CN110751060A - Portable motion mode real-time identification system based on multi-source signals - Google Patents
Portable motion mode real-time identification system based on multi-source signals Download PDFInfo
- Publication number
- CN110751060A CN110751060A CN201910932479.5A CN201910932479A CN110751060A CN 110751060 A CN110751060 A CN 110751060A CN 201910932479 A CN201910932479 A CN 201910932479A CN 110751060 A CN110751060 A CN 110751060A
- Authority
- CN
- China
- Prior art keywords
- signal acquisition
- signals
- real
- time
- signal
- 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.)
- Granted
Links
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
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/40—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture
-
- 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
A portable motion mode real-time identification system based on multi-source signals comprises a portable multi-source signal acquisition device and an upper computer which exchanges data with the portable multi-source signal acquisition device through Wi-Fi; the upper computer sets the working mode of the portable multi-source signal acquisition device and controls the portable multi-source signal acquisition device to be started and closed; preprocessing collected multi-source signals, detecting a motion starting point, extracting real-time features, training a classification model and identifying a motion mode in real time; filing and managing the information and the experimental data of the testee; the method can develop an individualized experimental paradigm for a user, complete paradigm design and generation, and provide a control method for synchronization of experimental paradigm presentation and signal acquisition; the invention can realize real-time transmission, processing, display, storage of multi-source signals and real-time automatic identification of motion modes, and also provides a way for designing and generating experimental paradigm; an ideal signal acquisition front end and an implementation platform are provided for real-time identification research and application of the motion mode.
Description
Technical Field
The invention relates to the technical field of rehabilitation aids and human-computer interfaces, in particular to a portable motion mode real-time identification system based on multi-source signals.
Background
The motion pattern recognition technology can realize recognition and prediction of the activity patterns of the whole or partial limbs of the human body in different scenes, and can be widely applied to many fields. Such as: in the field of artificial limb rehabilitation, the control of artificial limbs of the disabled can be realized by identifying the motion mode; in the field of human-computer interaction, the human motion intention is identified, so that the motion intention is converted into a control instruction to be controlled and interacted with a machine.
According to the difference of the types of the adopted signals, the human motion pattern recognition can be roughly divided into two types, namely, the motion pattern recognition is carried out by acquiring images or video information of human activities through an image-based method; the other method is a sensor-based method, and various signals generated during human body movement are collected through a sensor attached to a human body, and then the signals are utilized to identify the movement mode. The former limits the range of motion of a person, and the recognition effect depends largely on environmental conditions (lighting, shading, etc.). The latter has no strict limitation on the range of human activities, and is convenient to wear and more widely used.
Surface electromyography (sEMG) is a bioelectric signal generated during the activity of the neuromuscular system, is the synthesis of action potentials of motor units in a plurality of muscle fibers on the time dimension and the space dimension of the surface of a human body, and reflects the activity of the neuromuscular system to a certain extent. Surface electromyogram signals have been widely used as research means in the research of motor dysfunction diagnosis, motor medicine, rehabilitation medicine, and the like. With the development of signal processing technology and artificial intelligence technology, the importance of surface electromyographic signals in motion pattern recognition is more and more prominent.
In addition, by measuring an acceleration signal and a gyroscope signal generated by the limb in the movement process, on one hand, the spatial displacement trend of the limb generated along with the time can be acquired, and on the other hand, the inclination angle information of the limb relative to the gravity acceleration can be acquired, so that the movement track and posture change of the limb in the space can be acquired. The limb movement information has important value in the research fields of limb function evaluation, human-computer interfaces and the like.
The surface electromyogram signal is convenient to collect, accurate and noninvasive, fine action pattern recognition can be carried out, and the electromyogram signal generated by the muscle of the disabled limb can still be used for recognizing the movement pattern for the patient with the disabled limb. Acceleration and gyroscope signals have good identification performance on a motion mode with a larger scale, and a good identification effect can be realized only by few sensors. Therefore, the system development which can realize myoelectricity, acceleration, gyroscope signal synchronous acquisition and motion mode real-time identification has important practical value and wide application prospect.
However, the existing related devices and systems can only complete the electromyographic signal acquisition alone. If acceleration information needs to be acquired, this is often done by adding additional signal acquisition equipment. The myoelectricity and acceleration signals acquired by the independent acquisition module are difficult to realize synchronization among different signals, and the integration and portability of the system are reduced.
Secondly, in the field of motion pattern recognition research based on multi-source signals, most of the research is still in an off-line analysis stage, namely, the recognition algorithm can not be guaranteed to meet the real-time requirement.
