CN111176441A - Surface myoelectricity-based man-machine interaction training method and device and storage medium - Google Patents
Surface myoelectricity-based man-machine interaction training method and device and storage medium Download PDFInfo
- Publication number
- CN111176441A CN111176441A CN201911182105.2A CN201911182105A CN111176441A CN 111176441 A CN111176441 A CN 111176441A CN 201911182105 A CN201911182105 A CN 201911182105A CN 111176441 A CN111176441 A CN 111176441A
- Authority
- CN
- China
- Prior art keywords
- action
- effective signal
- surface electromyographic
- myoelectricity
- signal segment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 51
- 230000003993 interaction Effects 0.000 title claims abstract description 45
- 230000009471 action Effects 0.000 claims abstract description 119
- 210000003205 muscle Anatomy 0.000 claims abstract description 59
- 238000012545 processing Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 230000003068 static effect Effects 0.000 claims abstract description 7
- 238000013528 artificial neural network Methods 0.000 claims description 25
- 230000033001 locomotion Effects 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000003909 pattern recognition Methods 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 6
- 230000003183 myoelectrical effect Effects 0.000 abstract description 22
- 230000005236 sound signal Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 15
- 230000008569 process Effects 0.000 description 13
- 238000002567 electromyography Methods 0.000 description 11
- 210000002569 neuron Anatomy 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 210000000245 forearm Anatomy 0.000 description 7
- 230000006872 improvement Effects 0.000 description 7
- 238000000718 qrs complex Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 230000002232 neuromuscular Effects 0.000 description 5
- 230000002829 reductive effect Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000000747 cardiac effect Effects 0.000 description 4
- 238000012546 transfer Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 230000037433 frameshift Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 210000000707 wrist Anatomy 0.000 description 3
- 210000003984 auditory pathway Anatomy 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 210000000239 visual pathway Anatomy 0.000 description 2
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 1
- 206010049816 Muscle tightness Diseases 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 210000001015 abdomen Anatomy 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000005183 dynamical system Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000001698 laser desorption ionisation Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004118 muscle contraction Effects 0.000 description 1
- 210000001087 myotubule Anatomy 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000007830 nerve conduction Effects 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 230000035790 physiological processes and functions Effects 0.000 description 1
- 230000002360 prefrontal effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 210000003019 respiratory muscle Anatomy 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/015—Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The invention discloses a human-computer interaction training method, a device and a storage medium based on surface myoelectricity, wherein the human-computer interaction training method based on the surface myoelectricity comprises the following steps: collecting a surface electromyographic signal generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals of the target object in a static state and an action state; preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals; obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold; and extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment. According to the technical scheme provided by the invention, the myoelectric signals of the human body are collected and subjected to signal processing, and are converted into game instructions, and the game instructions are fed back to the human body through the game pictures or sound signals, so that the diversity of the forms of the myoelectric biological rehabilitation training method is improved.
Description
Technical Field
The invention relates to the technical field of human-computer interaction training methods, in particular to a human-computer interaction training method and device based on surface myoelectricity and a storage medium.
Background
The myoelectric biological rehabilitation training is to record weak electric signals when muscles contract autonomously by means of a surface myoelectric receiving device, provide feedback signals through a visual or auditory pathway by taking the weak electric signals as a source, change internal functions which cannot be realized by people, and convert the internal functions into recognizable audio-visual signals, so that a patient can learn to control the patient to play a training role by guiding and self-training of a doctor.
The existing myoelectric biological rehabilitation training method is single in form.
Disclosure of Invention
The embodiment of the invention provides a human-computer interaction training method, a human-computer interaction training device and a storage medium based on surface electromyography.
The embodiment of the invention provides a human-computer interaction training method based on surface myoelectricity, which is applied to a human-computer interaction training device and comprises the following steps:
collecting a surface electromyographic signal generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals of the target object in a static state and an action state;
preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals;
obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold;
extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
generating a corresponding control instruction according to the action characteristics corresponding to the muscle type;
and making a corresponding response action according to the control instruction.
As an improvement, the step of obtaining a valid signal segment according to the processed surface electromyogram signal and a preset threshold specifically includes:
sequentially judging whether the processed surface electromyographic signals are larger than a preset threshold value or not;
when the processed surface electromyographic signal is larger than a preset threshold value, recording a time point as an action starting point;
after the action starting point, when the processed surface electromyogram signal is not greater than a preset threshold value, recording a time point as an action ending point, and repeatedly and sequentially judging whether the processed surface electromyogram signal is greater than the preset threshold value to act; and the time period between the action starting point and the action ending point is an effective signal segment.
As an improvement, the step of sequentially determining whether the processed surface electromyogram signal is greater than a preset threshold specifically includes:
acquiring a noise threshold and an electrocardio interference threshold; wherein the noise threshold is an average of absolute values of background noise band levels; the electrocardio-interference threshold is a Short-time Zero-crossing Ratio (ZCR) threshold;
and sequentially judging whether the processed surface electromyographic signals are larger than a noise threshold and an electrocardio-interference threshold.
As an improvement, the step of extracting the motion feature of the effective signal segment and identifying the muscle type emitting the motion feature according to the motion feature of the effective signal segment specifically includes:
calculating a Root Mean Square (RMS) value by a first formula, wherein the first formula is:
n is the Data total point number of the surface electromyographic signals of the effective signal segment, and Data (i) represents the ith surface electromyographic signal Data point of the surface electromyographic signals of the effective signal segment;
and extracting the action characteristics of the effective signal section through the electromyographic root mean square value.
As an improvement, the step of extracting the motion feature of the effective signal segment and identifying the muscle type emitting the motion feature according to the motion feature of the effective signal segment specifically includes:
carrying out binarization processing on the action characteristic sequence of the effective signal segment by regions to obtain a character string SX；
Calculating a string SXThe number of new modes C (n);
normalizing the number C (n) of the new modes to obtain the LZC complexity; wherein, the LZC complexity is expressed by a formula two, and the formula two is as follows:
n is the total number of data points of the surface electromyographic signals of the effective signal segments;
and extracting the action characteristics of the effective signal segment through the LZC complexity.
