrieval of sign language database on gesture frame sequence;
B4, by the order respectively combination of two of adjacent two incidental words of node according to node, in combination, pointed to the words of next node by the words of a upper node;
B5, all combinations are indexed out respectively in probability database the probability of each combination;
B6, find out each combination composition sentence in probability and maximum sentence, draw gesture frame sequence recognition result.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the bending membership function in described steps A 1 is:
Wherein, X ∈ U
0, U
0represent digital flexion angle, U
0=[0,120], at U
0on set up three fuzzy set A of digital flexion angle
0represent that case of bending is the state of " stretching ", A
1=represent that case of bending is the state of " half is bending ", A
2=represent that case of bending is the state of " holding ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the pitching membership function in described steps A 2 is:
Wherein, x ∈ U
1, U
1represent the palm angle of pitch, U
1=[-90,90], at U
1on set up three fuzzy set B of the angle of pitch
0represent that the angle of pitch is the state of " bowing ", B
1=represent that the angle of pitch is the state of " level ", B
2=represent that the angle of pitch is the state of " facing upward ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the inclination membership function in described steps A 3 is:
Wherein, y ∈ U
2, U
2represent palm pitch angle, U
2=[-180,180], at U
2on set up three fuzzy set C at pitch angle
0represent that pitch angle is the state of " "Left"-deviationist ", C
1=represent that pitch angle is the state of " level ", C
2=represent that pitch angle is the state of " Right deviation ", C
3=represent that pitch angle is the state of " upset level ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the driftage membership function in described steps A 3 is:
Wherein, z ∈ U
3, U
3represent palm crab angle, U
3=[0,360], at U
3on set up three fuzzy set D of crab angle
0represent that crab angle is the state of " front ", D
1=represent that crab angle is the state on " right side ", D
2=represent that crab angle is the state of " afterwards ", D
3=represent that crab angle is the state on " left side ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, described sign language database is using digital flexion state and palm towards the gesture frame formed as index, and words corresponding to gesture is as indexed content.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the probability of the words combination in described probability database obtains for utilizing language model training tool SRILM.
The invention has the beneficial effects as follows:
The fuzzy sign Language Recognition Method of a kind of data glove of the present invention is by carrying out Fuzzy Processing to hand action data, effectively prevent because hand size differs the lower problem of the discrimination that causes, and by conjunction with sign language database and probability database, make the present invention can choose the identification of current gesture optimum according to front and back gesture, greatly improve the accuracy rate of gesture identification.
Embodiment
With reference to figure 1, the fuzzy sign Language Recognition Method of a kind of data glove of the present invention, comprises the following steps:
A, obtain hand motion data Fuzzy Processing is carried out to it, obtain gesture frame sequence;
B, according to sign language database and probability database, identifying processing is carried out to the gesture frame sequence obtained, obtain gesture frame sequence recognition result.
With reference to figure 2, as the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, described steps A comprises:
The angle of bend of A1, each finger obtained in hand motion data, and according to the bending membership function preset, draw the case of bending of each corresponding finger;
A2, the palm angle of pitch obtained in hand motion data, and calculate according to the pitching membership function preset, what value of obtaining a result was maximum is corresponding pitch attitude;
A3, the palm pitch angle obtained in hand motion data, and calculate according to the inclination membership function preset, what value of obtaining a result was maximum is corresponding heeling condition;
A4, the palm crab angle obtained in hand motion data, and calculate according to the driftage membership function preset, what value of obtaining a result was maximum is corresponding driftage state;
A5, according to the pitch attitude, heeling condition and the driftage state that calculate, combine the rule preset, draw the palm of correspondence towards;
A6, according to palm towards with each case of bending pointed, draw gesture frame, and and then draw gesture frame sequence.
With reference to figure 3, as the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, it is characterized in that: described step B comprises:
B1, acquisition gesture frame sequence, extract gesture frame according to order from the beginning to the end;
B2, the gesture frame of extraction put into respectively successively corresponding node;
B3, from sign language database, extract all words corresponding to each gesture frame successively, and added in corresponding node, until all gesture frames all complete the retrieval of sign language database on gesture frame sequence;
B4, by the order respectively combination of two of adjacent two incidental words of node according to node, in combination, pointed to the words of next node by the words of a upper node;
B5, all combinations are indexed out respectively in probability database the probability of each combination;
B6, find out each combination composition sentence in probability and maximum sentence, draw gesture frame sequence recognition result.