Thirdly, the existing motion mode real-time identification system software does not have the design and presentation functions of a specific experimental paradigm. When a classification model is trained, signal acquisition of a specific training paradigm is often required, and experimental stimulation or experimental prompt is generally presented through other external software or tools, so that a training task stimulation sequence is difficult to realize synchronous acquisition with signals, and difficulty is caused in positioning and selecting signal segments in subsequent electromyographic signal analysis. In conclusion, the existing motion pattern recognition system has the defects of lack of multi-source signal synchronous acquisition, poor portability, incapability of running motion pattern recognition in real time, no experimental paradigm design and no presentation function and the like, so that the system is greatly limited in practical use and causes much inconvenience.
Disclosure of Invention
In order to overcome the defects and shortcomings of the existing system, the invention aims to provide a portable motion mode real-time identification system based on multi-source signals, which can realize the synchronous acquisition of electromyographic signals and motion information on hardware, has the characteristics of strong wireless transmission capability, high system hardware integration level and strong system software function, and has the advantages of simple structure, small volume, low power consumption, high precision and convenience in wearing; on the aspect of software, real-time transmission, processing, display, storage and motion mode identification of multi-source signals can be realized, and a way capable of designing, generating and presenting an experimental paradigm is also provided; the system provides an ideal acquisition front end and an implementation platform for real-time identification of the motion mode.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a portable motion mode real-time identification system based on multi-source signals comprises a portable multi-source signal acquisition device and an upper computer which exchanges data with the portable multi-source signal acquisition device through Wi-Fi;
the portable multi-source signal acquisition device comprises a myoelectric signal acquisition front end, an inertia measurement module, a wireless MCU module and a power management module;
the electromyographic signal acquisition front end is connected with an electromyographic electrode attached to the body surface of muscle, and comprises an ADS1299 conversion chip and an electromyographic signal preprocessing circuit, wherein the ADS1299 conversion chip is used for preprocessing, amplifying and sampling the electromyographic signal, and finally the signal is sent to a wireless MCU module;
the inertial measurement module is used for collecting acceleration signals and gyroscope signals, and finally the signals are sent to the wireless MCU module and can be synchronized with the collection of the electromyographic signal collection front end;
the wireless MCU module is used for acquiring myoelectric signals, acceleration signals and gyroscope signals from the myoelectric acquisition front end and the inertia measurement module, and packaging the multi-source signals and sending the multi-source signals to the upper computer through Wi-Fi;
the power management module is connected with the electromyographic signal acquisition front end, the inertia measurement module and the wireless MCU module and provides voltage required by normal work for the electromyographic signal acquisition front end, the inertia measurement module and the wireless MCU module;
the upper computer is provided with motion mode real-time identification system software and provides the following functions: the portable multi-source signal acquisition device is used for setting the working mode of the portable multi-source signal acquisition device before starting acquisition and controlling the start and the stop of the portable multi-source signal acquisition device; preprocessing collected multi-source signals, detecting a motion starting point, extracting real-time features, training a classification model and identifying a motion mode in real time; filing and managing the information and the experimental data of the testee; the method can develop an individualized experimental paradigm for a user, complete paradigm design and generation, and provide a control method for synchronization of experimental paradigm presentation and signal acquisition.
The electromyographic signal acquisition front end in the acquisition device comprises a 24-bit high-precision A/D ADS1299 conversion chip and an electromyographic signal preprocessing circuit, wherein the ADS1299 conversion chip is provided with 8-channel differential input, an 8-channel programmable amplifier, a driving reference circuit and a communication and control interface, and the communication and control interface is connected with a wireless MCU module.
The electromyographic signal preprocessing circuit comprises a first resistor 401, a second resistor 402, a first capacitor 403, a second capacitor 404 and a third capacitor 405; one end of the first resistor 401 is connected with an output end of a positive electrode, the other end of the first resistor 401 is connected with one ends of the first capacitor 403 and the second capacitor 404, one end of the second resistor 402 is connected with an output end of a negative electrode, the other end of the first resistor 402 is respectively connected with the other end of the first capacitor 403 and one end of the third capacitor 405, the other end of the second capacitor 404 and the other end of the third capacitor 405 are connected to an AGND end, one end of the second capacitor 404 is connected with a differential positive input end of one channel of the ADS1299 conversion chip, and the other end of the third capacitor 405 is connected with a differential negative input end.
The motion mode is identified in real time, and the steps are as follows:
(1) carrying out band-pass filtering on the collected electromyographic signals, and carrying out mean value filtering on the collected acceleration and gyroscope signals;
(2) finishing the detection of the motion starting point based on the filtered electromyographic signals;
(3) according to the detected motion starting point, intercepting a myoelectric signal, an acceleration signal and a gyroscope signal with a certain length behind the starting point as an active segment signal;
(4) extracting the characteristics of the signals of the active segment to form a characteristic vector;
(5) normalizing the maximum and minimum values of the feature vectors;
(6) the steps 1-4 are repeated to collect different types of motion modes, feature vector sample sets of the different types of motion modes are collected, and then samples are used for training a classifier algorithm;
(7) and utilizing the trained classifier algorithm to identify the multi-source signals generated by the subsequent motion mode in real time.
