CN114626405A - Real-time identity recognition method and device based on electromyographic signals and electronic equipment - Google Patents
Real-time identity recognition method and device based on electromyographic signals and electronic equipment Download PDFInfo
- Publication number
- CN114626405A CN114626405A CN202210114585.4A CN202210114585A CN114626405A CN 114626405 A CN114626405 A CN 114626405A CN 202210114585 A CN202210114585 A CN 202210114585A CN 114626405 A CN114626405 A CN 114626405A
- Authority
- CN
- China
- Prior art keywords
- target
- electromyographic
- time
- real
- electromyographic signals
- 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
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 27
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 27
- 238000004590 computer program Methods 0.000 claims description 17
- 238000013507 mapping Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 7
- 230000001960 triggered effect Effects 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 6
- 230000003183 myoelectrical effect Effects 0.000 description 36
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000003491 array Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000002567 electromyography Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 210000003205 muscle Anatomy 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 210000000554 iris Anatomy 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000005452 bending Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 210000003169 central nervous system Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000035479 physiological effects, processes and functions Effects 0.000 description 1
- 230000035484 reaction time Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/389—Electromyography [EMG]
- A61B5/395—Details of stimulation, e.g. nerve stimulation to elicit EMG response
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Public Health (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Surgery (AREA)
- Medical Informatics (AREA)
- Heart & Thoracic Surgery (AREA)
- Computational Linguistics (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Fuzzy Systems (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention provides a real-time identity recognition method and device based on an electromyographic signal and an electronic device, wherein the real-time identity recognition method based on the electromyographic signal comprises the following steps: acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal; determining a target time-frequency characteristic of the target electromyographic signal; and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information. The method can realize that the terminal equipment receives the original electromyographic signals collected by the electromyographic bracelet in real time and performs time-frequency feature extraction and identity recognition processing on the original electromyographic signals received in real time, thereby realizing the purpose of performing real-time identity recognition based on the electromyographic signals and improving the recognition rate of the identity information on the premise of meeting the real-time property.
Description
Technical Field
The invention relates to the technical field of biological feature recognition, in particular to a real-time identity recognition method and device based on an electromyographic signal and an electronic device.
Background
In the continuous and rapid development of internet technology, in order to ensure the security of personal information, a biometric identification technology is increasingly paid attention by people and is widely used, for example, biometric features of human unique physiological features such as human faces, fingerprints, irises, voices and the like are applied to various identity identification systems such as credit card transactions, Automatic Teller Machine (ATM) security, bank security, smart phones and the like for identity identification, but considering that the biometric features are easy to forge, recreate and forge, how to improve the security of personal identities is still a popular research object.
In the related art, identification is performed based on an electromyographic signal, the event electromyographic signal is first divided from the entire electromyographic signal stream, and then the event electromyographic signal is processed using an artificial Neural network (CNN) or a Convolutional Neural Network (CNN), so as to obtain identification information.
However, the existing identification methods based on electromyographic signals all process electromyographic signals offline, and researchers focus on only improving the identification rate of identifying the identities based on the electromyographic signals offline, but neglecting the requirements of practical application scenarios, so that the real-time requirements cannot be met on the premise of ensuring the identification rate.
Disclosure of Invention
The invention provides a real-time identity recognition method and device based on an electromyographic signal and electronic equipment, which are used for overcoming the defect that the real-time requirement cannot be met on the premise of ensuring the recognition rate when the identity recognition is carried out based on the electromyographic signal in the prior art, and improving the recognition rate of identity information on the premise of meeting the real-time requirement when the identity recognition is carried out based on the electromyographic signal.
The invention provides a real-time identity recognition method based on electromyographic signals, which comprises the following steps:
acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
determining a target time-frequency characteristic of the target electromyographic signal;
and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
According to the real-time identity recognition method based on the electromyographic signals, the method for acquiring the target electromyographic signals representing limb movement aiming at the acquired original electromyographic signals comprises the following steps:
carrying out sliding window segmentation processing on the acquired original electromyographic signals to obtain sliding window electromyographic signals corresponding to each sliding window;
calculating the average absolute value of the electromyographic signals of the sliding window;
and obtaining a target electromyographic signal representing limb movement based on the average absolute value and a preset dynamic threshold value.
According to the real-time identity recognition method based on the electromyographic signals, the method for obtaining the target electromyographic signals representing limb movement based on the average absolute value and the preset dynamic threshold comprises the following steps:
selecting a reference sliding window electromyographic signal from the plurality of sliding window electromyographic signals;
determining a preset dynamic threshold value based on a reference average absolute value of the reference sliding window electromyographic signal;
and determining a corresponding target electromyographic signal representing limb movement when the average absolute value is larger than the preset dynamic threshold value.