Such as, gesture sequence S has two gesture frames to be followed successively by A and B, suppose that gesture A has words " you " and " that ", gesture B has words " good " and " just ", supposes that the probability of " hello " is 0.0052, and the probability of " you just " is " 0.00045 ", the probability of " OK " is 0.0078, the probability of " that just " is 0.00032, then the recognition result of gesture sequence S is that sentence of maximum probability, and namely recognition result is " OK ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the bending membership function in described steps A 1 is:
Wherein, X ∈ U
0, U
0represent digital flexion angle, U
0=[0,120], at U
0on set up three fuzzy set A of digital flexion angle
0represent that case of bending is the state of " stretching ", A
1=represent that case of bending is the state of " half is bending ", A
2=represent that case of bending is the state of " holding ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the pitching membership function in described steps A 2 is:
Wherein, x ∈ U
1, U
1represent the palm angle of pitch, U
1=[-90,90], at U
1on set up three fuzzy set B of the angle of pitch
0represent that the angle of pitch is the state of " bowing ", B
1=represent that the angle of pitch is the state of " level ", B
2=represent that the angle of pitch is the state of " facing upward ".
If the palm angle of pitch is 42, the membership function substituted into by x=42 in formula two calculates, and draws B
0(42)=0.1, B
1(42)=0.3, B
2(42)=0, B
1>B
0>B
2, then the angle of pitch of the x value of this input is level.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the inclination membership function in described steps A 3 is:
Wherein, y ∈ U
2, U
2represent palm pitch angle, U
2=[-180,180], at U
2on set up three fuzzy set C at pitch angle
0represent that pitch angle is the state of " "Left"-deviationist ", C
1=represent that pitch angle is the state of " level ", C
2=represent that pitch angle is the state of " Right deviation ", C
3=represent that pitch angle is the state of " upset level ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the driftage membership function in described steps A 3 is:
Wherein, z ∈ U
3, U
3represent palm crab angle, U
3=[0,360], at U
3on set up three fuzzy set D of crab angle
0represent that crab angle is the state of " front ", D
1=represent that crab angle is the state on " right side ", D
2=represent that crab angle is the state of " afterwards ", D
3=represent that crab angle is the state on " left side ".
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, described sign language database is using digital flexion state and palm towards the gesture frame formed as index, and words corresponding to gesture is as indexed content.
As the further improvement of the fuzzy sign Language Recognition Method of described a kind of data glove, the probability of the words combination in described probability database obtains for utilizing language model training tool SRILM.
With reference to figure 4, wherein, a gesture frame sequence in the present invention refers to a complete sign language sentence, comprises N number of gesture frame.A gesture frame is made up of 4 bytes (32), wherein the 32nd reservation, 26th ~ 30 and 11st ~ 15 " palm towards " states being respectively left hand and the right hand, 16th ~ 25 and the 1st ~ the 10th are respectively the digital flexion state storing left hand and the right hand, one of them digital flexion state accounts for 2 positions, thumb, forefinger, middle finger, the third finger and little finger of toe sort respectively from a high position to low level, and the 0th is check bit.
Sign language frame sequence is arranged in chronological order by multiple sign language frame, a corresponding multiple word of sign language frame or word, because a gesture represents the different meanings at different linguistic context scapes, and gesture is carried out abstract in the method, made some only have the gesture of JND to be abstracted into identical sign language frame.And gesture identification part is exactly the statistical probability based on context of the sign language frame in sign language frame sequence extracted by most suitable words and form optimal sentence with other sign language frames in sign language frame sequence.
Sign language database and probability database must be established before Sign Language Recognition.
The foundation of sign language database refers to reference to " Chinese Sign Language " upper volume two, Part I introduction method the content of the inside is utilized to propose to go the state orientation of case of bending and the palm pointed and form sign language frame, and using the sign language frame of these 4 bytes (32) as index, and words corresponding to gesture is as indexed content, under words corresponding to identical gesture is placed on an index, when with this indexed search, the full content of this index will be drawn.As: " you " is the same with the gesture of " that ", then their index is the same, supposes that this index is A, will draw " you ", " that " during search A.
Probability database refers to the frequency of words appearance of single words in corpus, and and then occurs the set of frequency of another word after each word; So-called corpus refers to sign language aspect works and expressions for everyday use sentence or article.For single words and word with firmly occur another word probability be then utilize language model training tool SRILM to obtain.And they are stored with form below.Drawing the probability of this words or words combination with words or words synthetic rope, is 0.000000147332 as " greenery " index out probability, with " as " " appearance " index out probability " 0.000145517086 ".The index of probability database is based on two words, if 2-gram in can not find this two contaminations, then 1-gram find this two words independent respectively probability, be set to P1 and P2, then these two contamination probability are P=P1*P2*e, e is natural constant about 2.71828182845905.
More than that better enforcement of the present invention is illustrated, but the invention is not limited to described embodiment, those of ordinary skill in the art also can make all equivalent variations or replacement under the prerequisite without prejudice to spirit of the present invention, and these equivalent distortion or replacement are all included in the application's claim limited range.