The features extracted in the step (4) comprise:
features extracted from the electromyographic signals:
mean absolute value of amplitude (MAV); wilson Amplitude (WAMP); a Wavelength (WL); a 2 nd order moment (SM 2); linear autoregressive coefficients of order 6.
Features extracted from the acceleration signal and the gyroscope signal:
mean (Mean); standard deviation (STD); an excess mean ratio (MCR); a maximum value (MAX); minimum value (MIN); a maximum and minimum index; quartering distances (IQR); signal amplitude area (SMA).
The step (6) is specifically as follows: the classifier algorithm used is a support vector machine, wherein the algorithm adopts a support vector machine using a nonlinear kernel function, the kernel function is RBF, and two parameters of the classifier are as follows: a kernel function parameter gamma and a penalty factor C, and determining the optimal value of the kernel function parameter gamma and the penalty factor C by a grid search method in a training stage, wherein the method specifically comprises the following steps:
(1) taking values of gamma and C, respectively generating from 2-16To 216Equally dividing the two into values according to indexes;
(2) combining each value of gamma and C, and performing cross validation on a classifier corresponding to each combination by using training data to obtain the cross validation accuracy under the combination;
(3) and selecting a combination with the highest cross validation accuracy, wherein the parameters gamma and C corresponding to the combination are the optimal values of the parameters.
The method can develop an individualized experimental paradigm for a user, complete paradigm design and generation, and provide a control method for synchronization of experimental paradigm presentation and signal acquisition, and comprises the following implementation steps:
(1) collecting stimulation materials needed by the experiment, wherein the types of the stimulation materials comprise pictures, music and characters;
(2) setting the appearance sequence, the stimulation interval, the appearance duration and the appearance frequency of each stimulation material according to an experimental paradigm;
(3) starting the signal acquisition and simultaneously starting the experimental paradigm timer to work;
(4) presenting a stimulus when the experimental paradigm timer reaches a starting time point for the stimulus;
(5) when the experimental paradigm timer reaches the end time point of a certain stimulus, closing the presentation of the stimulus;
(6) when the experimental paradigm timer reaches the end time point of the last stimulus, closing the stimulus and simultaneously ending the signal acquisition;
(7) the signal and corresponding stimulation sequence collected this time are saved.
Compared with the prior art, the invention has the following remarkable advantages:
(1) the portable multi-source signal acquisition device supports independent or synchronous acquisition of electromyographic signals, acceleration signals and gyroscope signals, and supports wider application.
(2) The portable multi-source signal acquisition device uses a high-integration-level analog front-end chip and a high-performance MCU with a Wi-Fi radio frequency module, the myoelectric signal acquisition front end, the inertia measurement module and the analog voltage conversion circuit are separated from the wireless MCU and the digital voltage conversion circuit to be designed into two printed circuit boards, the space utilization rate is further improved, the size of the whole device is not more than 60mm 50mm 30mm, and the signal acquisition device is high in integration level, small in size, convenient to carry, low in power consumption and high in accuracy.
(3) The experimental paradigm generation and experimental control unit is embedded in the motion mode real-time identification software, so that the problems of delay caused by stimulation presented by external software, difficulty in synchronization of signals and stimulation and the like are solved, and a convenient research platform is provided for the research of human-computer interfaces, rehabilitation robots and the like.
(4) The motion mode real-time recognition algorithm based on multi-source signal fusion with real-time guarantee is provided, and 96% of four hand actions can be correctly classified under the condition that the time delay is lower than 200 ms.
Drawings
FIG. 1 is a schematic diagram of the overall design of the system of the present invention.
Fig. 2 is a schematic diagram of a hardware structure of the portable multi-source signal acquisition device of the present invention.
Fig. 3 is a software architecture diagram of the real-time motion pattern recognition software according to the present invention.
Fig. 4 is a circuit diagram of an electromyographic signal preprocessing circuit in the electromyographic signal acquisition front end of the present invention.
FIG. 5 is a flowchart of a method for generating an experimental paradigm and an experimental control module in the real-time motion pattern recognition software according to the present invention.
FIG. 6 is a flow chart of a motion pattern real-time recognition method based on multi-source signals of the motion pattern real-time recognition module of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings.
