CN105373215B - Dynamic radio gesture identification method with decoding is encoded based on gesture - Google Patents

Abstract

The present invention discloses a kind of dynamic radio gesture identification method encoded based on gesture with decoding, comprises the following steps：Construct basic gesture collection, basic gesture off-line training, gesture coding, order hand signal acquisition to be checked, the identification of order hand signal, hand signal decoding.The dynamic radio gesture identification method that the present invention uses, suitable for a variety of radio and ultrasonic wave gesture recognition systems characterized by hand signal amplitude, phase, direction of arrival or Doppler frequency shift, versatility is good, and high to the identification probability of wireless gesture, good reliability.

Description

Dynamic Wireless Gesture Recognition Method Based on Gesture Coding and Decoding

technical field

The invention belongs to the technical field of gesture recognition in human-computer interaction, in particular to a dynamic wireless gesture recognition method based on gesture coding and decoding with high recognition probability and good reliability.

Background technique

In various fields such as smart robots, smart TVs, smart phones, computers, game consoles, smart cars, and smart homes, using dynamic gestures to wirelessly operate computers has become a new type of human-computer interaction.

Existing dynamic wireless gesture recognition methods include wearable and bare hand gesture recognition systems, such as the literature "B.Raj, K.Kalgaonkar, C.Harrison, and P.Dietz, Ultrasonic Doppler sensing in HCI, IEEE Pervasive Computing, vol.11, no.2, pp.24-29, Feb.2012". Wearable systems need to wear gloves containing sensors such as accelerometers and magnetometers to recognize gestures, which is inconvenient to use. The bare-hand gesture recognition system uses non-physical contact to interact with the machine, and does not need to wear special gloves to recognize gestures, which is convenient to use. The most common means of realization are visual (light wave), ultrasonic, and radio wave identification systems.

The visual gesture system system uses the camera to capture images and recognize gestures, but there are the following limitations: 1. The recognition distance is short, and the image captured by the camera will become worse as the distance increases; 2. The system needs continuous lighting, and the system fails in the dark; 3. Sensitive to light changes, it is generally not suitable for mobile or outdoor environments; 4. The human-machine needs to be transmitted in line of sight, and there must be no obstructions between the camera and the human room; 5. The camera may invade user privacy.

Radio wave and ultrasonic bare hand gesture recognition technology has many advantages:

(1). No need to wear equipment, easy to use.

(2). Long-distance gesture control can be realized.

(3). Less affected by environmental factors such as temperature and brightness. Can be used in dark, outdoor environments.

(4). Can work in non-line-of-sight environment.

(5). Does not affect the user's privacy.

For details, refer to "S.Gupta, D.Morris, S.Patel, and D.Tan, Soundwave: using the doppler effect to sense gestures, Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, May 05-10, 2012, Austin, Texas , USA, 2012", "F.Adib and D.Katabi, See through walls with Wi-Fi! Proceedings of the ACMSIGCOMM 2013 conference on SIGCOMM, August 12-16, 2013, Hong Kong, China, 2013", "Q. Pu, S.Gupta, S.Gollakota, and S.Patel, Whole-home gesture recognition using wireless signals, ACM International Conference on Mobile Computing and Networking (MOBICOM), 2013", "F.Adib, Z.Kabelac, D.Katabi, and R.C. Miller, 3Dtracking via body radio reflections, Usenix NSDI, 2014" and "B. Kellogg, V. Talla, and S. Gollakota, Bringing gesture recognition to all devices, Usenix NSDI, 2014".

However, in the wireless channel where radio waves and ultrasonic waves are located, the correct recognition of gestures faces a series of challenges, including:

(1). Influence of the channel environment: There are many unfavorable factors such as interference, fading, and multipath in the wireless channel, which increases the probability of misrecognition of the wireless gesture recognition system.

(2). Impact on gesture design: The current gesture design is in a state of "one system, one method", lacking a systematic and reliable design method. It is often difficult to design a set of gestures that are highly discriminative so that even the smallest distance between two gestures in the set can be distinguished. This is the intrinsic reason for gesture misrecognition.

(3). Manipulator influence: Gestures are performed by humans, and humans may misoperate, resulting in wrong gesture commands.