As an improvement, the calculation of the string SXThe step of counting the number C (n) of new modes specifically includes:
setting a character string S (S) to be solved1,s2,...,sn) And another character string Q (Q)1,q2,...,qn)；
Initializing S to a binarization sequence SXAnd Q is the binarization sequence SXThe second element of (1);
when Q is a sub-character of SQv, then concatenating the next character of the pending sequence to Q; where SQv is the string of SQ minus the last character, SQ denotes the concatenation of S and Q, SQ ═ S (S)1,s2,...,sn,q1,q2,...,qn)；
When Q is not a substring of SQv, it indicates that Q is a new pattern, then Q is concatenated to S, i.e. S equals SQ, and the number of new patterns is recorded with C (n);
reconstructing Q, and taking binary sequence S from QXRepeatedly judging whether Q is a sub-symbol of SQv and executing corresponding steps until Q is obtained to the binary sequence S to be solvedXThe last element of (2).
As an improvement, the step of extracting the motion feature of the effective signal segment and identifying the muscle type emitting the motion feature according to the motion feature of the effective signal segment specifically includes:
obtaining a known learning sample, and training an offline BP neural network according to the known learning sample to obtain a weight and a threshold required by pattern recognition; the known learning samples comprise a plurality of LZC complexities, muscle types corresponding to the LZC complexities one by one and emitted action characteristics;
and inputting the real-time LZC complexity into the trained offline BP neural network to obtain the action characteristics of the muscle type.
As an improvement, the step of generating a corresponding control instruction according to the action characteristic of the muscle type specifically includes:
acquiring action characteristics of muscle types; wherein the action characteristics of the muscle type comprise an electromyographic root mean square value or an LZC complexity;
and the game server generates a corresponding control instruction according to the action characteristics of the muscle type.
A second aspect of an embodiment of the present invention provides a surface myoelectric-based human-computer interaction training device, including:
the system comprises at least one electromyographic signal collector, a data processing unit and a data processing unit, wherein the at least one electromyographic signal collector is used for collecting surface electromyographic signals generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals in a static state and a motion state;
the band-pass digital filter is electrically connected with the at least one electromyographic signal collector respectively and is used for preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals;
the terminal is used for obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold value;
the terminal is further configured to: extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
generating a corresponding control instruction according to the action characteristics of the muscle type;
and making a corresponding response action according to the corresponding control instruction.
The first aspect of the embodiments of the present invention provides a storage medium, where a human-computer interaction training program based on surface myoelectricity is stored, and when the human-computer interaction training program based on surface myoelectricity is executed by at least one processor, the human-computer interaction training method based on surface myoelectricity is implemented.
According to the technical scheme provided by the invention, the myoelectric signals of the human body are collected and subjected to signal processing, and are converted into game instructions, and the game instructions are fed back to the human body through the game pictures or sound signals, so that the diversity of the forms of the myoelectric biological rehabilitation training method is improved.
Drawings
One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.
Fig. 1 is a schematic flow chart of a human-computer interaction training method based on surface electromyography according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a muscle type testing action in a human-computer interaction training method based on surface electromyography according to an embodiment of the present invention;
FIG. 3 is a four action LZC complexity profile of FIG. 2;
fig. 4 is a schematic diagram of training a BP neural network in a human-computer interaction training method based on surface myoelectricity according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a surface myoelectricity-based special game in a human-computer interaction training method according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a human-computer interaction training device based on surface myoelectricity according to an embodiment of the present invention.
Detailed Description
At present, the myoelectric biological rehabilitation training records weak electric signals when muscles contract autonomously by means of surface myoelectric receiving equipment, provides feedback signals through a visual or auditory pathway by taking the weak electric signals as a source, changes the internal functions which cannot be realized by people, converts the internal functions into visual and audio signals which can be realized, enables patients to learn to control the patients by doctors to guide and train the patients, and plays a training role. The existing myoelectric biological rehabilitation training method is single in form.
In order to solve the technical problems, the scheme provides a human-computer interaction training method, a device and a storage medium based on surface electromyography, wherein the form diversity of the electromyography biological rehabilitation training method is improved by acquiring the electromyography signals of a human body, processing the electromyography signals, converting the electromyography signals into game instructions and feeding back the game instructions to the human body through game pictures or sound signals.
In order to make the aforementioned objects, features and advantages of the present application more comprehensible, the present application is described in further detail with reference to the accompanying drawings and specific examples.
Referring to fig. 1, in order to achieve the above object, the present invention provides a human-computer interaction training method based on surface myoelectricity, which is implemented by a human-computer interaction training device, and specifically includes:
step S10: collecting a surface electromyographic signal generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals of the target object in a static state and an action state; in one embodiment, the target subject is a skin of a subject to be trained, wherein the skin includes, but is not limited to, prefrontal muscle, anterior wrist flexor of right forearm, and posterior extensor of right forearm; it should also be noted that surface myoelectricity generally implies information about the physiological state of the muscles and nervous system of the human body, reflecting to some extent the intention of neuromuscular activity.
Step S20: preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals; the preprocessing is mainly performed to filter noise and reduce interference.
Step S30: and obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold value.
It should be noted that: in the process of processing the surface electromyographic signals, effective electromyographic signals are often used as feedback signals, so the start and stop points of the electromyographic signals generated by human body actions in the collected signals need to be accurately known, and the electromyographic energy is calculated to judge whether the electromyographic energy reaches a trigger threshold value. In addition, when the detected part is close to the heart, for example, the surface electromyogram acquisition of a respiratory muscle group is easily influenced by the electrocardiographic QRS waveform, so that the electrocardiographic interference needs to be effectively reduced in the start-stop point detection algorithm, and the start-stop point needs to be accurately found.
Therefore, in the above step S30, in order to realize that a valid signal segment is obtained from the processed surface electromyogram signal and the preset threshold, the following steps S31 to S33 may be performed:
step S31: sequentially judging whether the processed surface electromyographic signals are larger than a preset threshold value or not;
in the above step S31, in order to sequentially determine whether the processed surface electromyogram signal is greater than the preset threshold, the following steps S311 to S312 may be performed:
step S311: acquiring a noise threshold and an electrocardio interference threshold; wherein the noise threshold is an average of absolute values of background noise band levels; the electrocardio-interference threshold is a Short-time Zero-crossing Ratio (ZCR) threshold;
specifically, to realize the acquisition noise threshold in step S311, the noise may be acquired through the following step S3111:
step S3111: when the muscle is in a relaxed state, calculating the mean value mu and the standard deviation sigma of the absolute values of the first 500 sampling points; it should be noted that, in the present embodiment, it is assumed that the background noise conforms to a gaussian distribution, and thus most of the noise amplitudes are within a range of plus or minus three times the standard deviation of the mean value, where the noise threshold is μ +3 σ in the present embodiment. Definition of LDIs the length of the signal X (N) to be processed and is divided into N frames with the length LFThe frame is shifted to M, and,the noise threshold criterion value of the ith frame is
j ∈ {1,2, 3., N }. In this embodiment, in order to improve the efficiency of the algorithm, the mean of the absolute values of each data frame is calculated to obtain the criterion value of the noise threshold, as shown in the following equation (1-1).
Generally, if the positioning accuracy of the start-stop point needs to be improved, a smaller frame length L needs to be selectedFAnd frame shift M, however, if there is transient spike noise in the electromyographic signal, a smaller frame length and frame shift may easily cause erroneous determination, so in order to improve the accuracy of start and stop point detection, it is necessary to perform filtering processing on each frame of data, that is, to remove high-frequency spike noise and to retain low-frequency signal.
In order to achieve the above-mentioned purpose of improving the accuracy of the start and stop point detection, the following steps S3112 to S3112 may be performed:
step S3112: noise threshold criterion for each frame
And performing m-order low-pass filtering processing to remove high-frequency spike noise.
The discrete transfer function of the m-order low-pass filtering is shown as formula (1-2), wherein
Z-transform representing a noise threshold criterion for each frame, Y (Z) representing a Z-transform of a filtered noise threshold criterion, a0～amAnd b0～bmIs a parameter of the low pass filter. The energy of the surface electromyogram signal is mainly concentrated in the range of 10-250 Hz, so that the cut-off frequency of the low-pass filter is 250Hz, high-frequency spike noise can be effectively removed, and more detailed information can be retained.
Step S3113: and performing moving average processing on the multi-frame filtered Y (j) to further eliminate peak interference.
It should be noted that the processing of the moving average is performed to further eliminate the factor of accidental variation, filter out spike interference, reduce the probability of misjudgment of the start and stop points of the surface electromyogram signal, and enhance the reliability of the start and stop point detection algorithm. The calculation formula is shown as formula (1-3), wherein LAvgThe number of frames for calculating the noise threshold criterion of the moving average should not be selected too large to reduce the accuracy of the detection of the start and stop points of the surface electromyographic signalAvgThe value is 64.
Specifically, in order to obtain the cardiac electric interference threshold in step S311, the following steps S3114 to S3115 may be performed:
the following idea of obtaining the electrocardio-interference threshold is introduced:
generally, the electrocardio waveform of a human is mainly composed of QRS complexes, the frequency is concentrated in a frequency range below 30Hz, the energy of the surface electromyogram signal is mainly concentrated in a frequency range of 10-250 Hz, the distribution is not as concentrated as the QRS complexes, if a frame length with enough length is selected for observation, the frequency of the surface electromyogram signal and the frequency of the QRS complexes passing through a zero line are obviously different, and the frequency of the surface electromyogram passing through the zero line is more than that of the QRS complexes. Therefore, a proper frame length is selected, the zero crossing rate of the frame is calculated, and after the zero crossing rate of the frame is compared with the electrocardio interference zero crossing rate threshold value, whether the frame belongs to the surface electromyogram signal or the electrocardio QRS complex can be judged.
Step S3114: calculating the appropriate frame length LFAnd a cardiac electrical interference zero crossing rate threshold. Wherein, define NAThe mean frequency of the main energy of the surface electromyography and the electrocardio is fEMGAnd fECG，fsFor the sampling frequency, the correlation is as follows (1-4):
it should be noted that the dominant frequency of the surface electromyographic signals may be chosen to be about 100Hz and the dominant frequency of the electrocardiosignals may be chosen to be about 30Hz, i.e. fECG≈30Hz，fEMG100 Hz. By analyzing L according to the formulas (1-4)F/fsTo estimate and calculate NADue to the threshold value NAMust be an integer, and the threshold value NAMust not differ too much, the sampling frequency f of the system of this embodiments2kHz, so the appropriate observation frame length is LF64, the electrocardio interference zero-crossing rate threshold is NA＝4。
Step S3115: the surface electromyogram signal and the electrocardio QRS complex are judged by calculating the zero crossing rate of each frame of data and comparing the zero crossing rate with the threshold value of the electrocardio interference zero crossing rate, and the calculation formula of the zero crossing rate of each frame of data is shown as the following formula (1-5).
In the formula (I), the compound is shown in the specification,
step S312: and sequentially judging whether the processed surface electromyographic signals are larger than a noise threshold and an electrocardio-interference threshold.
Step S32: when the processed surface electromyographic signal is larger than a preset threshold value, recording a time point as an action starting point; the preset threshold is a noise threshold and an ecg threshold, and this is also the case in step S33, which is not repeated herein.
Step S33: after the action starting point, when the processed surface electromyogram signal is not greater than a preset threshold value, recording a time point as an action ending point, and repeatedly and sequentially judging whether the processed surface electromyogram signal is greater than the preset threshold value to act; and the time period between the action starting point and the action ending point is an effective signal segment.
More specifically, the noise threshold and the cardiac electric interference threshold are respectively used for distinguishing the surface myoelectricity from the noise level and distinguishing the surface myoelectricity from the cardiac electric QRS waveform. The noise threshold is an average threshold of absolute values of background noise band levels of the surface electromyogram signals, the electrocardio interference threshold is a Short-time Zero-crossing rate threshold, and the Short-time Zero-crossing rate (ZCR) refers to the number of times that the surface electromyogram signals pass through a Zero line in a certain period of time. When the detected signal is higher than the noise threshold and the electrocardio-interference threshold, recording as a start point and a stop point; when the signal for detecting the starting point is lower than two threshold values, it is recorded as the end point. In the process, proper frame length (window length) and frame shift are required to be selected, the average value of the absolute values of the signals is calculated frame by frame, compared with the noise threshold and the electrocardio interference threshold, and the windows are shifted in sequence, so that the starting point and the end point of muscle contraction can be detected.
Step S40: extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
in one embodiment, in the step S40, in order to extract the motion feature of the valid signal segment and identify the muscle type emitting the motion feature according to the motion feature of the valid signal segment, the following steps S41-S42 may be performed:
step S41: calculating a Root Mean Square (RMS) value by a first formula, wherein the first formula is:
n is the total number of Data points of the surface electromyographic signals of the effective signal segment, and Data (i) represents the ith surface electromyographic signal Data point of the surface electromyographic signals of the effective signal segment.
Step S42: extracting the action characteristics of the effective signal section through the electromyographic root mean square value;
in another embodiment, in the above step S40, in order to extract the motion feature of the valid signal segment and identify the muscle type emitting the motion feature according to the motion feature of the valid signal segment, the following steps S43-S46 may be performed:
the following idea of extracting the feature value of the surface electromyogram signal by the nonlinear angle is introduced:
the motor neuromuscular system is essentially a highly nonlinear dynamical system, and in order to effectively and reliably acquire the characteristics of the surface electromyographic signals, nonlinear dynamical information of the surface electromyographic signals needs to be extracted. When the muscle is in contraction and relaxation or the muscle is in different fatigue degrees, the number of the muscle tissue movement units, the discharge frequency of each movement unit and the nerve conduction velocity are different, and from the kinetic point of view, the muscle tissue movement units are represented by different dynamic movement complexity degrees of the motor neuromuscular system. Therefore, in the embodiment, the complexity of the surface electromyogram signal is obtained through a correlation algorithm, and the characteristic value of the surface electromyogram signal is extracted from a nonlinear angle. In the embodiment, Lempel-Ziv complexity (which will be simply referred to as LZC complexity hereinafter) is employed to describe the complexity of the surface electromyogram signal. Complexity is a measure of the degree of randomness of a signal, and generally speaking, the larger this value, the higher the degree of randomness.
Step S43: carrying out binarization processing on the action characteristic sequence of the effective signal segment by regions to obtain a character string SX；
In the above step S43, to implement the binary processing for the action feature sequence of the effective signal segment with different partitions, a character string S is obtainedXThe following steps S431 to S433 may be performed:
step S431: firstly, solving the mean value of the sequence, obtaining two intervals through comparison of the mean values, and then carrying out similar interval division on each interval to finally obtain four intervals;
step S432: for a certain time signal value, if the certain time signal value is in the same interval with the last time value, the binary value of the certain time signal value is the same as the last time;
step S433: if the time value is smaller than the previous time value, the binary value is 0, otherwise, the binary value is 1.
Specifically, the collected surface electromyogram signal sequence in this embodiment is Data (N), the average value of the total sequence is T, the total number of sequence points is N, and the result after binarization is a character string SX＝(s1s2…sN) Then, the calculation expression of the inter-partition binarization method is shown as the following formula (1-6):
step S44: calculating a string SXThe number of new modes C (n);
in the above step S44, the character string S is calculatedXThe number of new modes C (n), the following steps S441 to S445 may be performed:
step S441: setting a character string S (S) to be solved1,s2,...,sn) And another character string Q (Q)1,q2,...,qn)；
Step S442: initializing S to a binarization sequence SXAnd Q is the binarization sequence SXThe second element of (1);
step S443: when Q is a sub-character of SQv, then concatenating the next character of the pending sequence to Q; where SQv is the string of SQ minus the last character, SQ denotes the concatenation of S and Q, SQ ═ S (S)1,s2,...,sn,q1,q2,...,qn)；
Step S444: when Q is not a substring of SQv, it indicates that Q is a new pattern, then Q is concatenated to S, i.e. S equals SQ, and the number of new patterns is recorded with C (n);
step S445: reconstructing Q, and taking binary sequence S from QXRepeatedly judging whether Q is a sub-symbol of SQv and executing corresponding steps until Q is obtained to the binary sequence S to be solvedXThe last element of (2).
Step S45: normalizing the number C (n) of the new modes to obtain the LZC complexity; wherein the LZC complexityExpressed by formula two, formula two is:
n is the total number of data points of the surface electromyographic signals of the effective signal segments;
step S46: and extracting the action characteristics of the effective signal segment through the LZC complexity.
In the above step S46, to extract the action feature of the effective signal segment by the LZC complexity, the following steps S461 to S462 may be performed:
the following idea of extracting the motion characteristics of the effective signal segment through LZC complexity is introduced:
the process of extracting the action characteristics of the effective signal section through the LZC complexity is also a machine learning process and can also be a training mode of a neural network; the action learning is divided into two conditions, the first condition is single-channel electromyogram value learning, and the root mean square value of a single-channel electromyogram signal is mainly collected to obtain a trigger threshold value; and in the second case, four gestures are recognized through myoelectric values of two channels, myoelectric complexity of the two channels is mainly collected, a weight and a threshold value required by pattern recognition are obtained through offline BP neural network training, and then the weight and the threshold value are written into a game for online recognition, wherein online learning is required for several times before online recognition and is used for correcting the weight and the threshold value. The BP neural network has strong nonlinear mapping capability and is widely applied to the fields of classification, fitting, compression and the like. Learning of the BP neural network is supervised learning, and therefore a known learning sample is required. According to functional requirements, 4-style mini games are developed for myoelectric biofeedback training in the embodiment, wherein the 4-style mini games comprise a dish game, a pixilated bird game, an angry bird game and an airplane playing game.
Several examples of single-channel electromyogram learning are described below:
examples of the seed-vegetable game: the myoelectric signal of the target neuromuscular activity is detected, a corresponding trigger threshold value is set according to the difficulty of game selection, a new dish grows on a game interface after the trigger threshold value is exceeded or is lower than the trigger threshold value, and the new dish is fed back to a patient, so that the muscle tension degree of the patient can be controlled.
Example of angry bird games: the myoelectric signals of the target neuromuscular activity are detected, a corresponding trigger threshold is set according to the difficulty of game selection, the bird is emitted out after the trigger threshold is exceeded or is lower than the trigger threshold, the emitting distance is determined by the myoelectric value, and the muscle strength of a patient can be adjusted.
Example of a pixel bird game: through detecting the myoelectric signal when the target nerve muscle is active, a corresponding trigger threshold value is set according to the difficulty of game selection, and the size of the myoelectric signal is required to be maintained all the time to control the bird in the game to fly over the obstacle, so that a patient can learn to adjust the muscle coordination.
An example of electromyogram value learning for two channels is described below:
example of a game of airplane play: the muscle actions of left palm rotation, right palm rotation, fist unfolding and fist grasping are identified by detecting myoelectric signals of forearm anterior wrist flexor and posterior extensor, and the pistol is controlled to move leftwards and rightwards, and shoot bullets singly and shoot continuously, so that the forearm muscle of a patient is effectively and pertinently trained; the specific experimental scheme is as follows: a healthy adult is selected as the subject, and the subject is 30 years old and has moderate weight. Electrodes are pasted on the positions of the front wrist flexor and the rear extensor of the right forearm of an experimental object, alcohol is needed to be used for cleaning the skin before the electrodes are pasted, electrode plates are needed to be pasted along the direction of muscle fibers, and the distance between the two electrodes is 2cm and the two electrodes are pasted on the belly of the muscle. Let the subject sit on the stool, right front arm is laid down on the desktop, accomplishes each 60 groups of four kinds of actions of exhibition fist, palm of clenching left side and palm of turning right side respectively to gather the flesh electricity signal of action process at every turn and regard as off-line action learning data, take a rest for 2 to 3 minutes between the action at every turn, in order to ensure that muscle is not in the fatigue state.
Generally speaking, each game carries out animation feedback by receiving myoelectricity acquisition signals, and the software part mainly comprises modules of surface myoelectricity signal real-time acquisition, preprocessing, action starting and stopping point judgment, surface myoelectricity signal characteristic extraction, online muscle activity identification, game animation feedback and the like. After the game is entered, the corresponding muscle activity size needs to be learned in advance as a threshold criterion, and four gesture recognition in the airplane game obtains a weight value and a threshold value required by pattern recognition through offline BP neural network training and then writes the weight value and the threshold value into the game for online recognition, wherein the weight value and the threshold value need to be learned online for several times before online recognition.
Step S461: obtaining a known learning sample, and training an offline BP neural network according to the known learning sample to obtain a weight and a threshold required by pattern recognition; the known learning samples comprise a plurality of LZC complexities, muscle types corresponding to the LZC complexities one by one and emitted action characteristics;
in step S461, in order to obtain a known learning sample, and train the offline BP neural network according to the known learning sample to obtain a weight and a threshold required for pattern recognition, the following steps S4611 to S4611 may be performed:
the following idea of training the offline BP neural network is introduced:
when training is started, initializing a weight value as a random value, and inputting a learning sample with known output to calculate the network output of the learning sample; then calculating the error between the output value and the target output value, and modifying the weight backwards layer by layer according to the error and a relevant criterion so that the error is gradually reduced; and circulating the steps until the error value is not reduced any more, and finishing the training of the network. The core of the BP network is to attribute the error between the network output and the target output to unreasonable network weights and thresholds, and then continuously modify the weights and thresholds through back propagation so that the error is not reduced any more until the target output is achieved. In this embodiment, the BP network structure is defined to have three layers, including M input neurons, I hidden layer neurons, and J output layer neurons. Wherein the mth neuron of the input layer is set to xmThe ith neuron of the hidden layer is set to kiThe j-th neuron of the output layer is set as yj. From xmTo kiThe connection weight is omegamiFrom kiTo yjThe connection weight is omegaij. The hidden layer transfer function is a Sigmoid function, and the output layer transfer function is a linear function.
Step S4611: transmitting an input vector with the length of M to the three-layer BP network to obtain an output vector with the length of J;
step S4612: the input and output of each layer are denoted u and v, respectively, and the actual output of the network is obtained as
Step S4613: obtaining a desired output of the network as d (n) — [ d1,d2,...dJ]The error signal of the nth iteration is defined as ej(n)＝dj(n)-Yj(n) defining the error energy as
The working process of training the offline BP neural network is described in more detail below:
step a: the forward propagation process of the working signal:
input signal of whole BP neural network
Input to the ith neuron of the hidden layer equals
Weighted sum of
Taking the transfer function f (-) of the hidden layer as a Sigmoid function, the output of the ith neuron of the hidden layer is obtained
Input to jth neuron of output layer is equal to
Weighted sum of
Output of jth neuron of output layer
Error of j-th neuron of output layer
Total error of network
Step b: error signal back propagation:
in the process of modifying the weight, adjustment needs to be performed in the opposite direction layer by layer along the network. So the weight ω between the hidden layer and the output layer is modified firstijCalculating error pair omega according to steepest descent methodijGradient of (2)
Then, the adjustment is performed along the direction reverse to the direction
ωij(n+1)＝Δωij(n)+ωij(n)；
Wherein the gradient can be derived by partial derivation
Therefore, the correction amount of the weight
Order to
Then
Error signal layer by layerPropagate forward and then modify the weights ω between the input and hidden layersmiSimilarly, there are
Wherein the derivation of partial derivatives is used to derive
And (4) carrying out weight adjustment for multiple times through the steps to gradually reduce the error until the error is not reduced any more, and then finishing the training of the BP neural network.
Referring to fig. 2-4, we can intuitively explain the training process through experiments; referring to fig. 2, fig. 2 is a schematic diagram illustrating a muscle type testing action of a target object in a human-computer interaction training method based on surface myoelectricity, wherein the muscle type testing actions from left to right are a fist unfolding, a fist making, a left palm turning and a right palm turning in sequence; FIG. 3 shows the results of experimental data acquisition and LZC complexity calculation to obtain four forearm movements at LZC1-LZC2Distribution over a plane; from FIG. 3, it can be clearly seen that the four actions are in LZC1-LZC2The plane is provided with a partitionable clustering area, and the separability is good. Therefore, the method can be intuitively seen that the LZC (Lempel-Ziv-Chic) of the signal complexity index can be calculated1And LZC2The signal characteristic of forearm action classification recognition is used for realizing the triggering of a rehabilitation game. Therefore, in an embodiment of the present disclosure, a new version of feed forward net of a neural network toolbox is used on MATLAB to create a BP network, and a training function train bfg corresponding to a quasi-newton method is used to perform network training, where a hidden layer of the BP network is 1 layer, the number of hidden nodes is 4, test data is divided into two parts, 40 groups are training data, and 20 groups are test data. The schematic diagram 4 of the BP neural network trained with the kit is shown, and the experimental results of the test are shown in table 1.
TABLE 1 recognition results after training of four actions
| Movement of | Exhibition fist | Fist making | Left hand rotation | Right hand rotation of |
| Exhibition fist | ||||
| 20 | 0 | 0 | 0 | |
| Fist making | 0 | 19 | 0 | 0 |
| |
0 | 0 | 20 | 0 |
| Right hand rotation of |
0 | 1 | 0 | 20 |
| Recognition rate | 100％ | 95％ | 100％ | 100％ |
As can be seen from table 1, the average recognition rate of the 4 actions is 98.75%, please refer to fig. 4, next, in an embodiment, the input layer weight net.iw {1,1} and the threshold net.b {1} of the trained BP Neural Network (Back prediction Neural Network), the hidden layer initial weight net.lw {2,1} and the threshold net.b {2} and the hidden layer function net.layers {1} and the output layer function net.layers {1} are obtained, and are respectively written into the game application code to perform online action recognition. The process of on-line identification is to manually calculate the output result of the new input signal according to the above-mentioned forward propagation process of the working signal. It should be noted that the feedwardnet function of the tool box automatically performs normalization and denormalization in MATLAB, so if the BP neural network trained by the feedwardnet function is manually calculated, manual normalization and denormalization are required. In addition, considering that the pre-trained neural network cannot meet any tested target object, when any target object is used for the first time, action information needs to be input in advance, and weight and threshold of the BP neural network are corrected online according to the error signal direction propagation process mentioned above.
Step S462: and inputting the real-time LZC complexity into the trained offline BP neural network to obtain the action characteristics of the muscle type.
Step S50: generating a corresponding control instruction according to the action characteristics corresponding to the muscle type;
in the above step S50, to generate a corresponding control instruction according to the action characteristic corresponding to the muscle type, the following steps S501 to S502 may be executed:
step S501: acquiring action characteristics of the muscle type, including myoelectricity root mean square value and LZC complexity;
step S502: the game server generates a corresponding control instruction according to the action characteristics of the muscle type; and generating a corresponding control instruction comprises generating a dish planting instruction, a moving instruction and an attack instruction.
Referring to fig. 5, fig. 5(a) is a schematic diagram illustrating a use status of a dedicated game based on surface myoelectricity in a human-computer interaction training method based on surface myoelectricity according to an embodiment of the present invention, in which, in an embodiment, a game server uses a standalone airplane shooting mini game, and fig. 5(b) shows an interface of the game; four trigger instructions are set in the game, the gun head is arranged leftwards, the gun head is arranged rightwards, the bullet is shot and the bullet is shot continuously, therefore, four actions are required to be recognized as the trigger instructions, and the four actions comprise fist unfolding, fist making, left palm rotating and right palm rotating. When the target object enters the game after first registration, the action of each instruction needs to be learned, and relevant action information is input and used for modifying the weight and the threshold of the BP neural network trained offline. For example, the screen prompts 'please input the action of the gun head to the left', the palm left-turning is assumed as the target action, the action signal is input through the myoelectricity acquisition line, the system requires continuous repeated action, and the repeated action frequency is determined according to the learning condition of the BP network and is not more than 10 times at most. By analogy, the myoelectric complexity of the other three actions needs to be learned. After the action learning is finished, the rehabilitation game training can be formally entered.
Step S60: and making a corresponding response action according to the control instruction.
Based on the same inventive concept, the embodiment of the present application further provides a human-computer interaction training method based on surface myoelectricity and a corresponding human-computer interaction training device based on surface myoelectricity, and as the principle of solving the problem of the device in the embodiment of the present application is similar to the human-computer interaction training method based on surface myoelectricity described in the embodiment of the present application, the implementation of the device may refer to the implementation of the aforementioned data processing method, and repeated parts are not described again.
Referring to fig. 6, in order to achieve the above object, the present invention further provides a surface myoelectric-based human-computer interaction training device, including:
the electromyographic signal collector 10 is used for collecting surface electromyographic signals generated by a target object, and the electromyographic signal collector 10 is used for collecting the surface electromyographic signals generated by the target object; wherein the surface electromyographic signals comprise surface electromyographic signals in a static state and a motion state;
the band-pass digital filter 20 is electrically connected with the at least one electromyographic signal collector 10, and the band-pass digital filter 20 is used for preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals;
the terminal 30 is used for obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold value;
the terminal 30 is further configured to: extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
generating a corresponding control instruction according to the action characteristics of the muscle type;
and making a corresponding response action according to the corresponding control instruction.
In order to achieve the above object, the present invention further provides a storage medium, where a human-computer interaction training program based on surface myoelectricity is stored, and when the human-computer interaction training program based on surface myoelectricity is executed by at least one processor, the human-computer interaction training method based on surface myoelectricity is implemented.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A man-machine interaction training method based on surface myoelectricity is applied to a man-machine interaction training device and is characterized in that: the man-machine interaction training method based on the surface myoelectricity comprises the following steps:
collecting a surface electromyographic signal generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals of the target object in a static state and an action state;
preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals;
obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold;
extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
generating a corresponding control instruction according to the action characteristics corresponding to the muscle type;
and making a corresponding response action according to the control instruction.
2. The surface myoelectricity-based human-computer interaction training method of claim 1, characterized in that: the step of obtaining an effective signal segment according to the processed surface electromyographic signal and a preset threshold specifically comprises:
sequentially judging whether the processed surface electromyographic signals are larger than a preset threshold value or not;
when the processed surface electromyographic signal is larger than a preset threshold value, recording a time point as an action starting point;
after the action starting point, when the processed surface electromyogram signal is not greater than a preset threshold value, recording a time point as an action ending point, and repeatedly and sequentially judging whether the processed surface electromyogram signal is greater than the preset threshold value to act; and the time period between the action starting point and the action ending point is an effective signal segment.
3. The surface myoelectricity-based human-computer interaction training method of claim 2, characterized in that: the step of sequentially judging whether the processed surface electromyographic signals are greater than a preset threshold specifically comprises:
acquiring a noise threshold and an electrocardio interference threshold; wherein the noise threshold is an average of absolute values of background noise band levels; the electrocardio-interference threshold is a Short-time Zero-crossing Ratio (ZCR) threshold;
and sequentially judging whether the processed surface electromyographic signals are larger than a noise threshold and an electrocardio-interference threshold.
4. The surface myoelectricity-based human-computer interaction training method of claim 1, characterized in that: the step of extracting the action features of the effective signal segment and identifying the muscle type sending the action features according to the action features of the effective signal segment specifically comprises the following steps:
calculating a Root Mean Square (RMS) value by a first formula, wherein the first formula is:
n is the Data total point number of the surface electromyographic signals of the effective signal segment, and Data (i) represents the ith surface electromyographic signal Data point of the surface electromyographic signals of the effective signal segment;
and extracting the action characteristics of the effective signal section through the electromyographic root mean square value.
5. The surface myoelectricity-based human-computer interaction training method of claim 1, characterized in that: the step of extracting the action features of the effective signal segment and identifying the muscle type sending the action features according to the action features of the effective signal segment specifically comprises the following steps:
carrying out binarization processing on the action characteristic sequence of the effective signal segment by regions to obtain a character string SX；
Calculating a string SXThe number of new modes C (n);
normalizing the number C (n) of the new modes to obtain the LZC complexity; wherein, the LZC complexity is expressed by a formula two, and the formula two is as follows:
n is the total number of data points of the surface electromyographic signals of the effective signal segments;
and extracting the action characteristics of the effective signal segment through the LZC complexity.
6. The surface myoelectricity-based human-computer interaction training method of claim 5, wherein: the calculation character string SXThe step of counting the number C (n) of new modes specifically includes:
setting a character string S (S) to be solved1,s2,...,sn) And another character string Q (Q)1,q2,...,qn)；
Initializing S to a binarization sequence SXAnd Q is the binarization sequence SXThe second element of (1);
when Q is a sub-character of SQv, then concatenating the next character of the pending sequence to Q; where SQv is the string of SQ minus the last character, SQ denotes the concatenation of S and Q, SQ ═ S (S)1,s2,...,sn,q1,q2,...,qn)；
When Q is not a substring of SQv, it indicates that Q is a new pattern, then Q is concatenated to S, i.e. S equals SQ, and the number of new patterns is recorded with C (n);
reconstructing Q, and taking binary sequence S from QXRepeatedly judging whether Q is a sub-symbol of SQv and executing corresponding steps until Q is obtained to the binary sequence S to be solvedXThe last element of (2).
7. The surface myoelectricity-based human-computer interaction training method of claim 5, wherein: the step of extracting the action features of the effective signal segment and identifying the muscle type sending the action features according to the action features of the effective signal segment specifically further comprises:
obtaining a known learning sample, and training an offline BP neural network according to the known learning sample to obtain a weight and a threshold required by pattern recognition; the known learning samples comprise a plurality of LZC complexities, muscle types corresponding to the LZC complexities one by one and emitted action characteristics;
and inputting the real-time LZC complexity into the trained offline BP neural network to obtain the action characteristics of the muscle type.
8. The surface myoelectricity-based human-computer interaction training method of claim 4 or 5, wherein: the step of generating a corresponding control instruction according to the action characteristics corresponding to the muscle type specifically includes:
acquiring action characteristics of muscle types; wherein the action characteristics of the muscle type comprise an electromyographic root mean square value or an LZC complexity;
and the game server generates a corresponding control instruction according to the action characteristics of the muscle type.
9. The human-computer interaction training device based on the surface myoelectricity is characterized in that: the human-computer interaction training device based on the surface myoelectricity comprises:
the system comprises at least one electromyographic signal collector, a data processing unit and a data processing unit, wherein the at least one electromyographic signal collector is used for collecting surface electromyographic signals generated by a target object; wherein the surface electromyographic signals comprise surface electromyographic signals in a static state and a motion state;
the band-pass digital filter is electrically connected with the at least one electromyographic signal collector respectively and is used for preprocessing the surface electromyographic signals to obtain processed surface electromyographic signals;
the terminal is used for obtaining an effective signal section according to the processed surface electromyographic signal and a preset threshold value;
the terminal is further configured to: extracting the action characteristics of the effective signal segment, and identifying the muscle type sending the action characteristics according to the action characteristics of the effective signal segment;
generating a corresponding control instruction according to the action characteristics of the muscle type;
and making a corresponding response action according to the corresponding control instruction.
10. A storage medium, characterized by: the storage medium is stored with a surface myoelectricity-based human-computer interaction training program, which when executed by at least one processor implements the surface myoelectricity-based human-computer interaction training method according to claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911182105.2A CN111176441A (en) | 2019-11-27 | 2019-11-27 | Surface myoelectricity-based man-machine interaction training method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911182105.2A CN111176441A (en) | 2019-11-27 | 2019-11-27 | Surface myoelectricity-based man-machine interaction training method and device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111176441A true CN111176441A (en) | 2020-05-19 |
Family
ID=70648839
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911182105.2A Pending CN111176441A (en) | 2019-11-27 | 2019-11-27 | Surface myoelectricity-based man-machine interaction training method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111176441A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112244882A (en) * | 2020-10-30 | 2021-01-22 | 北京中科心研科技有限公司 | Disease early warning method and device based on multi-modal physiological data |
| CN112826516A (en) * | 2021-01-07 | 2021-05-25 | 京东数科海益信息科技有限公司 | Electromyographic signal processing method, device, equipment, readable storage medium and product |
| CN114138111A (en) * | 2021-11-11 | 2022-03-04 | 深圳市心流科技有限公司 | Full-system control interaction method of myoelectric intelligent bionic hand |
| CN114388100A (en) * | 2022-01-14 | 2022-04-22 | 清华大学 | Method and device for establishing muscle mapping model and computer equipment |
| CN114676737A (en) * | 2022-05-26 | 2022-06-28 | 深圳市心流科技有限公司 | Dynamic regulation method for sampling frequency of electromyographic signal |
| CN115299950A (en) * | 2022-08-12 | 2022-11-08 | 歌尔股份有限公司 | Electromyographic signal acquisition circuit, wearable device and control method |
| WO2023102908A1 (en) * | 2021-12-10 | 2023-06-15 | 深圳大学 | Multi-modal strength training assistance method and system |
| CN116725556A (en) * | 2023-07-12 | 2023-09-12 | 中国科学院苏州生物医学工程技术研究所 | Motion intention recognition method and device based on surface electromyographic signals |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103617411A (en) * | 2013-10-17 | 2014-03-05 | 杭州电子科技大学 | Myoelectricity signal identification method based on complexity, fractal dimension and fractal length |
| KR20140109104A (en) * | 2013-03-05 | 2014-09-15 | 삼성전자주식회사 | Emg sensor system, and method of operation of emg sensor system |
| CN104107134A (en) * | 2013-12-10 | 2014-10-22 | 中山大学 | Myoelectricity feedback based upper limb training method and system |
| CN108324272A (en) * | 2018-03-13 | 2018-07-27 | 天津大学 | A kind of device for checking and examining anxiety degree based on LZC complexities |
| CN108629311A (en) * | 2018-05-02 | 2018-10-09 | 尚谷科技（天津）有限公司 | A kind of action identification method based on biological pulsation |
| CN109101910A (en) * | 2018-07-31 | 2018-12-28 | 湖南师范大学 | A kind of Magnetotelluric signal denoising method screened based on noise |
-
2019
- 2019-11-27 CN CN201911182105.2A patent/CN111176441A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140109104A (en) * | 2013-03-05 | 2014-09-15 | 삼성전자주식회사 | Emg sensor system, and method of operation of emg sensor system |
| CN103617411A (en) * | 2013-10-17 | 2014-03-05 | 杭州电子科技大学 | Myoelectricity signal identification method based on complexity, fractal dimension and fractal length |
| CN104107134A (en) * | 2013-12-10 | 2014-10-22 | 中山大学 | Myoelectricity feedback based upper limb training method and system |
| CN108324272A (en) * | 2018-03-13 | 2018-07-27 | 天津大学 | A kind of device for checking and examining anxiety degree based on LZC complexities |
| CN108629311A (en) * | 2018-05-02 | 2018-10-09 | 尚谷科技（天津）有限公司 | A kind of action identification method based on biological pulsation |
| CN109101910A (en) * | 2018-07-31 | 2018-12-28 | 湖南师范大学 | A kind of Magnetotelluric signal denoising method screened based on noise |
Non-Patent Citations (2)
| Title |
|---|
| 中国航空学会信息融合分会: "《第六届中国信息融合大会论文集 下》", 30 November 2014 * |
| 徐伟钊: "一种多通道表面肌电采集仪的研制方法", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112244882A (en) * | 2020-10-30 | 2021-01-22 | 北京中科心研科技有限公司 | Disease early warning method and device based on multi-modal physiological data |
| CN112244882B (en) * | 2020-10-30 | 2023-06-02 | 北京中科心研科技有限公司 | Disease early warning method and device based on multi-mode physiological data |
| CN112826516A (en) * | 2021-01-07 | 2021-05-25 | 京东数科海益信息科技有限公司 | Electromyographic signal processing method, device, equipment, readable storage medium and product |
| CN114138111A (en) * | 2021-11-11 | 2022-03-04 | 深圳市心流科技有限公司 | Full-system control interaction method of myoelectric intelligent bionic hand |
| CN114138111B (en) * | 2021-11-11 | 2022-09-23 | 深圳市心流科技有限公司 | Full-system control interaction method of myoelectric intelligent bionic hand |
| WO2023102908A1 (en) * | 2021-12-10 | 2023-06-15 | 深圳大学 | Multi-modal strength training assistance method and system |
| CN114388100A (en) * | 2022-01-14 | 2022-04-22 | 清华大学 | Method and device for establishing muscle mapping model and computer equipment |
| CN114388100B (en) * | 2022-01-14 | 2022-09-02 | 清华大学 | Method and device for establishing muscle mapping model and computer equipment |
| CN114676737A (en) * | 2022-05-26 | 2022-06-28 | 深圳市心流科技有限公司 | Dynamic regulation method for sampling frequency of electromyographic signal |
| CN115299950A (en) * | 2022-08-12 | 2022-11-08 | 歌尔股份有限公司 | Electromyographic signal acquisition circuit, wearable device and control method |
| CN116725556A (en) * | 2023-07-12 | 2023-09-12 | 中国科学院苏州生物医学工程技术研究所 | Motion intention recognition method and device based on surface electromyographic signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111176441A (en) | Surface myoelectricity-based man-machine interaction training method and device and storage medium | |
| Khushaba et al. | A framework of temporal-spatial descriptors-based feature extraction for improved myoelectric pattern recognition | |
| Reaz et al. | Techniques of EMG signal analysis: detection, processing, classification and applications | |
| Barsakcioglu et al. | Control of spinal motoneurons by feedback from a non-invasive real-time interface | |
| Dai et al. | Extracting and classifying spatial muscle activation patterns in forearm flexor muscles using high-density electromyogram recordings | |
| CN108681396A (en) | Man-machine interactive system and its method based on brain-myoelectricity bimodal nerve signal | |
| Dai et al. | Prediction of individual finger forces based on decoded motoneuron activities | |
| US9907489B2 (en) | Systems and methods for hierarchical pattern recognition for simultaneous control of multiple-degree of freedom movements for prosthetics | |
| JP2021194540A (en) | System and method for detecting stable arrhythmia heartbeat and for calculating and detecting cardiac mapping annotations | |
| Kshirsagar et al. | Classification of ECG-signals using artificial neural networks | |
| CN114822761A (en) | Wrist rehabilitation training system based on muscle cooperation and variable stiffness impedance control | |
| Jaber et al. | HD-sEMG gestures recognition by SVM classifier for controlling prosthesis | |
| Yu et al. | Wrist torque estimation via electromyographic motor unit decomposition and image reconstruction | |
| Naik et al. | Evaluation of higher order statistics parameters for multi channel sEMG using different force levels | |
| Guo et al. | Towards semi-supervised myoelectric finger motion recognition based on spatial motor units activation | |
| Shukri et al. | Investigation on Elman neural network for detection of cardiomyopathy | |
| Rivela et al. | Processing of surface EMG through pattern recognition techniques aimed at classifying shoulder joint movements | |
| Xu et al. | ART2 neural network for surface EMG decomposition | |
| Amiruddin et al. | Feature reduction and arrhythmia classification via hybrid multilayered perceptron network | |
| Herle | Movement intention detection from SEMG signals using time-domain features and discriminant analysis classifiers | |
| Zhao et al. | Online Decomposition of Surface Electromyogram into Individual Motor Unit Activities Using Progressive FastICA Peel-off | |
| CN115607167A (en) | Lightweight model training method, atrial fibrillation detection method, device and system | |
| Ouelli et al. | Multivariate autoregressive modeling for cardiac arrhythmia classification using multilayer perceptron neural networks | |
| Ali et al. | Identification of cardiomyopathy disease using hybrid multilayered perceptron network | |
| CN112773379B (en) | Myoelectric signal processing method, processing equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200519 |
|
| RJ01 | Rejection of invention patent application after publication |