Referring to fig. 1 and 2, a portable motion mode real-time identification system based on multi-source signals comprises a portable multi-source signal acquisition device and an upper computer which exchanges data with the portable multi-source signal acquisition device through Wi-Fi;
the portable multi-source signal acquisition device comprises a myoelectric signal acquisition front end, an inertia measurement module, a wireless MCU module and a power management module;
the electromyographic signal acquisition front end is connected with an electromyographic electrode attached to the body surface of muscle, and comprises an ADS1299 conversion chip and an electromyographic signal preprocessing circuit, wherein the ADS1299 conversion chip is used for preprocessing, amplifying and sampling the electromyographic signal, and finally the signal is sent to a wireless MCU module;
the electromyographic signal acquisition front end comprises an ADS1299 conversion chip with 24-bit high-precision A/D and an electromyographic signal preprocessing circuit. The ADS1299 conversion chip is provided with 8-channel differential input, an 8-channel programmable amplifier, a driving reference circuit and a communication and control interface, wherein the communication and control interface is connected with the wireless MCU module. Compared with an A/D conversion circuit built by common discrete devices, the ADS1299 conversion chip has the advantages of higher acquisition precision, better filtering effect, more comprehensive functions and smaller volume, thereby achieving the effects of portability and accuracy.
The electromyogram signal preprocessing circuit (as shown in fig. 4) includes a first resistor 401, a second resistor 402, a first capacitor 403, a second capacitor 404, and a third capacitor 405. One end of the first resistor 401 is connected to an output end of the positive electrode, the other end of the first resistor 401 is connected to one end of the first capacitor 403 and one end of the second capacitor 404, one end of the second resistor 402 is connected to an output end of the negative electrode, the other end of the first resistor 402 is connected to the other end of the first capacitor 403 and one end of the third capacitor 405, the other end of the second capacitor 404 and the other end of the third capacitor 405 are connected to the AGND terminal, one end of the second capacitor 404 is connected to a differential positive input terminal of one channel of the ADS1299 conversion chip, and the other end of the third capacitor 405 is connected to a differential negative input terminal. The main functions of the preprocessing circuit are as follows: firstly, before analog-to-digital conversion is carried out on a signal, anti-aliasing filtering is carried out on the signal, and meanwhile, a differential filtering mode is adopted between two differential input ends, so that the problem of reduction of a common mode rejection ratio caused by mismatch of capacitance values can be greatly improved; the second is to significantly reduce possible high frequency EMI interference.
The inertia measurement module is used for collecting acceleration signals and gyroscope signals, A/D conversion is carried out, data are sent to the wireless MCU through an I2C protocol, and finally the signals are sent to the wireless MCU module and can be synchronized with collection of the electromyographic signal collection front end.
The wireless MCU module is used for acquiring myoelectric signals, acceleration signals and gyroscope signals from the myoelectric acquisition front end and the inertia measurement module, and packaging the multi-source signals and sending the multi-source signals to the upper computer through Wi-Fi.
The power management module is connected with the electromyographic signal acquisition front end, the inertia measurement module and the wireless MCU module and provides voltage required by normal work for the three.
The upper computer is used for setting the working mode of the portable multi-source signal acquisition device before the acquisition is started and controlling the start and the stop of the portable multi-source signal acquisition device; preprocessing collected multi-source signals, detecting a motion starting point, extracting real-time features, training a classification model and identifying a motion mode in real time; filing and managing the information and the experimental data of the testee; the method can develop an individualized experimental paradigm for a user, complete paradigm design and generation, and provide a control method for synchronization of experimental paradigm presentation and signal acquisition.
Specifically, the motion mode real-time identification software in the upper computer is designed by adopting the idea of a three-layer architecture, namely an interface interaction layer, a service logic layer and a data exchange layer, wherein the layers are clear in labor division and complete the functions of all modules together; the software architecture diagram is shown in fig. 3, and the following modules are formed:
(1) real-time Wi-Fi communication interface: the system is used for exchanging data with the multi-source signal acquisition device through Wi-Fi;
(2) the acquisition device control module: the system is used for setting the working mode of the multi-source signal acquisition device before the acquisition is started and controlling the acquisition device to be started and closed;
(3) the information management and storage module: structured management is carried out on the information of the subject and the experimental data by utilizing a MySQL database;
(4) the experimental paradigm generation and experimental control module: the method is convenient for a user to develop an individualized experimental paradigm, can complete paradigm design and generation, and provides a synchronous control method for experimental paradigm presentation and signal acquisition;
(5) the motion mode real-time identification module: the system is used for preprocessing the collected multi-source signals, detecting a motion starting point, extracting real-time features, training a classification model and identifying a motion mode in real time;
the module provides a way for a user to design and generate an experimental paradigm and a control method for stimulus sequence presentation and signal acquisition synchronization, and the implementation steps are as follows:
(1) collecting stimulation materials needed by the experiment, wherein the types of the stimulation materials comprise pictures, music and characters;
(2) setting the appearance sequence, the stimulation interval, the appearance duration and the appearance frequency of each stimulation material according to an experimental paradigm;
(3) starting the signal acquisition and simultaneously starting the experimental paradigm timer to work;
(4) presenting a stimulus when the experimental paradigm timer reaches a starting time point for the stimulus;
(5) when the experimental paradigm timer reaches the end time point of a certain stimulus, closing the presentation of the stimulus;
(6) when the experimental paradigm timer reaches the end time point of the last stimulus, closing the stimulus and simultaneously ending the signal acquisition;
(7) the signal and corresponding stimulation sequence collected this time are saved.