On the other hand, in many applications, there is an urgent need for reliable gesture recognition technology with low probability of false recognition, such as bomb disposal robot control in military or anti-terrorism, military aircraft take-off and landing command, smart car control, etc. Therefore, there is a contradiction between the urgent need for high-reliability gesture recognition and the unreliable gesture recognition results.

In a word, the problems existing in the prior art are: the recognition probability is not high enough to meet the requirements of high-reliability gesture control, and the versatility is not strong.

Contents of the invention

The purpose of the present invention is to provide a dynamic wireless gesture recognition method based on gesture coding and decoding, which has high recognition probability, good reliability and strong versatility.

The technical solution to realize the object of the present invention is: a dynamic wireless gesture recognition method based on gesture encoding and decoding, comprising the following steps:

10) Constructing a basic gesture set: constructing a basic gesture set according to multiple basic gestures;

20) Offline training of basic gestures: collect each basic gesture for basic gestures, detect the signal emitted by the transmitter reflected by it multiple times, obtain the characteristics of each basic gesture, store all the characteristics of basic gestures, and establish basic gesture characteristics Library;

30) Gesture coding: coding basic gestures into command gestures corresponding to gesture control commands, constructing the command gestures includes basic gesture combination coding, error detection coding and error correction coding;

40) Acquisition of the command gesture signal to be checked: the transmitter transmits the signal, and the receiver receives the signal reflected by the command gesture to obtain the command gesture signal to be checked;

50) command gesture signal recognition: the command gesture signal to be checked is divided into multiple sub-gestures, each sub-gesture corresponds to an unknown basic gesture, and the command gesture signal recognition is realized through the recognition of each sub-gesture, and each sub-gesture recognition is to detect each sub-gesture The signal features of each sub-gesture are compared with the basic gesture features in turn to find out the basic gestures that match each sub-gesture, and these basic gestures are combined in turn to form the judgment result of the command gesture signal to be checked;

60) Gesture signal decoding: Perform error detection decoding and error correction decoding on the judgment result of the gesture signal to be detected to obtain a gesture control command corresponding to the gesture signal to be detected.

Compared with the prior art, the present invention has the remarkable advantages of high recognition probability, good reliability and strong versatility.

The above advantages are achieved by:

1. Simplify the types of gestures: use only a few basic gestures and make full use of the coding theory in communication to construct other gestures, avoiding the cumbersome gesture training process when the number of gestures is large; when the number of gestures in the basic gesture set is small, It is easy to construct basic gestures so that the minimum distance of the gesture set (that is, the distance between the two gestures with the smallest distance) is larger; and the greater the minimum distance between gestures, the smaller the probability of misjudgment;

2. Improve versatility: Gesture combination coding and error detection and error correction coding are independent of specific systems (such as ultrasound, radio, Doppler or signal amplitude), and are universal;

3. Occasional misjudgments of basic gestures can be corrected: after digitization of gestures, error correction coding technology in digital communication can be used. Misjudgment can be corrected to ensure correct gesture recognition.

The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Description of drawings

FIG. 1 is a flowchart of a dynamic wireless gesture recognition method based on gesture encoding and decoding in the present invention.

Fig. 2a is a schematic diagram of a basic gesture of ternary (gesture 0).

Fig. 2b is a schematic diagram of a basic gesture of a ternary system (gesture 1).

Fig. 2c is a schematic diagram of a basic gesture of ternary system (gesture 2).

Fig. 3a is a schematic diagram of combined coding of a specific gesture in the ternary equal-length coding mode.

Fig. 3b is a schematic diagram of an error detection coding of a specific gesture in a ternary equal-length coding mode.

Fig. 3c is a schematic diagram of error correction coding of a specific gesture in the ternary equal-length coding mode.

Fig. 4a is a coding table of ternary unequal-length codes for recognizing nine kinds of gesture commands.

Fig. 4b is a schematic diagram of error detection and decoding of ternary non-equal-length codes for recognizing nine gesture commands.

Fig. 5a is a coding table of ternary non-equal-length codes for recognizing nine gesture commands.

Fig. 5b is a schematic diagram of error correction decoding of ternary non-equal-length codes for recognizing nine gesture commands.

detailed description

As shown in Figure 1, the dynamic wireless gesture recognition method based on gesture coding and decoding of the present invention comprises the following steps:

10) Constructing a basic gesture set: constructing a basic gesture set according to multiple basic gestures;

The plurality of basic gestures satisfy: each basic gesture is divided into multiple motion time periods of equal number, within the time period of the same serial number, change the direction of motion, range of motion or movement between each basic gesture relative to the receiver range, to maximize the minimum norm distance between the feature vectors of each basic gesture or to make the feature vectors of each basic gesture orthogonal (even if the angle between the vectors is 90 degrees).