According to the real-time identity recognition method based on the electromyographic signals, the step of determining the target time-frequency characteristics of the target electromyographic signals comprises the following steps:
dividing the target electromyographic signals into a plurality of continuous target sub-frequency bands;
performing preset feature extraction on each target sub-frequency band to obtain target time-frequency features of the target electromyographic signals; the preset features comprise entropy features, percentile features, median features, mean features, variance features, standard deviation features, root mean square values, zero crossing rate features and average zero crossing rate features.
According to the real-time identity recognition method based on the electromyographic signals, provided by the invention, when the target time-frequency feature is one-dimensional, the target time-frequency feature is recognized and processed based on the preset convolution neural network to obtain target identity information, and the method comprises the following steps:
inputting the target time-frequency characteristics into a preset one-dimensional convolutional neural network for convolution, pooling and full-connection processing to obtain target characteristic vectors;
and obtaining target identity information corresponding to the target feature vector based on a preset mapping relation between the feature vector and the identity identification information.
According to the real-time identity recognition method based on the electromyographic signals, before the step of acquiring the target electromyographic signals representing limb movement aiming at the collected original electromyographic signals, the method further comprises the following steps:
acquiring original electromyographic signals collected when a user executes a limb movement task within a target time length, wherein the original electromyographic signals are electromyographic signal streams and comprise electromyographic signals triggered by limb movement and electromyographic signals in a non-limb movement state.
The invention also provides a real-time identity recognition device based on the electromyographic signals, which comprises:
the acquisition module is used for acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
the determining module is used for determining the target time-frequency characteristics of the target electromyographic signals;
and the identification module is used for identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of any one of the electromyographic signal-based real-time identity recognition methods.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when being executed by a processor, implements the steps of the method for real-time identification based on electromyographic signals, as set forth in any of the above.
The present invention also provides a computer program product comprising a computer program which, when being executed by a processor, implements the steps of any of the above-mentioned electromyographic signal based real-time identity recognition methods.
According to the electromyographic signal-based real-time identity recognition method, the electromyographic signal-based real-time identity recognition device and the electronic equipment, a target electromyographic signal representing limb movement is obtained according to an acquired original electromyographic signal, then a target time-frequency characteristic of the target electromyographic signal is determined, and the target time-frequency characteristic is recognized and processed based on a preset convolutional neural network to obtain target identity information. Therefore, the purpose of real-time identification based on the electromyographic signals is achieved, and the identification rate of the identity information can be improved on the premise of meeting the real-time property.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a real-time identification method based on electromyographic signals according to the present invention;
FIG. 2 is a schematic diagram of the principle of decomposing a target electromyographic signal using discrete wavelet transform according to the present invention;
FIG. 3 is a schematic structural diagram of a real-time identification device based on electromyographic signals according to the present invention;
fig. 4 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
With the rapid development of scientific technology, in order to ensure the security of personal information, people increasingly pay more attention to biometric identification technology, and many biometric identification methods are derived based on biometric identification technology, and can identify a person or verify the claimed identity of a person by using unique physical features. At present, morphological and biological characteristics such as human face, fingerprint, iris, voice and the like are widely applied to various identity recognition systems, such as credit card transaction, ATM security, bank security, smart phones and the like. However, biometric identification methods that utilize the unique physiological characteristics of the human body may be attacked because these characteristics can still be forged, recreated, and forged. For example, voice authentication systems are susceptible to spoofing attacks, i.e., imposters attempting to spoof the authentication system through pre-recorded or synthesized victim voice samples.
Biometric identification using living body features is one method to overcome the shortcomings of traditional biometric identification techniques. Electromyographic signals are complex signals controlled by the nervous system and depend on anatomical and physiological characteristics of the muscle. Since the direct link between intact muscle, intact central nervous system and brain is individual and directly related to each person's physiology, it is universal, stable, unique and measurable. Thus, electromyographic signals may be used in an identification system. In recent years, the electromyography-based identity recognition method is more researched, for example, an artificial neural network algorithm is adopted for processing, an identity recognition experiment is carried out through electromyography signals acquired by electrodes, and the identity recognition accuracy is 81.6%; for example, the CNN algorithm is adopted for processing, and the identification experiment is carried out through the electromyographic signals acquired by the electrodes, so that the identification accuracy can be improved to 95.0%. However, since the existing identification systems based on electromyographic signals are all based on off-line processing, the electromyographic signals to be processed are completely segmented from the whole signals in advance. And thus are not suitable for real-life scenarios.
In summary, the currently existing identity recognition methods based on electromyographic signals are all based on offline processing, and researchers focus on the situation of recognition rate based on electromyographic identity recognition only in an offline state, and ignore the need of practical application scenarios.
Based on the method, the device and the electronic equipment, the real-time identity recognition method and device based on the electromyographic signals are provided, and the real-time property is met while the recognition rate is guaranteed. The real-time identification method, device and electronic device based on electromyographic signals according to the present invention are described below with reference to fig. 1 to 4.
Referring to fig. 1, a schematic flow chart of a real-time identity recognition method based on an electromyographic signal according to the present invention is shown, where an execution main body of the real-time identity recognition method based on an electromyographic signal may be a real-time identity recognition device based on an electromyographic signal, and the real-time identity recognition device based on an electromyographic signal may be implemented as part or all of a terminal device in a software, hardware, or a combination of software and hardware, and the terminal device at least has an identity recognition function. Alternatively, the terminal device may be a Personal Computer (PC), a portable device, a notebook Computer, a smart phone, a tablet Computer, a portable wearable device, or other electronic devices. The invention does not limit the concrete form of the terminal equipment.
It should be noted that the execution subject of the method embodiments described below may be part or all of the terminal device described above. The following method embodiments take the execution subject as an example of the terminal device.
As shown in fig. 1, the real-time identification method based on electromyographic signals includes the following steps:
and 110, acquiring a target electromyogram signal representing limb movement aiming at the acquired original electromyogram signal.
Specifically, the terminal equipment can receive an original myoelectric signal acquired by the myoelectric bracelet, the myoelectric bracelet can be preferably an MYO myoelectric bracelet, the MYO myoelectric bracelet is a myoelectric arm band developed by Thalmic laboratories and used for acquiring the myoelectric signal, and the MYO myoelectric bracelet has the characteristics of low consumption, low cost, small size, light weight, convenience in wearing and the like. Therefore, the MYO myoelectric bracelet can be worn on the arm of an experimenter, 8 electrodes in the MYO myoelectric bracelet record original myoelectric signals of the experimenter in real time in the process of demonstrating different gesture actions (such as palm opening, fist making, left bending and the like), and the captured original myoelectric signals are transmitted to the terminal device through the Bluetooth serial port and can be stored in a file form or directly processed. Because the electromyographic signals representing non-arm movements may exist in the original electromyographic signals, the start time identification and the end time identification of the arm movements may also not exist. Therefore, in order to ensure the accuracy of identity recognition, the terminal device does not directly perform online processing on the original electromyographic signals, but acquires target electromyographic signals representing limb movements (such as arm movements) from the original electromyographic signals, so as to improve the accuracy of identity recognition.
And step 120, determining the target time-frequency characteristics of the target electromyographic signals.
Specifically, the terminal device may determine, for a target electromyographic signal representing limb movement, a target time-frequency characteristic of the target electromyographic signal using a discrete wavelet transform method, because the discrete wavelet transform may decompose the target electromyographic signal into an approximation (low frequency) coefficient and a detail (high frequency) coefficient. For N-level discrete wavelet decomposition, 1 approximate coefficient array and N detail coefficient arrays can be obtained, wherein N is a positive integer; as shown in fig. 2, when the value of N is 2, the target electromyogram signal may be decomposed twice to obtain 1 approximate coefficient array and 2 detail coefficient arrays, and feature extraction is performed based on each coefficient array, where the feature extraction is usually used to extract identification information in the target electromyogram signal, that is, features representing the identity of an experimenter, so as to achieve the purposes of minimizing complexity and reducing signal processing cost.
And step 130, identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
Specifically, the terminal device inputs the target time-frequency characteristics of the target electromyographic signals into a preset convolutional neural network for identity recognition, and the convolutional neural network is added with an attention mechanism and derived based on an artificial neural network, so that the convolutional neural network has very strong expression capacity, the required preset convolutional neural network can be obtained after the convolutional neural network is trained, and the preset convolutional neural network has a mapping function from the characteristics to identity information. Based on the above, the target time-frequency characteristics are input to the preset convolutional neural network, so that the height mapping of the target time-frequency characteristics can be completed, and the target identity information can be obtained.
It should be noted that, since the training convolutional neural network is already referred to in many published documents or patents, it is not described herein in detail.
According to the real-time identity recognition method based on the electromyographic signals, the target electromyographic signals representing limb movement are obtained according to the collected original electromyographic signals, then the target time-frequency characteristics of the target electromyographic signals are determined, and the target time-frequency characteristics are recognized and processed based on the preset convolutional neural network to obtain the target identity information. The myoelectric bracelet can receive the original myoelectric signals collected by the myoelectric bracelet in real time by the terminal equipment, and the original myoelectric signals received in real time are subjected to time-frequency feature extraction and identity recognition processing, so that the purpose of real-time identity recognition based on the myoelectric signals is realized, and the recognition rate of identity information can be improved on the premise of meeting the real-time property.
Optionally, before step 110, the method may further include:
acquiring original electromyographic signals collected when a user executes a limb movement task within a target time length, wherein the original electromyographic signals are electromyographic signal streams and comprise electromyographic signals triggered by limb movement and electromyographic signals in a non-limb movement state.
The target duration can be determined based on the number of times of repeated execution of the tasks by the user and the acquisition time of each task. For example, when the tasks are repeatedly executed 2 times and the acquisition time of each task is 8 seconds, the corresponding target time length is 16 seconds. The number of users may be plural and each user may be an experimenter.
Specifically, when the limb movement is a gesture movement and the myoelectric signal is determined to be acquired based on the myoelectric bracelet (such as the myoelectric bracelet), the myoelectric bracelet can be worn at the height of the radial humeral joint of at least one experimenter, so as to record the myoelectric signal representing muscle movement triggered when the experimenter performs gesture movement tasks such as a fist making gesture and a palm stretching gesture in real time and the myoelectric signal under the non-gesture movement state that the experimenter does not make any gesture. And each experimenter repeats the same task M times, wherein M is a positive integer, the acquisition time of each task can be 8 seconds, each experimenter needs to finish one or more gesture opening actions within 8 seconds, and the time lasts for 1.5 s. Illustratively, when an myoelectric signal is acquired by using a myoelectric bracelet, the sampling rate of the myoelectric bracelet is 200Hz/s, that is, each sample data is 200 × 8 × 8, each experimenter has M sample data, at this time, the M sample data acquired by each experimenter is an original myoelectric signal acquired by the corresponding myoelectric bracelet within a target time length, each original myoelectric signal is a myoelectric signal stream, and each myoelectric data stream can be transmitted to a terminal device through a bluetooth serial port in real time, so that the terminal device receives the original myoelectric signal acquired by the user during executing a limb movement task within the target time length.
It should be noted that the terminal device may preset a target duration, or may set the target duration when the myoelectric bracelet is worn on the arm of each experimenter, so that the electrodes on the surface of the myoelectric bracelet perform signal acquisition within the target duration, and a task execution instruction can be set in the terminal device, the task execution instruction carries limb movement task information, the limb movement task information comprises tasks to be executed, the number of times of task repeated execution and the continuous execution time of each task, the task execution instruction can be automatically generated based on the condition that the arms of the experimenter wear the myoelectric bracelet, the limb movement task information carried during the generation of the task execution instruction can be presented in a text and/or audio manner, thereby enabling the experimenter to perform the limb movement task based on the seen and/or heard limb movement task information.
According to the real-time identity recognition method based on the electromyographic signals, the original electromyographic signals are obtained in a mode that a user executes limb movement tasks within the target time length. Because the original electromyographic signals received by the terminal equipment are electromyographic signal streams and comprise the electromyographic signals triggered by limb movement and the electromyographic signals in a non-limb movement state, the terminal equipment can process the original electromyographic signals on line, and the real-time requirement of an actual application scene is met.
Optionally, the specific implementation process of step 110 may include:
firstly, carrying out sliding window segmentation processing on an acquired original electromyographic signal to obtain a sliding window electromyographic signal corresponding to each sliding window; thirdly, calculating the average absolute value of the sliding window electromyographic signal; and finally, obtaining a target electromyographic signal representing limb movement based on the average absolute value and a preset dynamic threshold value.
Specifically, the raw electromyographic signals acquired by the terminal device are electromyographic signal streams and include electromyographic signals triggered by limb movement and electromyographic signals in a non-limb movement state, and because the raw electromyographic signals do not indicate when a gesture starts or ends and only once identification is performed on user gestures in an actual application scene, a limb detection method is used for extracting target electromyographic signals representing limb movement from the raw electromyographic signals to serve as switches for identity identification according to design standards of minimum memory, minimum power consumption budget and minimum processing time in combination with a system with an identity identification function in the terminal device. Moreover, when the limb detection method is specifically a gesture detection method, a peak detection algorithm can be further adopted to detect a target electromyographic signal corresponding to the gesture time in real time, namely, a sliding window is adopted to perform segmentation processing on an original electromyographic signal, two parameters, namely the size of the sliding window and the step length of the adjacent sliding window, are determined according to the recognition rate and the reaction time of the whole recognition system, and then the average absolute value of the electromyographic signal of the sliding window corresponding to the sliding window is calculated; since the mean absolute value feature is used to measure the distance of a group of numbers from their mean value in a sliding window, the larger the mean absolute value, the higher the probability of an event occurring. Therefore, a method of searching a local maximum value by using an average absolute value of electromyographic signals of adjacent sliding windows and a preset dynamic threshold value can be adopted to detect a limb movement signal (such as gesture detection), so as to determine a target electromyographic signal representing limb movement from the original electromyographic signals.
Note that when the sliding window myoelectric signal is represented by X ═ X1,x2,...,xmIn time, the formula for calculating the mean absolute value of the electromyographic signal of the sliding window may be:
xifor the ith data in the sliding window electromyogram signal,
the mean value of the electromyographic signal X of the sliding window is shown, and the MAD is the mean absolute value of the electromyographic signal X of the sliding window.
According to the real-time identity recognition method based on the electromyographic signals, the purpose of determining the target electromyographic signals representing limb movement according to the change of the adjacent mean absolute values and the relation with the preset dynamic threshold value is achieved by performing sliding window segmentation processing on the acquired original electromyographic signals and calculating the mean absolute value of the sliding window electromyographic signals corresponding to each sliding window, so that not only can limb movement events be detected in real time, but also a foundation can be laid for the subsequent real-time identity recognition.
Optionally, obtaining a target electromyographic signal representing limb movement based on the average absolute value and a preset dynamic threshold includes:
firstly, selecting a reference sliding window electromyographic signal from a plurality of sliding window electromyographic signals; then, determining a preset dynamic threshold value based on a reference average absolute value of the reference sliding window electromyographic signal; and finally, determining a corresponding target electromyographic signal representing limb movement when the average absolute value is larger than the preset dynamic threshold value.
Specifically, in order to avoid interference of noise signals, a preset dynamic threshold is adopted as a determination standard, that is, for the sliding window electromyogram signal corresponding to each sliding window and the mean absolute value of each sliding window electromyogram signal, the terminal device may select a reference sliding window electromyogram signal from the plurality of sliding window electromyogram signals, for example, the sliding window electromyogram signal corresponding to a first sliding window is used as the reference sliding window electromyogram signal, then the mean absolute value of the reference sliding window electromyogram signal is used as the mean absolute value, and the reference mean absolute value of the preset multiple is further determined to be the preset dynamic threshold, for example, the preset dynamic threshold may be 3 times of the reference mean absolute value. Then, according to the sequence of the obtained sliding window electromyographic signals (excluding the reference sliding window electromyographic signals), the average absolute values of the corresponding sliding window electromyographic signals are sequentially compared with a dynamic preset threshold value, and when the average absolute value greater than the dynamic preset threshold value appears for the first time, the corresponding sliding window electromyographic signals can be determined as target electromyographic signals representing limb movement.
It should be noted that, when the preset dynamic threshold is determined based on the mean absolute value of the sliding window electromyographic signal corresponding to the first sliding window, the sliding window may first slide twice on the original electromyographic signal and calculate the mean absolute value of the sliding window electromyographic signal corresponding to the sliding window twice, and based on this, determine whether the mean absolute value of the sliding window electromyographic signal corresponding to the sliding window for the second time is greater than the preset dynamic threshold, if not, slide once again and calculate the mean absolute value of the sliding window electromyographic signal corresponding to the sliding window for the third time, and determine whether the mean absolute value of the sliding window electromyographic signal corresponding to the sliding window for the third time is greater than the preset dynamic threshold. And analogizing in this way until a target electromyographic signal representing limb movement greater than a preset dynamic threshold is determined.
According to the real-time identity recognition method based on the electromyographic signals, the reference sliding window electromyographic signals and the reference mean absolute value of the reference sliding window electromyographic signals are selected from the plurality of sliding window electromyographic signals, the preset dynamic threshold value is determined, and the corresponding target electromyographic signals representing limb movement when the mean absolute values are larger than the preset dynamic threshold value are further determined, so that the interference of noise signals is avoided, the accuracy of subsequent identity recognition is ensured, and the real-time requirement is met.
Optionally, the specific implementation process of step 120 may include:
firstly, dividing the target electromyographic signals into a plurality of continuous target sub-frequency bands; and then, extracting preset features of each target sub-frequency band to obtain target time-frequency features of the target electromyographic signals.
The preset features comprise entropy features, percentile features, median features, mean features, variance features, standard deviation features, root mean square values, zero crossing rate features and average zero crossing rate features.
Specifically, considering that the discrete wavelet transform can be implemented as a filter bank and can decompose a signal into a plurality of continuous sub-frequency bands, and compared with a time-frequency characteristic mode for representing the signal by a continuous wavelet transform method, the discrete wavelet transform method avoids the problem that the calculation performance is seriously influenced by redundant parameters in the processing process; in addition, the discrete wavelet transform can simultaneously increase the processing speed. Therefore, the terminal device may divide the obtained target electromyogram signal into a plurality of continuous target sub-frequency bands by using a discrete wavelet transform method, and when performing N-level discrete wavelet decomposition, the target electromyogram signal may be divided into N +1 target sub-frequency bands, where the N sub-frequency bands are specifically characterized by N detail coefficient arrays, and the 1 sub-frequency band is specifically characterized by 1 approximate coefficient array.
Then, preset feature extraction is respectively carried out on each target sub-frequency band, in order to minimize the implementation complexity and reduce the signal processing cost, the invention adopts a statistical method, and 12 different types of preset features are respectively extracted aiming at each target sub-frequency band, wherein the method comprises the following steps: the target time-frequency characteristic of the target electromyographic signal is determined by the entropy characteristic, the 5 th percentile characteristic, the 25 th percentile characteristic, the 75 th percentile characteristic, the 95 th percentile characteristic, the median characteristic, the mean characteristic, the variance characteristic, the standard deviation characteristic, the root mean square value characteristic, the zero-crossing rate characteristic and the average zero-crossing rate characteristic, and 12 different types of preset characteristics extracted from each target sub-frequency band are determined as the target time-frequency characteristic of the target electromyographic signal.
According to the real-time identity recognition method based on the electromyographic signals, the target time-frequency characteristics of the target electromyographic signals are obtained by dividing the target electromyographic signals into a plurality of continuous target sub-frequency bands and then extracting the preset characteristics of each target sub-frequency band, and the computing performance and the processing speed of the terminal equipment are improved. Furthermore, the preset features comprise entropy features, percentile features, median features, mean features, variance features, standard deviation features, root mean square values features, zero crossing rate features and average zero crossing rate features, so that the integrity and diversity of the features required by identity recognition can be realized, and a powerful basis is provided for the accuracy of subsequent identity recognition.
Optionally, when the target time-frequency feature is one-dimensional, the specific implementation process of step 130 may include:
firstly, inputting target time-frequency characteristics into a preset one-dimensional convolution neural network for convolution, pooling and full-connection processing to obtain target characteristic vectors; and then, obtaining target identity information corresponding to the target feature vector based on a mapping relation between a preset feature vector and identity identification information.
Specifically, because the one-dimensional convolutional neural network achieves high-level performance in a plurality of applications such as biomedical data classification, structural health monitoring and early diagnosis, the conventional two-dimensional deep convolutional neural network only operates on two-dimensional data, and cannot operate on one-dimensional data. Compared with a two-dimensional deep convolution neural network, the one-dimensional deep convolution neural network is relatively easy to train, has low calculation complexity and is suitable for real-time low-cost application. Therefore, the identity information of the corresponding experimenter is identified by the preset one-dimensional convolutional neural network obtained after the one-dimensional convolutional neural network is trained. That is, the preset one-dimensional convolutional neural network can be used for identity recognition, and can realize high mapping of the characteristics of the one-dimensional target time-frequency characteristics to identity information by means of strong expression capacity, so that the identity information of an experimenter can be recognized.
Similar to the traditional two-dimensional convolutional neural network, the preset one-dimensional convolutional neural network mainly comprises three layers, including a convolutional layer, a pooling layer and a full-connection layer. The input layer receives one-dimensional characteristic signals, and the output layer is a full-connection layer with the neuron number equal to the identification class number.
It should be noted that the main difference between the one-dimensional convolutional neural network and the two-dimensional convolutional neural network is that a two-dimensional matrix of a convolution kernel and a feature map is replaced by a one-dimensional array. Therefore, the preset one-dimensional convolutional neural network not only has very strong expression capacity of the two-dimensional convolutional neural network, can represent the mapping from the characteristics to the identity information, but also is easier to train and smaller in calculation complexity, so that the real-time requirement can be better met.
Exemplarily, when the target time-frequency characteristic of the target electromyographic signal is one-dimensional, the target time-frequency characteristic can be expressed as a one-dimensional convolution input vector [ b, c, w ] and input into a preset one-dimensional convolution neural network, and the convolution layer of the input vector is firstly subjected to one-dimensional convolution processing to obtain a convolution output vector [ b, c', w ]; wherein b is the batch size, c is the number of channels, w is the length of the input one-dimensional target time-frequency feature, c' is the number of convolution kernels in the convolution layer, the size of each convolution kernel is [ k, c ], and k is the width of each convolution kernel. Pooling the convolution output vector [ b, c', w ], and obtaining a pooled output vector [ b, c, w/2] if the pooled layer is processed by 2 times of down sampling; and finally, carrying out full-connection processing on the pooled output vector [ b, c, w/2], wherein a general full-connection layer is the last layer of a preset one-dimensional convolutional neural network, and when the width of the pooled output vector [ b, c, w/2] input to the full-connection layer is 1, obtaining a target characteristic vector [ b, n ], wherein n is the number of target identity information. Finally, target Identity information corresponding to the target feature vector is obtained based on a preset mapping relationship between the feature vector and the Identity identification information, and target Identity information of each experimenter, such as an Identity document (id), a name and the like, is obtained.
It should be noted that the mapping relationship between the feature vector and the identity identification information may be preset and stored in the preset one-dimensional convolutional neural network, and the mapping relationship may be the mapping relationship between a plurality of experimenters performing experiments in batch and the identity information thereof, so that the corresponding target identity information can be identified when the preset one-dimensional convolutional neural network outputs the target feature vector [ b, n ].
The invention provides a real-time identity recognition method based on electromyographic signals, which comprises the steps of inputting target time-frequency characteristics into a preset one-dimensional convolutional neural network for convolution, pooling and full-connection processing, then outputting target characteristic vectors, further obtaining target identity information corresponding to the target characteristic vectors based on the mapping relation between the preset characteristic vectors and the identity recognition information, and achieving the purpose of rapidly obtaining the target identity information by combining the identity recognition function, the strong attention mechanism and the mapping relation of the preset one-dimensional convolutional neural network.
The real-time identification device based on the electromyographic signals provided by the invention is described below, and the real-time identification device based on the electromyographic signals described below and the real-time identification method based on the electromyographic signals described above can be referred to correspondingly.
Fig. 3 illustrates an electromyographic signal based real-time identification apparatus, as shown in fig. 3, the electromyographic signal based real-time identification apparatus 300 comprising: an obtaining module 310, configured to obtain a target electromyographic signal representing a limb movement for an acquired original electromyographic signal; a determining module 320, configured to determine a target time-frequency characteristic of the target electromyographic signal; and the identification module 330 is configured to perform identification processing on the target time-frequency feature based on a preset convolutional neural network to obtain target identity information.
Optionally, the obtaining module 310 may be specifically configured to perform sliding window segmentation processing on the acquired original electromyographic signals to obtain sliding window electromyographic signals corresponding to each sliding window; calculating the average absolute value of the electromyographic signals of the sliding window; and obtaining a target electromyographic signal representing limb movement based on the average absolute value and a preset dynamic threshold value.
Optionally, the determining module 320 may be specifically configured to select a reference sliding window electromyographic signal from the plurality of sliding window electromyographic signals; determining a preset dynamic threshold value based on a reference average absolute value of the reference sliding window electromyographic signal; and determining a corresponding target electromyographic signal representing limb movement when the average absolute value is larger than the preset dynamic threshold value.
Optionally, the determining module 320 may be further configured to divide the target electromyographic signal into a plurality of continuous target sub-frequency bands; performing preset feature extraction on each target sub-frequency band to obtain target time-frequency features of the target electromyographic signals; the preset features comprise entropy features, percentile features, median features, mean features, variance features, standard deviation features, root mean square values, zero crossing rate features and average zero crossing rate features.
Optionally, the identification module 330 may be specifically configured to input the target time-frequency feature to a preset one-dimensional convolutional neural network for convolution, pooling and full-link processing, so as to obtain a target feature vector; and obtaining target identity information corresponding to the target feature vector based on a preset mapping relation between the feature vector and the identity identification information.
Optionally, the obtaining module 310 may be further specifically configured to obtain a raw electromyographic signal collected when the user performs the limb movement task within the target duration, where the raw electromyographic signal is an electromyographic signal stream and includes an electromyographic signal triggered by limb movement and an electromyographic signal in a non-limb movement state.
Fig. 4 illustrates a physical structure diagram of an electronic device, and as shown in fig. 4, the electronic device 400 may include: a processor (processor)410, a communication interface (communication interface)420, a memory (memory)430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 are communicated with each other via the communication bus 440. The processor 410 may invoke logic instructions in the memory 430 to perform a method of real-time identification based on electromyographic signals, the method comprising:
acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
determining a target time-frequency characteristic of the target electromyographic signal;
and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
In addition, the logic instructions in the memory 430 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
In another aspect, the present invention further provides a computer program product, where the computer program product includes a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and when the computer program is executed by a processor, a computer can execute the electromyography-based real-time identification method provided by the above methods, and the method includes:
acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
determining a target time-frequency characteristic of the target electromyographic signal;
and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
In still another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the electromyogram signal-based real-time identification method provided by the above methods, the method including:
acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
determining a target time-frequency characteristic of the target electromyographic signal;
and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A real-time identity recognition method based on electromyographic signals is characterized by comprising the following steps:
acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
determining a target time-frequency characteristic of the target electromyographic signal;
and identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
2. The real-time identification method based on electromyographic signals according to claim 1, wherein the obtaining of the target electromyographic signals representing limb movement for the collected raw electromyographic signals comprises:
carrying out sliding window segmentation processing on the acquired original electromyographic signals to obtain sliding window electromyographic signals corresponding to each sliding window;
calculating the average absolute value of the electromyographic signals of the sliding window;
and obtaining a target electromyographic signal representing limb movement based on the average absolute value and a preset dynamic threshold value.
3. The real-time identification method based on electromyographic signals according to claim 2, wherein obtaining the target electromyographic signal representing limb movement based on the mean absolute value and a preset dynamic threshold comprises:
selecting a reference sliding window electromyographic signal from the plurality of sliding window electromyographic signals;
determining a preset dynamic threshold value based on a reference average absolute value of the reference sliding window electromyographic signal;
and determining a corresponding target electromyographic signal representing limb movement when the average absolute value is larger than the preset dynamic threshold value.
4. The real-time identification method based on electromyographic signals according to claim 1, wherein the determining the target time-frequency characteristics of the target electromyographic signals comprises:
dividing the target electromyographic signals into a plurality of continuous target sub-frequency bands;
performing preset feature extraction on each target sub-frequency band to obtain target time-frequency features of the target electromyographic signals; the preset features comprise entropy features, percentile features, median features, mean features, variance features, standard deviation features, root mean square values, zero crossing rate features and average zero crossing rate features.
5. The method according to claim 1, wherein when the target time-frequency feature is one-dimensional, the identifying process is performed on the target time-frequency feature based on a preset convolutional neural network to obtain target identity information, and the method includes:
inputting the target time-frequency characteristics into a preset one-dimensional convolutional neural network for convolution, pooling and full-connection processing to obtain target characteristic vectors;
and obtaining target identity information corresponding to the target feature vector based on a preset mapping relation between the feature vector and the identity identification information.
6. The real-time electromyographic signal based identity recognition method according to claim 1, wherein prior to said step of obtaining a target electromyographic signal representative of a limb movement for a raw acquired electromyographic signal, said method further comprises:
acquiring original electromyographic signals collected when a user executes a limb movement task within a target time length, wherein the original electromyographic signals are electromyographic signal streams and comprise electromyographic signals triggered by limb movement and electromyographic signals in a non-limb movement state.
7. A real-time identity recognition device based on electromyographic signals, comprising:
the acquisition module is used for acquiring a target electromyographic signal representing limb movement aiming at the acquired original electromyographic signal;
the determining module is used for determining the target time-frequency characteristics of the target electromyographic signals;
and the identification module is used for identifying the target time-frequency characteristics based on a preset convolutional neural network to obtain target identity information.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the electromyographic signal based real-time identification method according to any one of claims 1 to 6 when executing the program.
9. A non-transitory computer-readable storage medium, on which a computer program is stored, wherein the computer program, when being executed by a processor, implements the steps of the method for real-time identification based on electromyographic signals according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the electromyographic signal based real-time identification method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210114585.4A CN114626405A (en) | 2022-01-30 | 2022-01-30 | Real-time identity recognition method and device based on electromyographic signals and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210114585.4A CN114626405A (en) | 2022-01-30 | 2022-01-30 | Real-time identity recognition method and device based on electromyographic signals and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114626405A true CN114626405A (en) | 2022-06-14 |
Family
ID=81898631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210114585.4A Pending CN114626405A (en) | 2022-01-30 | 2022-01-30 | Real-time identity recognition method and device based on electromyographic signals and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114626405A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115114962A (en) * | 2022-07-19 | 2022-09-27 | 歌尔股份有限公司 | Control method and device based on surface electromyogram signal and wearable device |
-
2022
- 2022-01-30 CN CN202210114585.4A patent/CN114626405A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115114962A (en) * | 2022-07-19 | 2022-09-27 | 歌尔股份有限公司 | Control method and device based on surface electromyogram signal and wearable device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yuan et al. | Fingerprint liveness detection using an improved CNN with image scale equalization | |
| WO2021143353A1 (en) | Gesture information processing method and apparatus, electronic device, and storage medium | |
| KR102619981B1 (en) | Gesture classification apparatus and method using electromyogram signals | |
| CN110555468A (en) | Electroencephalogram signal identification method and system combining recursion graph and CNN | |
| Chakladar et al. | A multimodal-Siamese Neural Network (mSNN) for person verification using signatures and EEG | |
| López | Deep learning in biometrics: a survey | |
| Saini et al. | Don’t just sign use brain too: A novel multimodal approach for user identification and verification | |
| CN113180692A (en) | Electroencephalogram signal classification and identification method based on feature fusion and attention mechanism | |
| KR20120026232A (en) | Method and apparatus for generating an objected descriptor using extended curvature gabor filter | |
| Behera et al. | Variance-guided attention-based twin deep network for cross-spectral periocular recognition | |
| Huang et al. | Robust multi-feature collective non-negative matrix factorization for ECG biometrics | |
| Jia et al. | Excellent fine-tuning: From specific-subject classification to cross-task classification for motor imagery | |
| CN114626405A (en) | Real-time identity recognition method and device based on electromyographic signals and electronic equipment | |
| Zhang et al. | Artificial intelligence in physiological characteristics recognition for internet of things authentication | |
| Gürkan et al. | Evrişimsel sinir ağı ve QRS imgeleri kullanarak EKG tabanlı biyometrik tanıma yöntemi | |
| Sridhar et al. | A Neural Network Approach for EEG classification in BCI | |
| Choi et al. | Multi-modal biometrics based implicit driver identification system using multi-TF images of ECG and EMG | |
| Zhang et al. | ECGID: a human identification method based on adaptive particle swarm optimization and the bidirectional LSTM model | |
| CN114548262B (en) | Feature level fusion method for multi-mode physiological signals in emotion calculation | |
| CN114343679A (en) | Surface electromyogram signal upper limb action recognition method and system based on transfer learning | |
| Rathod et al. | Wading corvus optimization based text generation using deep CNN and BiLSTM classifiers | |
| Kumar et al. | A systematic review of palm and dorsal hand vein recognition techniques | |
| Islam | Exploring convolutional neural networks for facial expression recognition: A comprehensive survey | |
| Kumar et al. | A Critical Review of Finger Vein Recognition Techniques for Human Identification | |
| Lu et al. | LGL-BCI: A Lightweight Geometric Learning Framework for Motor Imagery-Based Brain-Computer Interfaces |
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 |