The method well solves the problems of delay caused by presentation of stimulation by external software, difficulty in synchronization of signals and stimulation and the like, and provides a convenient research platform for researches on human-machine interfaces, rehabilitation robots and the like.
The motion pattern real-time identification is shown in fig. 6, and the detailed steps are as follows:
(1) carrying out band-pass filtering on the collected electromyographic signals, and carrying out mean value filtering on the collected acceleration and gyroscope signals;
(2) finishing the detection of the motion starting point based on the filtered electromyographic signals;
(3) according to the detected motion starting point, intercepting a myoelectric signal, an acceleration signal and a gyroscope signal with a certain length behind the starting point as an active segment signal;
(4) extracting the characteristics of the signals of the active segment to form a characteristic vector;
(5) the values in the feature vector are normalized, the normalization function being as follows:
in the formula: x is the original value of the feature and z is the normalized value of the feature. max (x) is the maximum value in the feature vectors in the training samples, and min (x) is the minimum value in the feature vectors in the training samples.
(6) The steps 1-4 are repeated to collect different types of motion modes, feature vector sample sets of the different types of motion modes are collected, and then the samples are used for training the model;
(7) and utilizing the trained model to identify the multi-source signals generated by the subsequent motion mode in real time.
In order to ensure that the normal work of real-time gesture recognition is not influenced by real-time feature extraction, a system needs to construct a set of electromyographic signal feature set for real-time detection, and the information for effectively representing the movement intention is extracted under the condition that the algorithm complexity is low and the requirement for motion pattern recognition under a small time delay is met. Therefore, the algorithm selects the following time domain features and frequency domain features as the feature set for real-time identification of the movement intention.
In step (4) of the identification method, the extracted features are as follows,
features extracted from the electromyographic signals:
① mean absolute value of amplitude (MAV):
in the formula: n is the number of points in the sequence; x ═ x1,x2…xN) Is a time domain sequence of the electromyographic signals.
② Wilson Amplitude (WAMP):
in the formula, threshold is a preset threshold, and 40uV is taken in the algorithm;
wavelength (WL):
③ 2 order moment (SM 2);
in the formula: p ═ P1,P2…PN-1) Is a myoelectric signal power spectrum sequence.
Linear autoregressive coefficient of order ④ 6:
in the formula: p is the order of the AR model, and 6 is taken in the algorithm; a ═ a1,…aP) Are the AR model coefficients; w is aiIs a white noise error term.
Features extracted from the acceleration signal and the gyroscope signal:
MEAN (MEAN); standard deviation (STD); an excess mean ratio (MCR); a maximum value (MAX); minimum value (MIN); maximum minimum index (IMIN, IMAX); quartering distances (IQR); signal amplitude area (SMA).
In the step (6) of the motion mode real-time identification method based on the multi-source signals, the used classifier algorithm is a support vector machine, wherein the algorithm adopts the support vector machine using a nonlinear kernel function, the kernel function is RBF, and two parameters of the classifier are as follows: a kernel function parameter γ and a penalty factor C. The penalty factor C can represent the weight of penalty for the misclassification case. The higher C represents a high penalty for error conditions, so that overfitting is easy to occur, and under-fitting is easy to occur otherwise. Too large or too small a value of C tends to degrade the performance of the classification model. γ is an internal parameter of the RBF function that determines the width of the RBF, the size of which affects the mapping of the data to the distribution behind the feature space. The basic rule is that the larger the gamma is, the fewer the support vectors are, the higher the accuracy of the model to the training set sample, but the lower the accuracy to the test sample, namely overfitting occurs; the smaller the value of gamma, the more support vectors, the higher accuracy cannot be obtained in the training phase, and the accuracy in the recognition phase is affected. In the training stage, the optimal value is determined by a grid search method, and the specific method is as follows:
(1) taking values of gamma and C, respectively generating from 2-16To 216Equally dividing the two into values according to indexes;
(2) combining each value of gamma and C, and performing cross validation on a classifier corresponding to each combination by using training data to obtain the cross validation accuracy under the combination;
(3) and selecting a combination with the highest cross validation accuracy, wherein the parameters gamma and C corresponding to the combination are the optimal values of the parameters.
It should be noted that the above-mentioned embodiments of the present invention are only used for illustrating or explaining the principle of the present invention, and do not constitute a limitation to the present invention. Therefore, any modification, equivalent replacement, improvement and the like made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. Further, it is intended that the appended claims cover all such variations and modifications as fall within the scope and boundaries of the appended claims or the equivalents of such scope and boundaries.