Among them, the formula for calculating the angle between two vectors a=[a ₁ ,a ₂ ,…a _N ]﹑b=[b ₁ ,b ₂ ,…b _N ]:

The formula for calculating the norm distance between two vectors a﹑b:

One way to maximize the norm distance of each basic gesture feature vector: determine the number N of basic gestures, design a set of M (M>N) candidate basic gestures, measure each candidate gesture feature vector once, and select one N order subset, so that the distance between the two basic gestures with the smallest feature vector norm distance in this subset is the largest.

One way to make the feature vectors of each basic gesture orthogonal: if there are three basic gestures, and the basic gestures are divided into 12 segments, then construct three basic gestures so that their feature vectors a ₁ , a ₂ , and a ₃ are symbolized respectively has the following form:

a ₁ =[1 1 1 1 1 1 -1 -1 -1 -1 -1 -1]

a ₂ =[1 1 1 1 1 1 1 1 1 1 1 1]

a ₃ =[-1 -1 -1 1 1 1 1 1 1 -1 -1 -1]

It can be verified that the above three vectors are orthogonal to each other. Other forms of orthogonal vectors include the following forms and their variants:

a ₁ =[1 0 0 0 0 0 0 0 0 0 0 0]

a ₂ =[0 0 0 0 1 0 0 0 0 0 0 0]

a ₃ =[0 0 0 0 0 0 0 0 1 0 0 0]

and

a ₁ =[1 0 0 0 0 1 0 0 0 0 1 0]

a ₂ =[0 0 0 1 1 0 0 0 0 1 0 0]

a ₃ =[0 0 1 0 0 0 0 1 1 0 0 0]

The present invention is no longer enumerated one by one.

In the embodiment, a ternary basic gesture diagram as shown in FIG. 2 is obtained by using the method of maximizing the distance of the eigenvector norm.

In the embodiment, three basic gestures are selected, and the basic gestures are mapped to "0", "1", and "2", and ternary code is used. The Doppler frequency shift value caused by the gesture motion is used as the feature of the gesture. The Doppler shift is related to the relative velocity and direction of motion. Among them, the gesture "0": make a circular motion, first move towards the receiver and then move away from the receiver, and return to the starting point. Gesture "1": Make a circular motion, first move away from the receiver and then move towards the receiver, returning to the starting point. Gesture "2": Make a linear movement in the direction of the man-machine connection, first move away from the receiver and then move towards the receiver, return to the starting point, then move towards the receiver and then move away from the receiver, and return to the starting point.

20) Offline training of basic gestures: collect each basic gesture for basic gestures, detect the signal emitted by the transmitter reflected by it multiple times, obtain the characteristics of each basic gesture, store all the characteristics of basic gestures, and establish basic gesture characteristics Library;

Described basic gesture off-line training (20) step comprises:

21) Segmentation of basic gestures: for each basic gesture, divide its complete process into multiple segments of equal duration;

In the step of segmenting the basic gestures (21), the number of segments of the basic gestures is 5-30.

22) Feature scalar acquisition: discretize the signal features corresponding to each gesture to obtain the signal feature scalar of each gesture;

The characteristic scalar acquisition (22) step is specifically:

Take the average value of the multiple measurement values of each segment of the signal or use the median of the multiple measurement values as the characteristic scalar of each segment of the signal.

In the feature scalar acquisition (22) step, the signal features include signal amplitude, phase, signal direction of arrival or variation of Doppler frequency shift with gesture movement.

23) Feature vector acquisition: combining the signal feature scalars of the entire basic gesture in turn to form the signal feature vector of the basic gesture;

24) Establishment of the basic gesture feature library: the signal feature vectors of each basic gesture constitute the basic gesture feature library.

In the embodiment, the basic gesture is divided into 12 segments, and the Doppler frequency shift value of each segment is measured as the feature of the gesture, and the Doppler frequency shift of each segment is obtained after taking the average of multiple measurement values. Since different users complete the same gesture, the characteristic There may be differences in the value vectors. In order to eliminate the differences, the symbols of the Doppler frequency shift are taken and combined to form the Doppler frequency shift vectors of each gesture, which are recorded as G ₀ , G ₁ , and G ₂ . According to the Doppler frequency shift characteristic, a form of the Doppler frequency shift vector of each basic gesture measured in the training phase is as follows: G ₀ =[1 1 1 1 1 1 -1 -1 -1 -1 -1 -1], G ₁ = [-1 -1 -1-1 -1 -1 1 1 1 1 1], G ₂ =[1 1 1 -1 -1 -1 -1 -1 -1 -1 1 1]. The above three vectors constitute the basic gesture feature library.

30) Gesture coding: coding basic gestures into command gestures corresponding to gesture control commands, constructing the command gestures includes basic gesture combination coding, error detection coding and error correction coding;

The gesture coding (30) step includes:

31) Coding mode judgment: judge the gesture coding mode, if it is non-equal-length coding, then go to the fault-tolerant coding mode judgment (33) step;

32) Combined gesture generation: use the same number of basic gesture combinations to form a combined gesture, and realize the mapping from gesture commands to combined gestures.

33) Error-tolerant encoding mode judgment: judge the error-tolerant encoding mode, if it is an error-correcting encoding mode, then go to the error-correcting encoding generation (35) step;

34) Generation of error detection codes: perform error detection codes on gesture commands or combined gestures. If equal-length codes are used, at least two code elements in any two code groups in the code table of combined gesture error detection codes are different. Equal-length codes, any two code groups in the code table of gesture command error detection codes either have different lengths, or have at least two different code units;

35) Generation of error correction codes: perform error correction codes on gesture commands or combined gestures. If equal-length codes are used, any two code groups in the code table of combined gesture error correction codes have at least three different code elements. length coding, then any two code groups in the coding table of the gesture command error correction coding are either different in length, or have at least three different code elements.

As shown in FIG. 3 , it is a schematic diagram of combined coding, error detection coding and error correction coding of a specific gesture in the ternary equal-length coding mode.

Equal-length combination coding scheme: In the embodiment of the ternary equal-length coding mode, the three basic gestures shown in Figure 2 can be combined into nine gestures, which can be used to control nine commands, and one possible encoding of the nine gestures for

Equal-length error detection coding scheme: further perform combined gesture error detection coding on the above combined codes (when the probability of misoperation is not high, the operation mode is set to error detection coding, gestures are simple and misoperations are avoided at the same time). The coding length of the error detection coding is 3 bits, that is, an error detection bit is added on the basis of the combined coding. An error detection coding method of the above combined coding is as follows:

(1) Divide the above nine combined gestures into three groups vertically, the first group being 00 01 02. The third digit of the first group of error detection codes is the non-repetitive permutation and combination of 0, 1, and 2, a total of 6 types: 000 011 022, 000 012 021, 001 010 022, 001012 020, 002 010 021, 002 011 020;

(2) The second and third groups of third digits are the cyclic shifts of the previous group of third digits respectively. The cyclic shifts include left shift and right shift, and the shift directions of the second and third groups are the same. There are two types in total. Realized, a total of 12 encoding methods.

(3) The minimum distance of the above encoding method is 2: it is guaranteed that any two code groups have at least two code elements that are different, and when a code element is misjudged, this error can be found. For example, under the first encoding method, when "000" is misjudged as "001", this decoding error can be found.

According to the method described, all 12 error detection codes are:

On the basis of the above-mentioned gesture error detection code, continue to add a code to realize the gesture error correction code, that is, the length of the error correction code is 4, and a group of error correction codes removes the fourth digit and the remaining three bits constitute a group of error detection codes. In the embodiment of the present invention, the first type of error correction code is taken as an example to illustrate the construction process of the error correction code. The other 11 methods of constructing error-correcting codes using the above-mentioned error-detecting codes are exactly the same.

Equal-length error-correcting coding scheme: a ternary (4,2) code is used, that is, the coding length is 4, and 2 information bits are included. Each error detection code is divided into three groups vertically, the fourth digits of the first group are 0, 1, 2 non-repeating combinations, a total of 6 types; the fourth digits of the second and third groups are respectively the fourth digits of the previous group The cyclic shift (that is, the fourth digit of the second group is the cyclic shift of the fourth digit of the first group, the fourth digit of the third group is the cyclic shift of the fourth digit of the second group), the fourth digit of the second and third groups The shift direction of the bits is the same , and the cyclic shift method is just opposite to that of the third bit. Each set of error detection codes finally derives 6 sets of error correction codes, with a total of 72 error correction codes. The minimum distance between error correction codes is 3: any two codes in a set of codes have at least 3 symbols different.

When the error detection code adopts the first group of 12 groups:

00 0 01 1 02 2

10 1 11 2 12 0

20 2 21 0 22 1

The corresponding 6 sets of error correction codes are:

In the embodiment of the present invention, in order to facilitate memory, the second error correction code is adopted

0000 0112 0221

1011 1120 1202

2022 2101 2210

The characteristics of this group of error correction codes: the 1st bit+2nd bit=3rd bit of each code, the 2nd bit+3rd bit=4th bit (here the addition is modulo 3 addition). Therefore, simply memorize the command number corresponding to the gesture control command, and its encoding method can be calculated. For example, in robot control, "forward" corresponds to command No. 2, then the combination code is 01, the error detection code is 011, and the error correction code is 0112.

Non-equal-length error detection coding scheme: the present invention also provides a ternary non-equal-length gesture command error detection coding. Still taking nine gesture commands as an example, the form of the error detection coding is as follows:

The present invention also provides a binary non-equal-length error detection code, and the basic gesture is mapped to two coding symbols "0" and "1". The corresponding gestures of "0" and "1" are the same as above. Still taking the nine gesture commands as an example, the form after error detection coding is as follows:

Non-equal length error correction coding scheme: the present invention also provides a ternary non-equal length gesture command error correction coding, still taking nine gesture commands as an example, the form of the error correction coding is as follows:

The present invention also provides a binary unequal-length error- correcting code, and the basic gesture is mapped to two coding symbols "0" and "1". The corresponding gestures of "0" and "1" are the same as above. Still taking the nine gesture commands as an example, the form after combined coding and error correction coding is as follows:

In the above binary and ternary error detection codes, except for the unique unary code "0", any pair of codes whose code length is not less than 2 or code lengths are different, or at least 2 code elements are different, so it has the ability to detect wrong ability.

Likewise, the binary and ternary error correction codes described above each differ from the other in code length or code symbols. For example, in ternary non-equal-length error- correcting codes, the unique code length of the unary code "0" and the unique binary code "11", the other codes with the same length and code length not less than 3 have at least three symbols different, such as "000" and "111", so the error correction capability of the above codes comes from two aspects: (1) codes with a length of not less than 3 rely on error correction codes to correct gesture misjudgments; (2) codes with a length of less than 3 rely on gestures and A unique mapping between encoding lengths to avoid gesture misjudgments. At the same time, shorter codes are assigned to more frequently used control commands.

According to the above ideas, other forms of binary and ternary codes can also be constructed with slight changes, which are not listed one by one in the present invention.

40) Acquisition of the command gesture signal to be checked: the transmitter transmits the signal, and the receiver receives the signal reflected by the command gesture to obtain the command gesture signal to be checked;

In the step of obtaining (40) the gesture signal to be detected, the transmitted signal includes: a single-tone signal, a double-tone signal, an OFDM signal, and a frequency-modulated continuous wave signal.

50) command gesture signal recognition: the command gesture signal to be checked is divided into multiple sub-gestures, each sub-gesture corresponds to an unknown basic gesture, and the command gesture signal recognition is realized through the recognition of each sub-gesture, and each sub-gesture recognition is to detect each sub-gesture The signal features of each sub-gesture are compared with the basic gesture features in turn to find out the basic gestures that match each sub-gesture, and these basic gestures are combined in turn to form the judgment result of the command gesture signal to be checked;

In the gesture signal recognition (50) step, the sub-gesture recognition includes:

51) Sub-gesture signal segmentation: for each sub-gesture signal, its complete process is divided into multiple segments of equal duration, the number of segments is the same as the number of segments in the step of the basic gesture segment (21), and each segment is calculated The signal feature scalars are combined into sub-gesture signal feature vectors in sequence, and according to the preset mode, the absolute value or only the sign of each scalar in the sub-gesture signal feature vector is reserved;

52) Judgment of signal feature type: judge the signal feature type of the sub gesture signal, if it is the symbol of the measured value, then go to the norm calculation (54) step;

53) Angle calculation: calculate the angle between the sub gesture signal feature vector and all feature vectors in the basic gesture feature library, and go to the gesture matching (55) step;

54) Norm calculation: calculate the norm of the difference between the sub gesture signal feature vector and all feature vectors in the basic gesture feature library;

55) Sub-gesture matching: Find the basic gesture with the minimum norm of the angle between the basic gesture feature library and the sub-gesture signal feature vector or the difference with the sub-gesture signal feature vector, and use it as the basic gesture that matches the sub-gesture signal .

The division of sub-gestures in the measurement process needs to distinguish between the beginning and the end of gestures. One existing method pauses between different gestures for a short time to identify the beginning of a new gesture. See "Q. Pu, S. Gupta, S. Gollakota, and S. .Patel, Whole-home gesture recognition using wireless signals, ACM International Conference on Mobile Computing and Networking (MOBICOM), 2013". In the sub-gesture recognition embodiment, for each sub-gesture, the sub-gesture signal is divided into 12 segments, the Doppler frequency shift value of each segment signal is measured, and the sign of the measured value is taken to form a measured feature vector. Calculate the distance between the measured feature vector and the Doppler frequency shift vectors G ₀ , G ₁ , and G ₂ of each basic gesture obtained in the training stage, and the basic gesture with the smallest distance between the Doppler frequency shift vector and the measured feature vector is the decision gesture. Here, the distance between two vectors is defined as the norm of the difference between the two vectors.

For example, when the basic gesture is the gesture "0", the measurement error causes the measured Doppler frequency shift vector to be G _m =[-1 11 1 1 1 -1 1 -1 -1 -1 -1]. Then the distances d _0,m , d _1,m , d _2,m between G _m and G ₀ , G ₁ , G ₂ are respectively Obviously, the gesture measured by the decision is the basic gesture "0".

60) Gesture signal decoding: Perform error detection decoding and error correction decoding on the judgment result of the gesture signal to be detected to obtain a gesture control command corresponding to the gesture signal to be detected.

The gesture signal decoding (60) step includes:

61) Error-tolerant encoding mode judgment: judge the gesture error-tolerant encoding mode, if it is an error-correcting mode, then go to the gesture error-correcting decoding (63) step;

62) Gesture error detection and decoding: determine whether there is an error in gesture recognition, if an error is found, stop decoding, gesture control becomes an invalid operation, otherwise decode and go to the gesture control command output (64) step;

63) Gesture error correction decoding: use the optimal maximum likelihood decoding or minimum distance decoding to correct errors in gesture recognition and obtain decoding results;

64) Gesture control command output: output the gesture control command corresponding to the decoding result.

After the basic gesture judgment is completed, a set of codes, that is, the measured values of the original gesture codes, is obtained, which is called "measurement codes" in the present invention. If the measurement error is so large that occasional misjudgments occur in the above basic gesture judgments, the "measurement code" will be different from the original code. It can be further processed by gesture error-tolerant decoding: the embodiment of the present invention provides a ternary equal-length error detection decoding and error correction decoding of a gesture recognition system that recognizes nine gesture commands, and a ternary non-equal Decoding methods for long error detection decoding and error correction decoding.

Ternary equal-length error detection decoding: According to the encoding characteristics of the ternary equal-length error detection coding table, as shown in the following table, judge whether each digit of the "measurement code" satisfies the following relationship (wherein addition is modulo 3 addition): 1st place + 2nd place = 3rd place? If they are equal, strip the third bit to get the decoding result. If they are not equal, an error message will be given, and the current gesture operation is invalid.

00 0 01 1 02 2

10 1 11 2 12 0

20 2 21 0 22 1

Trinary equal-length error-correcting decoding: Calculate the distance between the "measurement code" and all codes in the following code table in turn. The distance is defined as: the number of different code elements corresponding to two codes. If the actual gesture code is "0000", the "measurement code" judged by the measurement error is "0002", the "measurement code" and the stopwatch

0000 0112 0221

1011 1120 1202

2022 2101 2210

Each coding distance in is:

1 2 3

3 4 2

2 3 4

Then the distance between "0000" and "0002" is the smallest, and after error correction decoding obtains "0000", the third and fourth bits are stripped to obtain the decoding result.