Claims (7)
1. A portable motion mode real-time identification system based on multi-source signals is characterized by comprising a portable multi-source signal acquisition device and an upper computer which exchanges data with the portable multi-source signal acquisition device through Wi-Fi;
the portable multi-source signal acquisition device comprises a myoelectric signal acquisition front end, an inertia measurement module, a wireless MCU module and a power management module;
the electromyographic signal acquisition front end is connected with an electromyographic electrode attached to the body surface of muscle, and comprises an ADS1299 conversion chip and an electromyographic signal preprocessing circuit, wherein the ADS1299 conversion chip is used for preprocessing, amplifying and sampling the electromyographic signal, and finally the signal is sent to a wireless MCU module;
the inertial measurement module is used for collecting acceleration signals and gyroscope signals, and finally the signals are sent to the wireless MCU module and can be synchronized with the collection of the electromyographic signal collection front end;
the wireless MCU module is used for acquiring myoelectric signals, acceleration signals and gyroscope signals from the myoelectric acquisition front end and the inertia measurement module, and packaging the multi-source signals and sending the multi-source signals to the upper computer through Wi-Fi;
the power management module is connected with the electromyographic signal acquisition front end, the inertia measurement module and the wireless MCU module and provides voltage required by normal work for the electromyographic signal acquisition front end, the inertia measurement module and the wireless MCU module;
the upper computer is provided with motion mode real-time identification system software and provides the following functions: the portable multi-source signal acquisition device is used for setting the working mode of the portable multi-source signal acquisition device before starting acquisition and controlling the start and the stop of the portable multi-source signal acquisition device; preprocessing collected multi-source signals, detecting a motion starting point, extracting real-time features, training a classification model and identifying a motion mode in real time; filing and managing the information and the experimental data of the testee; the method can develop an individualized experimental paradigm for a user, complete paradigm design and generation, and provide a control method for synchronization of experimental paradigm presentation and signal acquisition.
2. The portable real-time motion pattern recognition system based on multi-source signals according to claim 1, wherein the electromyographic signal acquisition front end in the acquisition device comprises a 24-bit high-precision A/D ADS1299 conversion chip and an electromyographic signal preprocessing circuit, the ADS1299 conversion chip is provided with 8-channel differential input, an 8-channel programmable amplifier, a driving reference circuit and a communication and control interface, and the communication and control interface is connected with the wireless MCU module.
3. The portable real-time motion pattern recognition system based on multi-source signals of claim 2, wherein the electromyographic signal preprocessing circuit comprises a first resistor 401, a second resistor 402, a first capacitor 403, a second capacitor 404 and a third capacitor 405; one end of the first resistor 401 is connected with an output end of a positive electrode, the other end of the first resistor 401 is connected with one ends of the first capacitor 403 and the second capacitor 404, one end of the second resistor 402 is connected with an output end of a negative electrode, the other end of the first resistor 402 is respectively connected with the other end of the first capacitor 403 and one end of the third capacitor 405, the other end of the second capacitor 404 and the other end of the third capacitor 405 are connected to an AGND end, one end of the second capacitor 404 is connected with a differential positive input end of one channel of the ADS1299 conversion chip, and the other end of the third capacitor 405 is connected with a differential negative input end.
4. The portable real-time multi-source signal-based motion pattern recognition system of claim 1, wherein the real-time motion pattern recognition comprises the following steps:
(1) carrying out band-pass filtering on the collected electromyographic signals, and carrying out mean value filtering on the collected acceleration and gyroscope signals;
(2) finishing the detection of the motion starting point based on the filtered electromyographic signals;
(3) according to the detected motion starting point, intercepting a myoelectric signal, an acceleration signal and a gyroscope signal with a certain length behind the starting point as an active segment signal;
(4) extracting the characteristics of the signals of the active segment to form a characteristic vector;
(5) normalizing the maximum and minimum values of the feature vectors;
(6) the steps 1-4 are repeated to collect different types of motion modes, feature vector sample sets of the different types of motion modes are collected, and then samples are used for training a classifier algorithm;
(7) and utilizing the trained classifier algorithm to identify the multi-source signals generated by the subsequent motion mode in real time.
5. The portable real-time multi-source signal-based motion pattern recognition system of claim 4, wherein the features extracted in the step (4) comprise:
features extracted from the electromyographic signals:
mean absolute value of amplitude (MAV); wilson Amplitude (WAMP); a Wavelength (WL); a 2 nd order moment (SM 2); linear autoregressive coefficients of order 6.
Features extracted from the acceleration signal and the gyroscope signal:
mean (Mean); standard deviation (STD); an excess mean ratio (MCR); a maximum value (MAX); minimum value (MIN); a maximum and minimum index; quartering distances (IQR); signal amplitude area (SMA).
6. The portable real-time motion pattern recognition system based on multi-source signals according to claim 4, wherein the step (6) is specifically as follows: the classifier algorithm used is a support vector machine, wherein the algorithm adopts a support vector machine using a nonlinear kernel function, the kernel function is RBF, and two parameters of the classifier are as follows: a kernel function parameter gamma and a penalty factor C, and determining the optimal value of the kernel function parameter gamma and the penalty factor C by a grid search method in a training stage, wherein the method specifically comprises the following steps:
(1) taking values of gamma and C, respectively generating from 2-16To 216Equally dividing the two into values according to indexes;
(2) combining each value of gamma and C, and performing cross validation on a classifier corresponding to each combination by using training data to obtain the cross validation accuracy under the combination;
(3) and selecting a combination with the highest cross validation accuracy, wherein the parameters gamma and C corresponding to the combination are the optimal values of the parameters.
7. The system of claim 1, wherein the system is capable of developing an individualized experimental paradigm for a user, completing paradigm design and generation, and providing a control method for synchronization of experimental paradigm presentation and signal acquisition, and comprises the following steps:
(1) collecting stimulation materials needed by the experiment, wherein the types of the stimulation materials comprise pictures, music and characters;
(2) setting the appearance sequence, the stimulation interval, the appearance duration and the appearance frequency of each stimulation material according to an experimental paradigm;
(3) starting the signal acquisition and simultaneously starting the experimental paradigm timer to work;
(4) presenting a stimulus when the experimental paradigm timer reaches a starting time point for the stimulus;
(5) when the experimental paradigm timer reaches the end time point of a certain stimulus, closing the presentation of the stimulus;
(6) when the experimental paradigm timer reaches the end time point of the last stimulus, closing the stimulus and simultaneously ending the signal acquisition;
(7) the signal and corresponding stimulation sequence collected this time are saved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910932479.5A CN110751060B (en) | 2019-09-29 | 2019-09-29 | Portable motion mode real-time identification system based on multi-source signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910932479.5A CN110751060B (en) | 2019-09-29 | 2019-09-29 | Portable motion mode real-time identification system based on multi-source signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110751060A true CN110751060A (en) | 2020-02-04 |
| CN110751060B CN110751060B (en) | 2021-02-19 |
Family
ID=69277406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910932479.5A Active CN110751060B (en) | 2019-09-29 | 2019-09-29 | Portable motion mode real-time identification system based on multi-source signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110751060B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112075938A (en) * | 2020-06-03 | 2020-12-15 | 深圳市润谊泰益科技有限责任公司 | Distributed data acquisition system, method and storage medium |
| CN117528131A (en) * | 2024-01-05 | 2024-02-06 | 青岛美迪康数字工程有限公司 | AI integrated display system and method for medical image |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105496418A (en) * | 2016-01-08 | 2016-04-20 | 中国科学技术大学 | Arm-belt-type wearable system for evaluating upper limb movement function |
| CN105786189A (en) * | 2016-04-28 | 2016-07-20 | 深圳大学 | Finger independent action recognition method and system based on MMG signal |
| CN106237604A (en) * | 2016-08-31 | 2016-12-21 | 歌尔股份有限公司 | Wearable device and the method utilizing its monitoring kinestate |
| CN107169302A (en) * | 2017-06-12 | 2017-09-15 | 北京工业大学 | A kind of stroke patient information gathering and management system |
| CN107562191A (en) * | 2017-08-03 | 2018-01-09 | 天津大学 | The online brain-machine interface method of fine Imaginary Movement based on composite character |
| CN107995855A (en) * | 2015-03-26 | 2018-05-04 | 拜欧米特制造有限责任公司 | For planning and performing the method and system of joint replacement flow using motion capture data |
| CN108968952A (en) * | 2018-05-30 | 2018-12-11 | 燕山大学 | A kind of brain myoelectricity and Inertia information synchronous acquisition device |
| CN108992066A (en) * | 2018-08-15 | 2018-12-14 | 东北大学 | Portable lower limb behavior pattern real-time identifying system and method based on electromyography signal |
| KR20190048351A (en) * | 2017-10-31 | 2019-05-09 | 권혁준 | Smart health care system for checking exercise |
| CN109753868A (en) * | 2018-11-14 | 2019-05-14 | 深圳卡路里科技有限公司 | Appraisal procedure and device, the Intelligent bracelet of athletic performance |
| CN109992113A (en) * | 2019-04-09 | 2019-07-09 | 燕山大学 | A kind of MI-BCI system and its control method induced based on more scenes |
-
2019