Ternary non-equal-length error detection decoding: judge the length of the "measurement code", when the code length is 1, directly output the decoding result "0", when the code length is 2, calculate the "measurement code" and code table For the distance of each binary code in , refer to the code table below, here is [00 11 22]. If the minimum distance is zero, the "measurement code" is directly output as a decoding; if the minimum distance is greater than zero, an error prompt is given, and the judgment of the basic gesture is wrong. When the code length is 3, calculate the distance between the "measurement code" and each ternary code [000 011 022 101 202] in the code table. If the minimum distance is zero, the "measurement code" is directly output as a decoding; if the minimum distance is greater than zero, an error prompt is given, and the judgment of the basic gesture is wrong.

0 00 11

22 000 011

022 101 202

Ternary non-equal length error correction decoding: judge the length of the "measurement code", when the code length is 1, directly output the decoding result "0". When the encoding length is 2, the decoding result "11" is directly output. When the code length is 3, calculate the distance between the "measurement code" and each ternary code in the code table, refer to the code table below, here is [000 111 222], the ternary code with the smallest distance from the "measurement code" is used as the translation code output. When the code length is 4, the processing is the same as when the code length is 3.

0 11 000

111 222 0000

0111 0222 1012

According to the decoding result, the gesture control command can be obtained, and the corresponding command can be executed.

In the method of the present invention, only a small number of basic gestures are used, and other gestures can be constructed by making full use of the coding theory in communication, which avoids the cumbersome gesture training process when the amount of gestures is large; when the number of gestures in the basic gesture set is small, it is easy Construct basic gestures so that the minimum distance (that is, the distance between the two gestures with the smallest distance) is larger; and the greater the minimum distance between gestures, the smaller the probability of misjudgment;

In the method of the present invention, gesture combination coding and error detection and error correction coding are independent of specific systems (such as ultrasound, radio, Doppler or signal amplitude), and have universal applicability;

After the gesture is digitized, the error correction coding technology in digital communication can be used. When the wireless channel fading, the influence of interference or human misoperation cause occasional misjudgment in the basic gesture judgment, the misjudgment can be corrected to ensure correct gesture recognition.

Therefore, by adopting the dynamic wireless gesture recognition method based on gesture encoding and decoding of the present invention, the wireless gesture recognition probability is high, the reliability is good, and it can be used in various wireless gesture recognition systems.

Claims (10)

1. A dynamic wireless gesture recognition method based on gesture encoding and decoding, characterized in that, comprising the steps: 10) Constructing a basic gesture set: constructing a basic gesture set according to multiple basic gestures; 20) Offline training of basic gestures: collect each basic gesture for basic gestures, detect the signal emitted by the transmitter reflected by it multiple times, obtain the characteristics of each basic gesture, store all the characteristics of basic gestures, and establish basic gesture characteristics library; 30) Gesture coding: coding basic gestures into command gestures corresponding to gesture control commands, constructing the command gestures includes basic gesture combination coding, error detection coding and error correction coding; 40) Acquisition of the command gesture signal to be checked: the transmitter transmits the signal, and the receiver receives the signal reflected by the command gesture to obtain the command gesture signal to be checked; 50) command gesture signal recognition: divide the command gesture signal to be checked into multiple sub-gestures, each sub-gesture corresponds to an unknown basic gesture, and realize command gesture signal recognition through sub-gesture recognition, and each sub-gesture recognition is to detect each sub-gesture The signal features of each sub-gesture are compared with the basic gesture features in turn to find out the basic gestures that match each sub-gesture, and these basic gestures are combined in turn to form the judgment result of the command gesture signal to be checked; 60) Gesture signal decoding: Perform error detection decoding and error correction decoding on the judgment result of the gesture signal to be detected to obtain a gesture control command corresponding to the gesture signal to be detected. 2. The gesture recognition method according to claim 1, characterized in that, in the step of constructing a basic gesture set (10), the plurality of basic gestures satisfy: each basic gesture is divided into a plurality of motion time periods of equal quantity , within the time period of the same serial number, change the motion direction, motion range or motion range of each basic gesture relative to the receiver, so that the minimum norm distance between the feature vectors of each basic gesture is the largest or the characteristics of each basic gesture Vector Orthogonal. 3. gesture recognition method according to claim 1, is characterized in that: described basic gesture off-line training (20) step comprises: 21) Segmentation of basic gestures: for each basic gesture, divide its complete process into multiple segments of equal duration; 22) Feature scalar acquisition: discretize the signal features corresponding to each gesture to obtain the signal feature scalar of each gesture; 23) Feature vector acquisition: combining the signal feature scalars of the entire basic gesture in turn to form the signal feature vector of the basic gesture; 24) Establishment of the basic gesture feature library: the signal feature vectors of each basic gesture constitute the basic gesture feature library. 4. The gesture recognition method according to claim 3, characterized in that: in the step of segmenting the basic gestures (21), the number of segments of the basic gestures is 5-30. 5. gesture recognition method according to claim 3, is characterized in that, described feature scalar acquisition (22) step is specifically: The average value of the multiple measurement values of each section of signal is taken as the characteristic scalar of each section of signal, and the characteristic scalar of each section of signal adopts the absolute value of the measured value or only takes its sign. 6. The gesture recognition method according to claim 3, characterized in that, in the feature scalar acquisition (22) step, the signal features include signal amplitude, phase, signal direction of arrival or Doppler frequency shift with the gesture The amount of change in motion. 7. The gesture recognition method according to claim 1, characterized in that: the gesture coding (30) step comprises: 31) Coding mode judgment: judge the gesture coding mode, if it is non-equal-length coding, then go to the fault-tolerant coding mode judgment (33) step; 32) Combined gesture generation: use the same number of basic gesture combinations to form a combined gesture, and realize the mapping from gesture commands to combined gestures; 33) Error-tolerant encoding mode judgment: judge the error-tolerant encoding mode, if it is an error-correcting encoding mode, then go to the error-correcting encoding generation (35) step; 34) Generation of error detection codes: perform error detection codes on gesture commands or combined gestures. If equal-length codes are used, at least two code elements in any two code groups in the code table of combined gesture error detection codes are different. Equal-length codes, any two code groups in the code table of the gesture command error detection code are either different in length, or have at least two different code elements; 35) Generation of error correction codes: perform error correction codes on gesture commands or combined gestures. If equal-length codes are used, any two code groups in the code table of combined gesture error correction codes have at least three different code elements. length coding, then any two code groups in the coding table of the gesture command error correction coding are either different in length, or have at least three different code elements. 8. The gesture recognition method according to claim 1, characterized in that: in the step of obtaining (40) the gesture signal to be checked, the transmitting signal comprises: a single-tone signal, a dual-tone signal, an OFDM signal, Frequency modulated continuous wave signal. 9. The gesture recognition method according to claim 1, characterized in that: in the gesture signal recognition (50) step, the sub-gesture recognition comprises: 51) Sub-gesture signal segmentation: for each sub-gesture signal, its complete process is divided into multiple segments of equal duration, the number of segments is the same as the number of segments in the step of the basic gesture segment (21), and each segment is calculated The signal feature scalars are combined into sub-gesture signal feature vectors in sequence, and according to the preset mode, the absolute value or only the sign of each scalar in the sub-gesture signal feature vector is reserved; 52) Judgment of signal feature type: judge the signal feature type of the sub gesture signal, if it is the symbol of the measured value, then go to the norm calculation (54) step; 53) Angle calculation: calculate the angle between the sub gesture signal feature vector and all feature vectors in the basic gesture feature library, and go to the gesture matching (55) step; 54) Norm calculation: calculate the norm of the difference between the sub gesture signal feature vector and all feature vectors in the basic gesture feature library; 55) Sub-gesture matching: Find the basic gesture with the minimum norm of the angle between the basic gesture feature library and the sub-gesture signal feature vector or the difference with the sub-gesture signal feature vector, and use it as the basic gesture that matches the sub-gesture signal . 10. The gesture recognition method according to claim 1, characterized in that: the gesture signal decoding (60) step comprises: 61) Error-tolerant encoding mode judgment: judge the gesture error-tolerant encoding mode, if it is an error-correcting mode, then go to the gesture error-correcting decoding (63) step; 62) Gesture error detection and decoding: determine whether there is an error in gesture recognition, if an error is found, stop decoding, gesture control becomes an invalid operation, otherwise decode and go to the gesture control command output (64) step; 63) Gesture error correction decoding: use the optimal maximum likelihood decoding or minimum distance decoding to correct errors in gesture recognition and obtain decoding results; 64) Gesture control command output: output the gesture control command corresponding to the decoding result.