- 2019-09-29 CN CN201910932479.5A patent/CN110751060B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107995855A (en) * | 2015-03-26 | 2018-05-04 | 拜欧米特制造有限责任公司 | For planning and performing the method and system of joint replacement flow using motion capture data |
| CN105496418A (en) * | 2016-01-08 | 2016-04-20 | 中国科学技术大学 | Arm-belt-type wearable system for evaluating upper limb movement function |
| CN105786189A (en) * | 2016-04-28 | 2016-07-20 | 深圳大学 | Finger independent action recognition method and system based on MMG signal |
| CN106237604A (en) * | 2016-08-31 | 2016-12-21 | 歌尔股份有限公司 | Wearable device and the method utilizing its monitoring kinestate |
| CN107169302A (en) * | 2017-06-12 | 2017-09-15 | 北京工业大学 | A kind of stroke patient information gathering and management system |
| CN107562191A (en) * | 2017-08-03 | 2018-01-09 | 天津大学 | The online brain-machine interface method of fine Imaginary Movement based on composite character |
| KR20190048351A (en) * | 2017-10-31 | 2019-05-09 | 권혁준 | Smart health care system for checking exercise |
| CN108968952A (en) * | 2018-05-30 | 2018-12-11 | 燕山大学 | A kind of brain myoelectricity and Inertia information synchronous acquisition device |
| CN108992066A (en) * | 2018-08-15 | 2018-12-14 | 东北大学 | Portable lower limb behavior pattern real-time identifying system and method based on electromyography signal |
| CN109753868A (en) * | 2018-11-14 | 2019-05-14 | 深圳卡路里科技有限公司 | Appraisal procedure and device, the Intelligent bracelet of athletic performance |
| CN109992113A (en) * | 2019-04-09 | 2019-07-09 | 燕山大学 | A kind of MI-BCI system and its control method induced based on more scenes |
Non-Patent Citations (3)
| Title |
|---|
| TEXAS: "《Texas Instruments》", 31 January 2017, 《TEXAS INSTRUMENTS》 * |
| 刘磊: "基于多源信息的步态识别算法研究", 《中国博士学位论文全文数据库（信息科技辑）》 * |
| 杨勇: "脑电-肌电-惯性信息同步采集系统设计", 《中国优秀硕士学位论文全文数据库（基础科学辑）》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112075938A (en) * | 2020-06-03 | 2020-12-15 | 深圳市润谊泰益科技有限责任公司 | Distributed data acquisition system, method and storage medium |
| CN112075938B (en) * | 2020-06-03 | 2024-01-05 | 深圳市润谊泰益科技有限责任公司 | Distributed data acquisition system, method and storage medium |
| CN117528131A (en) * | 2024-01-05 | 2024-02-06 | 青岛美迪康数字工程有限公司 | AI integrated display system and method for medical image |
| CN117528131B (en) * | 2024-01-05 | 2024-04-05 | 青岛美迪康数字工程有限公司 | AI integrated display system and method for medical image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110751060B (en) | 2021-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108983973B (en) | Control method of humanoid smart myoelectric artificial hand based on gesture recognition | |
| CN102499797B (en) | Artificial limb control method and system | |
| CN105769173A (en) | Electrocardiogram monitoring system with electrocardiosignal denoising function | |
| CN202288542U (en) | Artificial limb control device | |
| Jaramillo et al. | Real-time hand gesture recognition with EMG using machine learning | |
| CN110555468A (en) | Electroencephalogram signal identification method and system combining recursion graph and CNN | |
| CN110058691A (en) | Based on Embedded wearable wireless dry electrode brain wave acquisition processing system and method | |
| CN111598451B (en) | Control work efficiency analysis method, device and system based on task execution capacity | |
| CN110751060B (en) | Portable motion mode real-time identification system based on multi-source signals | |
| CN111553307A (en) | Gesture recognition system fusing bioelectrical impedance information and myoelectric information | |
| CN108958482B (en) | Similarity action recognition device and method based on convolutional neural network | |
| CN101776981A (en) | Method for controlling mouse by jointing brain electricity and myoelectricity | |
| CN107212883B (en) | A kind of mechanical arm writing device and control method based on brain electric control | |
| Shin et al. | Korean sign language recognition using EMG and IMU sensors based on group-dependent NN models | |
| CN106569606A (en) | Smart home infrared control system and smart home infrared control method based on natural gesture identification | |
| CN109765996A (en) | Insensitive gesture detection system and method are deviated to wearing position based on FMG armband | |
| CN107822629A (en) | The detection method of extremity surface myoelectricity axle | |
| Chen et al. | Surface-electromyography-based gesture recognition using a multistream fusion strategy | |
| Guo et al. | Towards zero training for myoelectric control based on a wearable wireless sEMG armband | |
| CN113871028A (en) | Interactive rehabilitation system based on myoelectric intelligent wearing | |
| CN114707562A (en) | Electromyographic signal sampling frequency control method and device and storage medium | |
| CN110464348A (en) | The continuous amount of exercise recognition methods of joint of lower extremity and system based on electromyography signal | |
| CN116910464A (en) | Myoelectric signal prosthetic hand control system and method | |
| CN205083477U (en) | Real time control device based on body surface flesh | |
| Fu et al. | Ping pong motion recognition based on smart watch |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |