CN113171214B - Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method - Google Patents
Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method Download PDFInfo
- Publication number
- CN113171214B CN113171214B CN202110584128.7A CN202110584128A CN113171214B CN 113171214 B CN113171214 B CN 113171214B CN 202110584128 A CN202110584128 A CN 202110584128A CN 113171214 B CN113171214 B CN 113171214B
- Authority
- CN
- China
- Prior art keywords
- motor
- joint
- finger
- prosthetic hand
- feedback
- 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.)
- Active
Links
- 230000003183 myoelectrical effect Effects 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000033001 locomotion Effects 0.000 claims abstract description 74
- 230000009471 action Effects 0.000 claims abstract description 35
- 230000003993 interaction Effects 0.000 claims abstract description 26
- 238000001228 spectrum Methods 0.000 claims abstract description 22
- 238000012216 screening Methods 0.000 claims abstract description 14
- 210000003811 finger Anatomy 0.000 claims description 50
- 210000001503 joint Anatomy 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 19
- 230000005540 biological transmission Effects 0.000 claims description 18
- 210000000811 metacarpophalangeal joint Anatomy 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 13
- 210000003813 thumb Anatomy 0.000 claims description 13
- 210000001145 finger joint Anatomy 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000013145 classification model Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 210000004932 little finger Anatomy 0.000 claims description 5
- 208000031361 Hiccup Diseases 0.000 claims description 4
- 230000008030 elimination Effects 0.000 claims description 4
- 238000003379 elimination reaction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000003064 k means clustering Methods 0.000 claims description 3
- 238000004804 winding Methods 0.000 claims description 3
- 238000003780 insertion Methods 0.000 claims 3
- 230000037431 insertion Effects 0.000 claims 3
- 230000004913 activation Effects 0.000 description 14
- 210000003414 extremity Anatomy 0.000 description 12
- 230000006870 function Effects 0.000 description 10
- 230000036544 posture Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 230000009466 transformation Effects 0.000 description 7
- 238000005070 sampling Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 210000003205 muscle Anatomy 0.000 description 4
- 230000003595 spectral effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000001276 controlling effect Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 239000011664 nicotinic acid Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 238000002266 amputation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000003592 biomimetic effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 210000005224 forefinger Anatomy 0.000 description 1
- 210000004247 hand Anatomy 0.000 description 1
- 210000002478 hand joint Anatomy 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
- A61F2/00—Filters implantable into blood vessels; Prostheses, i.e. artificial substitutes or replacements for parts of the body; Appliances for connecting them with the body; Devices providing patency to, or preventing collapsing of, tubular structures of the body, e.g. stents
- A61F2/50—Prostheses not implantable in the body
- A61F2/54—Artificial arms or hands or parts thereof
- A61F2/58—Elbows; Wrists ; Other joints; Hands
- A61F2/583—Hands; Wrist joints
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
- A61F2/00—Filters implantable into blood vessels; Prostheses, i.e. artificial substitutes or replacements for parts of the body; Appliances for connecting them with the body; Devices providing patency to, or preventing collapsing of, tubular structures of the body, e.g. stents
- A61F2/50—Prostheses not implantable in the body
- A61F2/54—Artificial arms or hands or parts thereof
- A61F2/58—Elbows; Wrists ; Other joints; Hands
- A61F2/583—Hands; Wrist joints
- A61F2/586—Fingers
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
- A61F2/00—Filters implantable into blood vessels; Prostheses, i.e. artificial substitutes or replacements for parts of the body; Appliances for connecting them with the body; Devices providing patency to, or preventing collapsing of, tubular structures of the body, e.g. stents
- A61F2/50—Prostheses not implantable in the body
- A61F2/68—Operating or control means
- A61F2/70—Operating or control means electrical
- A61F2/72—Bioelectric control, e.g. myoelectric
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
- A61F2/00—Filters implantable into blood vessels; Prostheses, i.e. artificial substitutes or replacements for parts of the body; Appliances for connecting them with the body; Devices providing patency to, or preventing collapsing of, tubular structures of the body, e.g. stents
- A61F2/50—Prostheses not implantable in the body
- A61F2/68—Operating or control means
- A61F2002/6827—Feedback system for providing user sensation, e.g. by force, contact or position
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Transplantation (AREA)
- General Physics & Mathematics (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Vascular Medicine (AREA)
- General Engineering & Computer Science (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Cardiology (AREA)
- Mathematical Physics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Prostheses (AREA)
Abstract
The invention discloses a multipath feedback myoelectric control prosthetic hand based on a self-adaptive enhanced classifier and a method thereof, wherein a prosthetic hand body comprises a plurality of independently moving simulated fingers arranged on a palm platform, and the simulated fingers are controlled to move by a motor driving module; the signal acquisition module is used for acquiring an electromyographic signal of a prosthetic hand body wearer; the classification module is used for extracting action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, and performing integrated screening on the action recognition features, and then performing electromyographic classification by adopting a self-adaptive enhancement classifier to obtain a motion control instruction; the motor driving module is used for driving the finger-like action according to the motion control instruction; the feedback module is used for controlling the execution mode of the motor after the finger-like action according to the interaction force with the object, the working current of the motor and the joint pose when the finger-like action is performed. Under the condition of ensuring high accuracy, the calculated amount is reduced to the maximum extent, and the speed and accuracy of motion recognition of the prosthetic hand are improved.
Description
Technical Field
The invention relates to the technical field of rehabilitation robots, in particular to a multipath feedback myoelectric control prosthetic hand based on a self-adaptive reinforced classifier and a method thereof.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Wearing a biomimetic prosthetic hand is the most direct and effective method of compensation for hand amputees. The appearance of the bionic artificial hand is basically similar to that of a human hand; the structure simulates the physiological structure of the palm and the five fingers, and the motor is used for replacing muscles to drive the movement of the joints of the fingers; functionally, simple gripping or other action can be achieved.
However, to the inventors' knowledge, existing prosthetic hand (excluding robotic arms) controls have far failed to achieve a level of flexible control. The single prosthetic hand is controlled in single degree of freedom, and the motion mode is only opened and closed, which is far from the true flexible control; many prostheses with multiple degrees of freedom are designed to be a combination of a prosthetic hand and a mechanical arm, or the prosthetic hand is arranged on a huge base, so that a patient can hardly really wear the heavy prosthetic hand to operate, and particularly for patients with amputation at the wrist, the prosthetic hand does not need a complicated mechanical arm, but needs to have a prosthetic hand with smart control and light weight design.
For real-time control, existing classification-based or regression-based algorithms need to choose from accuracy and timeliness. When the signal slides, the larger the window length, the more comprehensive the information is, the higher the accuracy is, but the time delay is larger, and vice versa. At present, no very excellent algorithm can realize extremely high accuracy and timeliness at the same time.
Disclosure of Invention
In order to solve the problems, the invention provides a multipath feedback myoelectric control prosthetic hand and a multipath feedback myoelectric control prosthetic hand method based on a self-adaptive enhancement classifier, which are used for collecting motion intention signals of a prosthetic hand wearer in real time, controlling actions of the prosthetic hand by combining feedback information of interaction force with an object, working current of a motor and joint pose, realizing operation requirements in daily life, realizing flexible control and quick response, and improving effects of user experience.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
in a first aspect, the invention provides a multipath feedback myoelectric control prosthetic hand based on a self-adaptive enhanced classifier, which comprises a prosthetic hand body, a motor driving module, a signal acquisition module, a classification module and a feedback module;
the artificial hand body comprises a plurality of independently moving artificial fingers arranged on a palm platform, and the artificial fingers are controlled to move by a motor driving module;
the signal acquisition module is used for acquiring an electromyographic signal of a prosthetic hand body wearer;
the classification module receives the electromyographic signals and is configured to extract action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, and after the action recognition features are subjected to integrated screening, the electromyographic classification is carried out by adopting a self-adaptive enhancement classifier to obtain a motion control instruction;
the motor driving module is configured to drive the finger-like motion according to the motion control instruction;
the feedback module is configured to control an execution mode of the motor after the finger-like action according to interaction force with an object, working current of the motor and joint pose when the finger-like action is performed.
In a second aspect, the present invention provides a control method based on the prosthetic hand, including:
acquiring an electromyographic signal of a prosthetic hand body wearer;
obtaining a motion control instruction according to the classification of the electromyographic signals, and driving the finger-like motion; extracting action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, carrying out integrated screening on the action recognition features, and then carrying out electromyographic classification by adopting a self-adaptive enhancement classifier;
and acquiring interaction force with an object, working current of a motor and joint pose when the finger is imitated, so as to control an execution mode of the motor after the finger is imitated.
Compared with the prior art, the invention has the beneficial effects that:
the invention realizes the full-driving prosthetic hand with high freedom degree, wherein 14 degrees of freedom (thumb 2 and other four fingers 3) are independently driven by the corresponding encoder motors, thereby providing more flexible and accurate prosthetic hand control for users, solving the design and control problems of the flexible multi-freedom prosthetic hand with multi-path feedback, being different from the multi-freedom of the independent prosthetic hand, and not needing complex bases for supporting.
According to the invention, through short-time Fourier transform and logarithmic spectrum image feature extraction, and after integrated feature selection, the myoelectricity classification is performed by using the self-adaptive enhanced classifier, so that the calculated amount is reduced to the greatest extent under the condition of ensuring high accuracy, and the speed and accuracy of motion recognition of the prosthetic hand are improved.
The invention adopts the myoelectric arm loop as a main signal source for controlling the prosthetic hand, and carries out auxiliary control through the force sensor, the motor encoder and the current feedback path, and the interaction force, the joint pose and the working current of the prosthetic hand are fed back in real time in the motion process, so that the applicability and the safety of the grasping action are improved, and the using effect of the prosthetic hand is improved.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a block diagram of a multiple feedback control system for a prosthetic hand according to embodiment 1 of the present invention;
FIG. 2 is a diagram showing the overall structure of the prosthetic hand according to embodiment 1 of the present invention;
FIG. 3 is a schematic view of the structure of a prosthetic finger joint according to embodiment 1 of the present invention;
fig. 4 is a flow chart of myoelectricity feature extraction and classification provided in embodiment 1 of the present invention;
fig. 5 is a block diagram of an integrated myoelectric feature screening procedure provided in embodiment 1 of the present invention;
FIG. 6 is a flow chart of the feedback control gripping motion provided in embodiment 1 of the present invention;
wherein, 1, a palm platform, 2, a prosthetic hand thumb, 3, a prosthetic hand index finger, 4, a prosthetic hand middle finger, 5, a prosthetic hand ring finger, 6, a prosthetic hand little finger, 7, a direct current motor, 8, a driving gear, 9, a transmission gear, 10, a joint driven shaft, 11, a joint between finger joints with shafts, 12, a groove finger joint, 13, a metacarpophalangeal joint transmission shaft, 14, a finger joint with shafts, 15 and a groove finger joint.
The specific embodiment is as follows:
the invention is further described below with reference to the drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms also are intended to include the plural forms, and furthermore, it is to be understood that the terms "comprises" and "comprising" and any variations thereof are intended to cover non-exclusive inclusions, such as, for example, processes, methods, systems, products or devices that comprise a series of steps or units, are not necessarily limited to those steps or units that are expressly listed, but may include other steps or units that are not expressly listed or inherent to such processes, methods, products or devices.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Example 1
The embodiment provides a multipath feedback myoelectric control prosthetic hand based on a self-adaptive enhancement classifier, which comprises a prosthetic hand body, a motor driving module, a signal acquisition module, a classification module and a feedback module;
the artificial hand body comprises a plurality of independently moving artificial fingers arranged on a palm platform, and the artificial fingers are controlled to move by a motor driving module;
the signal acquisition module is used for acquiring an electromyographic signal of a prosthetic hand body wearer;
the classification module receives the electromyographic signals and is configured to extract action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, and after the action recognition features are subjected to integrated screening, the electromyographic classification is carried out by adopting a self-adaptive enhancement classifier to obtain a motion control instruction;
the motor driving module is configured to drive the finger-like motion according to the motion control instruction;
the feedback module is configured to control an execution mode of the motor after the finger-like action according to interaction force with an object, working current of the motor and joint pose when the finger-like action is performed.
As shown in fig. 1, a block diagram of a prosthetic hand system is shown, in a training stage, myoelectric signals are used as a data set for training a classifier, and in an actual grasping operation stage, the myoelectric signals are used as a control signal source of the classifier; the motor driving module outputs preset motion commands to each encoder motor according to the motion control instructions, and controls the motors to move to the designated positions, namely, the finger-like motion is driven;
meanwhile, the adaptability interaction with the object is also considered when the prosthetic hand is held, and the method is mainly realized through multipath feedback on the prosthetic hand; because the gripping posture is preset, is universal to all interactive objects and is limited by the muscle retention condition of the residual limb of the user, the effective myoelectric activation mode generated by the gripping posture is not enough to be distributed to each object with different shape, quality and texture, and therefore, a certain feedback control is needed to ensure that the gripping is safer and more reliable; the motion gesture of the prosthetic hand is fed back in real time by a winding encoder in a motor, the working current of the motor is realized by a current feedback circuit of a motor driving module, and the interaction between the prosthetic hand and a gripping object is measured and fed back by a force sensor.
The inner side of the prosthetic hand is stuck with a covering layer made of flexible materials to increase the contact area with an object, meanwhile, a piezoelectric sensor is arranged between the covering layer and the prosthetic hand shell, when the contact force of the prosthetic hand is overlarge due to interference between the objects in the movement process, in order to prevent damage to the object to be gripped and the driving motor, the prosthetic hand is enabled to enter a hiccup control mode by the feedback force, a strong current generated by motor stalling is prevented to damage a driving system under the condition of maintaining certain interaction force, and however, the interference caused by the environmental contact within a non-force sensor range or the prosthetic hand can directly send a command to an upper computer through a current feedback module in a motor driving module to stop the action in time; for gripping situations where there is no interaction, such as when the prosthetic hand is making various gestures, where there is no intervention of force feedback, the prosthetic hand is considered to complete the movement task after moving to the specified position.
As shown in fig. 2, the prosthetic hand comprises a palm platform 1, a prosthetic hand thumb 2, a prosthetic hand index finger 3, a prosthetic hand middle finger 4, a prosthetic hand ring finger 5 and a prosthetic hand little finger 6; the 5 direct current motors fixed in the metacarpophalangeal joints are fixed on the inner side of the palm through the fixing structure in the palm platform 1, and the direct current motors of the rest 9 fingertip joints are fixed in the similar structure in the fingertip joints; the flexible material covers the inner side of the prosthetic hand to increase the contact area with the object during the grasping, realize balanced atress, the pressure sensor is in the form of strip, fixes respectively in the below of overburden by fingertip to metacarpophalangeal joint, in order to measure the interaction of the force between hand and the object during the grasping.
The corresponding positions of the five fingers on the palm platform are sequentially provided with a thumb metacarpophalangeal joint and a interphalangeal joint, a forefinger metacarpophalangeal joint and an interphalangeal joint, a middle metacarpophalangeal joint and an interphalangeal joint, and a ring metacarpophalangeal joint and a little metacarpophalangeal joint and an interphalangeal joint, wherein the thumb of the prosthetic hand corresponds to two encoder direct current motors to respectively control the thumb metacarpophalangeal joint and one interphalangeal joint, the other four fingers respectively correspond to three encoder direct current motors to respectively control the metacarpophalangeal joint, and the motions of the near-end interphalangeal joint and the far-end interphalangeal joint are controlled by the mutually independent driving modules.
The palm platform 1, the artificial thumb 2, the direct current motor 7, the driving gear 8, the transmission gear 9, the joint driven shaft 10, the interphalangeal joint with shaft 11, the grooved interphalangeal joint 12, the metacarpophalangeal joint transmission shaft 13, the fingertip joint with shaft 14 and the fingertip joint with groove 15 are shown in fig. 3; the structure of the index finger 3, the middle finger 4, the ring finger 5 and the little finger 6 of the artificial hand are completely consistent, the interphalangeal joint and the fingertip joint of the thumb are completely consistent with the structures of the other four fingers, the artificial finger joint is formed by combining an interphalangeal joint 11 with an axis and a interphalangeal joint 12 with a groove, a direct current motor is arranged in the joint for driving the joint of the next stage to move, and the fingertip joint is formed by combining a fingertip joint 14 with an axis and a fingertip joint 15 with a groove;
the palm platform is designed with a protruding shaft connected with the proximal end at one side of five fingers, the protruding shaft connected with the proximal end of the five fingers is coaxially connected with the tail end of the corresponding position of the artificial finger of the palm platform so as to realize rotary motion, and a transmission gear is nested on the protruding shaft;
the artificial limb units of the index finger of the artificial limb, the middle finger of the artificial limb, the ring finger of the artificial limb and the little finger of the artificial limb all comprise a fingertip joint protruding member, a fingertip joint embedding member, a middle joint protruding member, a middle joint embedding member, a proximal joint protruding member, a proximal joint embedding member, a power gear, a transmission gear and a driving shaft; the fingertip joint is meshed with the transmission gear at the same horizontal plane through a driving shaft which is locked with the fingertip joint, the transmission (bevel) gear is meshed with the power (bevel) gear at an angle of 90 degrees, and torque generated by a motor is used for driving the fingertip joint;
the motor for driving the fingertip joint is positioned in the inner space of the middle joint, the motor for driving the middle joint is positioned in the proximal joint, the motor for driving the proximal joint is positioned in the palm platform, and the fingertip joint is taken as the extreme end of the finger without the motor;
the artificial hand thumb only has a fingertip joint and a interphalangeal joint, and comprises a fingertip joint protruding member, a fingertip joint embedding member, a proximal joint protruding member, a proximal joint embedding member, a power gear, a transmission gear and a driving shaft, and the movement and the transmission mode of the artificial hand thumb are the same as those of the other four fingers.
When the encoder rotating motor moves, the power of the encoder rotating motor is transmitted to the transmission gear nested on the finger joint of the stage through the bevel gear on the encoder rotating motor, and then the transmission gear drives the driving shaft meshed with the encoder rotating motor, the structure meshed with the transmission gear is integrally printed on the driving shaft, and the encoder rotating motor is structurally locked with the finger joint of the next stage, so that the driving shaft can drive the finger joint of the next stage to rotate, bend or stretch.
Taking the index finger of the prosthetic hand as an example in fig. 3, the interphalangeal joint is embedded with the palm platform 1 through the joint driven shaft 10, and meanwhile, the interphalangeal joint and the joint driven shaft are locked, so are the designs of the rest joints; when the direct current motor 7 works, the driving gear 8 drives the transmission gear 9 to move, then torque is transmitted to the joint driven shaft 10, the driven shaft and the interphalangeal joint are locked, and finally the joint driven shaft 10 drives the whole interphalangeal joint to realize the rotary motion of the joint.
As shown in fig. 4, which is a flow chart for myoelectric feature extraction and classification, surface myoelectric (Surface Electromyogrphy, sEMG) signals are widely used in many fields of rehabilitation devices, man-machine interaction, clinic, biomedicine and the like; in the classification of control commands for rehabilitation devices, in particular prosthetic hands, sEMG signals are currently the dominant choice for non-invasive devices, which can record electrical signals reflecting different hand muscle activities under non-invasive conditions;
the embodiment provides a medical expert system based on sEMG signals for classifying grasping action induced myoelectricity of a user so as to enhance action classification accuracy in the use of the prosthetic hand, wherein the medical expert system mainly comprises the following 3 modules: a spectrum-based logarithmic transformation image signal (Logarithmic Spectrogram-Based Graph Signal, LSGS), feature extraction techniques, and an adaptive enhancement k-means (AdaBoost k-means, AB-k-means) classifier.
In the actual use process, preprocessing the collected sEMG data of the user, which is the same as the training data set, then carrying out the same Hamming sliding window processing and short-time Fourier transformation on the data points collected in real time, obtaining a Laplace matrix and characteristic values thereof, and calculating relevant statistical characteristics; because the statistical features with good effects are integrated feature selection in the training stage, the step is not repeated in actual use, and the obtained feature value is directly input into a trained self-adaptive k-means action classifier to obtain a motion instruction.
In the embodiment, short-time Fourier transform is performed on EMG after sliding window, a spectrogram image and a weight matrix thereof are obtained after logarithmic processing is performed on a frequency density spectrum, and a Laplace feature matrix and a feature vector thereof are further obtained, so that various statistical parameter features are calculated, finally, the selected classification features need to be comprehensively subjected to chi-square feature selection, mutual information feature selection and recursive feature elimination, and the most effective data features are obtained through a Fisher formula; and finally, classifying by using a self-adaptive enhancement k-means classifier, classifying and identifying the extracted features in the surface electromyographic signals, wherein the algorithm combines k-means classification optimization results of a plurality of linear weak classifiers into a strong classifier.
Specifically, simple preprocessing is carried out on sEMG signals, and the sEMG signals are filtered and noise-reduced to be used as original signals for classifying motion modes; since physiological signals are usually unstable and nonlinear, the LSGS model is applied to analyze the processed electromyographic signals to obtain the activation patterns of the fixed features, and extract the features; extracting a feature set with higher weight from the feature set; finally, a novel classification model AB-k-means is adopted in the classification stage to map the selected sEMG features into different gripping poses, so that a final classification result is realized; the classifier adopted in the embodiment has fewer learned features, so that the operation amount and the hardware requirement are reduced, and more accurate results can be provided in shorter training time.
For the EMG, which is a high-dimensional non-stationary signal, the embodiment adopts a spectrogram in a logarithmic domain to perform time-frequency conversion on the electromyographic signal so as to facilitate further processing; the specific operation of the time-frequency conversion is to adopt Short-time Fourier transform (STFT), the data needs to be windowed for the Short-time Fourier transform, and a Hamming window containing 256 continuous sampling points is selected as a unit of the Fourier transform; each time the window is shifted by half window length, namely 128 sampling points, to completely cover the electromyographic signals to generate a spectrogram, wherein x represents the original electromyographic signal input, K represents a discrete-time label, z represents a Hamming window function selected by calculation, and K is a constant representing the window length of 256;
thus, x|f| is obtained for each frequency range of each window signal, the spectrum value is a complex number, and the logarithmic spectrum density S (f, j) is obtained by taking the modulus value of the complex value; in the process of calculating the logarithmic spectrum density, the logarithmic spectrum density needs to be marked by a time frame j;
the next step is to plot the log spectral density plot as shown in equation (1):
IS(f,j)＝log(S(f,j))(1)
the log function IS used for nonlinearly scaling the logarithmic spectral density, and meanwhile, the spectral components of the frequency band of the EMG signal can be enhanced, and the two-dimensional logarithmic spectrogram IS (f, j) in frequency and time can be obtained after the log function IS used for transformation.
For single task training, thousands of data sampling points can be recorded in a motion of a few seconds, and after short-time Fourier transform, log spectral density calculation and spectrogram drawing with a time tag are performed on the data sampling points, a pair of IS (f, j) with higher resolution and containing motion recognition characteristic information IS obtained; in order to facilitate model training and later motion recognition, the embodiment reduces the dimension of the spectrogram to 50×50 resolution by means of uniform sampling, and the image is sensitive to different motion frequencies and STFT amplitude values corresponding to the frequencies, and can be used as a reliable criterion for motion classification.
The reduced log spectrum iS denoted iS (f, j), which iS a constant matrix with a resolution of 50 x 50, and the variables f and j no longer represent the frequency versus time scale due to the reduced dimensions of the time variable and the frequency variable, and are therefore only used here as algebraic representations of the abscissa and the ordinate.
Further, the logarithmic spectrum image iS (f, j) iS changed from a 50×50 matrix to a 1×2500-scale row matrix G in a row-by-row arrangement, as shown in the formula (2):
G＝[g 1 ,g 2 ,g 3 ,......g n ],g 50(f-1)+j ＝iS(f,j) (2)
the spectrogram image at the moment has lost the two-dimensional resolution, and the adjacent elements in G are adjacent elements in the original two-dimensional logarithmic spectrogram imageA prime or an end element of one row and a first element of the next row; it is evident that in generating the weighting matrix, the weight calculation of the two cases should be differentiated, so that each element v in the corresponding index matrix v, v of equal scale generating one G n All represent G corresponding thereto n Is a two-dimensional coordinate of (c).
Then, a Gaussian kernel weighting function is applied to calculate the weight between different nodes, as shown in formulas (3) - (4):
wherein the exp function is an exponential function based on natural logarithm e, dist (v i ,v j ) To solve for v in the density spectrum i 、v j Euclidean distance between dist (g i ,g j ) Is the physical distance of the elements in the image signal G, and the parameters ψ, h, ω are empirically set to 5.01,0.2,0.3, respectively.
For the case where the euclidean distance or physical distance of an element is greater than the threshold h, when i=j, p i,j ＝q i,j =1; otherwise p i,j ＝q i,j =0; the mutual influence is defined by the position relation of the pixel points of the two spectrograms, and the final weighting matrix W is obtained by multiplying the mutual influence by the formula (5):
w i,j ＝p i,j ×q i,j (5)
under the condition that a weighting matrix W is known, a derivative diagonal matrix D is obtained through elementary transformation, and a Laplace matrix l is obtained and is the difference matrix of subtracting W from D.
The Laplace matrix l is subjected to eigenvalue decomposition to realize the transformation of the spectrum image, all eigenvalues of the l are solved by means of mathematical software, and are ordered according to the size, so that an N-dimensional eigenvalue matrix X= { χ is obtained 1 ,χ 2 ,χ 3 ,......χ n Characteristic value diagonal matrix Λ of } and NxN, and Laplacian matrix of the graph is a semi-positive definite matrix, so χ 1 =0, and for any χ n And the characteristic value decomposition result L is shown in a formula (6):
L＝XΛX T (6)
wherein X is T Representing a matrix of eigenvalue columns obtained by transposition of the matrix of eigenvalue rows.
As shown in fig. 5, the motion recognition feature obtained in the above step is selected, and various feature value selection methods have advantages and disadvantages, which are specifically embodied in various factors such as selection of data types, accuracy, time, and cost. Therefore, the embodiment adopts an integrated feature selection method, and simultaneously adopts a chi-square feature selection, mutual information and a recursive feature elimination mode to carry out integral feature screening so as to obtain a feature set.
To reduce the dimensionality of the eigenvalue vector extracted from G, a set of common statistical features is selected, namely: a first quartile, a second quartile, a standard deviation, a mean, a minimum, a mode, a maximum, a median, a range, a coefficient of variation, a skewness, a kurtosis, a trend, a sequence correlation, a self-similarity, a periodicity, a root mean square, a percentile; all the parameters are tested by adopting chi-square feature selection, mutual information and recursive feature elimination methods respectively, and threshold screening based on Fisher criteria is carried out on the final features to obtain the most effective classification features, as shown in a formula (7):
wherein p is k ，σ k ，u k Respectively representing the mean value, variance and the duty ratio of the k class of the feature in the final feature to be screened; from this, a fischer judgment rate is obtained, which is used to set a threshold for the final feature screening, as in equation (8):
wherein, the value of alpha is set to 0.75, ρ is the feature quantity to be finally reserved, and v represents the feature corresponding to the threshold value; it should be noted that too few features may result in insufficient classification information, while too many features may increase the computational effort and may also have an overfitting effect on the final result.
In order to classify and identify the extracted and screened features, the embodiment adopts a classification method of an adaptive enhancement (AdaBoost) iterative algorithm and k-means clustering to further improve the classification accuracy; specifically, adaBoost is responsible for training weak classifiers to recognize EMG-related hand grip gestures, while k-means is used to learn classification models to distinguish between different hand grip gestures; when misjudgment exists in k-means classification, adaBoost modifies the weight according to the error rate, so that an optimal single weak classifier is obtained.
The clustering principle of k-means is as follows: firstly, selecting K points in a classification space as initial centroids; assigning each object to the centroid nearest to it; obtaining a new centroid of the region according to the distribution result; the above steps are repeated until the centroid is no longer changed.
The self-adaptive enhancement classifier is a machine learning algorithm and is used for training a group of weak classifiers, the classifier mechanism is to update the weight and error rate of a training sample after each iteration, the training weight of the training set with the error classification is increased, the weight of the training set with the correct classification is reduced, and the effective combination of the weak classifiers is realized through the iteration of negative feedback, so as to design a strong classifier, such as a formula (9):
wherein h is i Represents the ith order weak classifier, θ i Is a preset performance threshold value, prevents the program from being trapped in infinite loop, v is a characteristic vector, f i Representing the f-order component of the feature vector v, s representing the regularization parameter.
First, two sets of labeled training set data M 1 、M 2 Input trainer, wherein each set of data comprises (x 1 ,y 1 ),(x 2 ,y 2 ).....(x n ,y n ) The method comprises the steps of carrying out a first treatment on the surface of the x represents the EMG feature set which is reserved and input after feature screening; y represents its corresponding tag; presetting s=0.1, and initializing a weight matrix w ij Initializing a cycle count parameter i=1, j=1;
the weight matrix is then normalized as shown in equation (10):
a simple weak classifier is defined as equation (11):
h(v,f i ,θ i ,s)＝aδ(v f >θ i )+b (11)
wherein, delta represents a step function, when the f-order component of the feature vector v is larger than a preset performance threshold, the function value is constant at 1, otherwise, the function value is constant at zero; a and b are adjustment coefficients for correcting errors;
the classifier is a simple linear function with weaker classification effect and uses a weighting matrix w ij Training k-means clustering, and obtaining error related parameters a, b and f for a clustering result i 、θ i Expression of (c), as in formula (12):
wherein, the liquid crystal display device comprises a liquid crystal display device,is the regularized clustering result, and the value range is [1, -1]The method comprises the steps of carrying out a first treatment on the surface of the At error E i On the premise of minimizing, the parameters a, b and f which are most in line with the current weak classifier are found i 、θ i The optimal classification effect is achieved; traversing each order of the feature vector from i=1 to i=n to obtainTo the best performance weak classifier for each order feature vector, the final strong classifier is obtained using equation (9).
The surface myoelectric electrode used in this embodiment performs EMG signal acquisition at a sampling frequency of 500Hz, and then selects a butterworth band-pass filter to filter noise signals lower than 15Hz and higher than 500Hz, and at the same time, 50Hz power frequency notch is required to be performed on the myoelectric signal in order to eliminate interference of a power supply line.
The user needs to collect myoelectric control data before using the prosthetic hand, familiarize with the control relation between different myoelectric activation modes and the prosthetic hand, train the accuracy of the classification model, the prosthetic hand wearer dynamically selects the number of effective actions according to the muscle persistence condition of the residual limb of the user, because each grasping action needs to be matched with the specific myoelectric activation mode, each myoelectric activation mode is required to be independent, repeatable and actively controlled by the user; for repeated training of each myoelectric activation mode, the user is allowed to decide the speed and the force during myoelectric activation by himself, so long as the repeatability of the movement is ensured.
In the movement process, after the user inputs basic information and is trained, the user controls myoelectric activation of the residual limb to serve as a main signal source of movement control, and the self-adaptive enhanced k-means classifier classifies the screened characteristics and outputs the result as a final control instruction to drive the prosthetic hand.
After the prosthetic hand acts, the feedback signal is used as a motion gating signal to perform enabling control on the motion of the prosthetic hand, as shown in table 1; wherein K is E 、K C 、K F Respectively representing encoder feedback, current feedback and interaction force feedback in a feedback signal, and defining events that a motor does not move to a designated position, the current does not reach a threshold value and the force feedback does not reach the threshold value as logic 0; events of the motor reaching a specified position, the current exceeding a threshold value, and the force feedback exceeding the threshold value are defined as logic 1;
TABLE 1 feedback control
Multiplex feedback takes three forms: force feedback, encoder feedback, current feedback. In prosthetic hand control systems, force feedback is to prevent damage caused by overload of interactive forces during gripping; the encoder feedback judges the number of turns of the motor through the change of the windings of the encoder, so that the rotation angle of each joint and the relative position between the joints are obtained; the current feedback is used for guaranteeing the most basic electrical safety of the prosthetic hand and preventing the prosthetic hand from being burnt out by faults and a circuit thereof; the multipath feedback can be the perception feedback similar to the sense of touch and body sense for the prosthetic hand, and can also monitor the working state of the controlled motor in real time, thereby endowing the prosthetic hand with higher reliability and flexibility;
the piezoelectric sensors are arranged in a matrix and uniformly distributed in the soft prosthetic hand, each node in the matrix network is a pressure sensitive resistor, the size and the position of the resistance change are determined by scanning the current change of the rows and the columns, so that the contact position and the stress condition of the prosthetic hand and an object are determined, the piezoelectric sensors cover the inner sides of the palm and the fingers and serve as the 'skin' of the prosthetic hand, and a feedback signal similar to 'touch' is provided.
The encoder direct current motor can determine the forward rotation or reverse rotation stroke of the motor through the change of the resistance value of the encoder, and convert the stroke of the motor into the angle change of joint rotation; the rotation postures of the 14 joints are synthesized to determine the current movement posture of the prosthetic hand, the current movement posture is transmitted to the microcontroller and the upper computer, and the motor encoder can enable the control system to acquire the movement posture of the prosthetic hand joint in real time, so that the prosthetic hand is controlled to accurately move to a fixed posture, and a feedback signal equivalent to 'body sense' is provided.
The current feedback function is integrated in the motor driving module, and the power supply is stable under the condition of normal motion of the prosthetic hand, so that the current feedback reflects the running power of the motor; when the motion of the prosthetic hand motor is blocked, the current is increased suddenly, and the control system receives excessive feedback current to stop the motion of the prosthetic hand; the motor driving module receives signal input from the microcontroller, instructions are sent to the motor driving module in the form of pulse width modulation (Pulse Width Modulation, PWM) signals, the power input of the motor is provided through external power supply, and the PWM signals are converted into external power supply levels in the motor driving module to drive the motor to move. Meanwhile, a current feedback interface on the module can feed back the current passing through the direct current motor to the upper computer.
Control of the gripping process by the feedback signal as shown in fig. 6, the judgment condition in the flow chart is to judge the feedback signal, wherein the different event branches cover the typical gripping range, specifically:
(1) Presetting a current threshold, a force threshold and a pose state threshold; in the motion process of the prosthetic hand, the interaction force and the feedback current of the encoder are both within threshold values, and when each joint of the prosthetic hand reaches the expected motion position, the end of one normal grasping motion can be judged through the feedback of the joint motion state of the encoder motor.
(2) The current feedback of the prosthetic hand is higher than a threshold value in the motion process, which represents that the motor is blocked to a certain extent; the force feedback is also above the threshold, and the interaction between the prosthetic hand and the object can be considered to hinder further flexion and extension of the joint, and the grasping purpose is achieved at this time, so that in order to protect the prosthetic hand and the object to be grasped, the motor enters a hiccup control mode, the necessary interaction force is maintained, and meanwhile overload current is prevented.
(3) The current feedback of the prosthetic hand is higher than the threshold value in the motion process, but the feedback force is lower, and under the condition that no interaction force exists, the current overload represents the fault of the motion of the prosthetic hand, at the moment, the control system stops the current motion, and the encoder motor is disconnected to ensure the electrical safety.
Example 2
The embodiment provides a control method based on the prosthetic hand, which comprises the following steps:
acquiring an electromyographic signal of a prosthetic hand body wearer;
obtaining a motion control instruction according to the classification of the electromyographic signals, and driving the finger-like motion; extracting action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, carrying out integrated screening on the action recognition features, and then carrying out electromyographic classification by adopting a self-adaptive enhancement classifier;
and acquiring interaction force with an object, working current of a motor and joint pose when the finger is imitated, so as to control an execution mode of the motor after the finger is imitated.
In the control method, user basic information is determined, a control system is initialized, training of a myoelectric activation mode is carried out for a user to wear the bionic prosthetic hand and a myoelectric acquisition arm ring, a preset prosthetic hand action posture is set, and the selected myoelectric activation mode needs to have specificity, spontaneity and repeatability;
mapping the effective myoelectric activation modes which can be realized by the user into presupposed artificial limb manual operation one by one, and enabling the user to be familiar with control connection in the artificial limb manual operation; adjusting the number of movements to match the prosthetic hand wearer's residual limb activation provides sufficient freedom of movement to match all possible myoelectric activation patterns.
Further, after the user is familiar with the rules of operating the prosthetic hand, carrying out a grasping test of preset concentrated actions, and training a classification model through the obtained data with the labels; selecting a spectrum-based logarithmic transformation image signal as a myoelectric classification material, extracting a Laplacian matrix from the myoelectric signal through sliding window short-time Fourier transformation, obtaining each characteristic quantity of a characteristic matrix, and carrying out integrated characteristic selection on the characteristic quantity as a classification input; finally, a plurality of linear weak classifiers are trained by the self-adaptive enhancement k-means algorithm to form a self-adaptive enhancement classifier, and the self-adaptive enhancement classifier is used as a myoelectricity classification tool;
in the movement process, a user uses myoelectric activation of the residual limb of the user as a main signal source of movement control, classifies the myoelectric signals as final control instructions to drive the prosthetic hand, monitors the position signal of the prosthetic hand in real time through feedback control grasping, finishes movement when the prosthetic hand moves to a designated position, and selects intermittent hiccup control or disconnection of a motor to prevent damage to the prosthetic hand and a grasping object caused by movement interference when the feedback force and the current exceed a threshold value.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (6)
1. The multipath feedback myoelectric control prosthetic hand based on the self-adaptive enhancement classifier is characterized by comprising a prosthetic hand body, a motor driving module, a signal acquisition module, a classification module and a feedback module;
the artificial hand body comprises a plurality of independently moving artificial fingers arranged on a palm platform, and the artificial fingers are controlled to move by a motor driving module;
the signal acquisition module is used for acquiring an electromyographic signal of a prosthetic hand body wearer;
the classification module receives the electromyographic signals and is configured to extract action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, and after the action recognition features are subjected to integrated screening, the electromyographic classification is carried out by adopting a self-adaptive enhancement classifier to obtain a motion control instruction;
the classification accuracy is further improved by adopting a classification method of a self-adaptive enhancement iterative algorithm and k-means clustering;
specifically, the adaptive enhancement iterative algorithm is responsible for training a weak classifier to recognize EMG-related hand grip gestures, and k-means is used to learn a classification model to distinguish between different hand grip gestures; when misjudgment exists in k-means classification, the self-adaptive enhancement iterative algorithm modifies the weight according to the error rate, so that an optimal single weak classifier is obtained;
the motor driving module is configured to drive the finger-like motion according to the motion control instruction;
the feedback module is configured to control an execution mode of the motor after the finger-like action according to interaction force with an object, working current of the motor and joint pose when the finger-like action is performed;
the corresponding positions of the five fingers on the palm platform are sequentially provided with thumb metacarpophalangeal joints and interphalangeal joints, index finger metacarpophalangeal joints and interphalangeal joints, middle metacarpophalangeal joints and interphalangeal joints, ring metacarpophalangeal joints and interphalangeal joints and little metacarpophalangeal joints and interphalangeal joints; the thumb of the artificial hand corresponds to two encoder direct current motors, respectively controls the metacarpophalangeal joint and one interphalangeal joint of the thumb, and the other four fingers respectively correspond to three encoder direct current motors, respectively controls the metacarpophalangeal joint, the proximal interphalangeal joint and the distal interphalangeal joint;
the encoder direct current motor determines the forward rotation or reverse rotation stroke of the motor through the change of the resistance value of the encoder, and converts the stroke of the motor into the angle change of joint rotation;
the feedback module comprises interaction force feedback, encoder direct current motor feedback and motor working current feedback; the interaction force feedback is the interaction of force between the simulated finger and the grasping object, which is measured and fed back by the force sensor; the encoder DC motor feeds back finger motion simulating gesture, and the number of turns of the motor is judged through the change of the encoder windings, so that the rotation angle of each joint and the relative position between the joints are obtained; the motor driving module feeds back the working current of the motor;
if the interaction force and the motor working current are both in the threshold range, judging that one normal grasping movement is finished when each joint of the simulated finger reaches the predicted movement position through joint movement state feedback of the encoder direct current motor;
if the interaction force and the working current of the motor exceed the threshold range and each joint of the simulated finger reaches the expected movement position, controlling the motor to enter a hiccup control mode;
and if the working current of the motor exceeds the threshold range and the interaction force is lower than the threshold, controlling the motor to stop running.
2. The multi-path feedback myoelectric control prosthetic hand based on the self-adaptive enhancement classifier according to claim 1, wherein in the classification module, the action recognition feature extraction process comprises short-time Fourier transform after sliding window is carried out on the myoelectric signals, a spectrum image and a weight matrix thereof are obtained after logarithmic processing is carried out on a frequency density spectrum, and a Laplace feature matrix and a feature vector are obtained according to the spectrum image and the weight matrix thereof, so that various action recognition features are obtained.
3. The adaptive boost classifier based multi-feedback myoelectric control prosthetic hand of claim 1 wherein in the classification module the integrated screening process includes chi-square feature selection, mutual information feature selection and recursive feature elimination of motion recognition features followed by screening by fischer formula.
4. The adaptive boost classifier based multi-feedback myoelectric control prosthetic hand of claim 1, wherein the prosthetic units of the prosthetic hand index finger, the prosthetic middle finger, the prosthetic ring finger and the prosthetic little finger on the palm platform each comprise a fingertip joint protrusion member, a fingertip joint insertion member, an intermediate joint protrusion member, an intermediate joint insertion member, a proximal joint protrusion member, a proximal joint insertion member, a power gear, a transmission gear and a drive shaft; when the motor moves, the power of the motor is transmitted to a transmission gear nested on the finger joint of the stage through a bevel gear, the transmission gear drives a driving shaft meshed with the motor, and the driving shaft and the finger joint of the next stage are locked, so that the driving shaft drives the finger joint of the next stage to rotate, bend or stretch.
5. The adaptive boost classifier based multi-feedback myoelectric control prosthetic hand of claim 1, wherein the palm platform is designed with a proximal connection protruding shaft on one side of the five fingers, the proximal connection protruding shaft of the five fingers is coaxially connected with the end of the corresponding position of the prosthetic hand finger thereof to realize rotational movement, and a transmission gear is nested on the protruding shaft.
6. The control method based on the prosthetic hand according to any one of claims 1 to 5, characterized by comprising:
acquiring an electromyographic signal of a prosthetic hand body wearer;
obtaining a motion control instruction according to the classification of the electromyographic signals, and driving the finger-like motion; extracting action recognition features from the electromyographic signals through short-time Fourier transform and logarithmic spectrum images, carrying out integrated screening on the action recognition features, and then carrying out electromyographic classification by adopting a self-adaptive enhancement classifier;
and acquiring interaction force with an object, working current of a motor and joint pose when the finger is imitated, so as to control an execution mode of the motor after the finger is imitated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110584128.7A CN113171214B (en) | 2021-05-27 | 2021-05-27 | Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110584128.7A CN113171214B (en) | 2021-05-27 | 2021-05-27 | Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113171214A CN113171214A (en) | 2021-07-27 |
| CN113171214B true CN113171214B (en) | 2023-10-24 |
Family
ID=76927267
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110584128.7A Active CN113171214B (en) | 2021-05-27 | 2021-05-27 | Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113171214B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114224578B (en) * | 2021-12-06 | 2022-08-16 | 深圳市心流科技有限公司 | Semi-physical simulation test system and method for artificial limb test |
| CN116028841B (en) * | 2022-09-01 | 2023-08-15 | 贵州大学 | Work system and method for identifying article category by tactile perception |
| CN117679048B (en) * | 2023-12-14 | 2024-05-24 | 太极计算机股份有限公司 | Orthopedic efficacy evaluation method, system, device and electronic equipment |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH08294503A (en) * | 1995-04-25 | 1996-11-12 | Kenzo Akazawa | Motor-driven artificial hand |
| CN106890038A (en) * | 2017-03-21 | 2017-06-27 | 上海师范大学 | Prosthetic hand control system and its control method based on MYO armlets |
| CN107049570A (en) * | 2017-03-13 | 2017-08-18 | 日照若比邻机器人科技有限公司 | Manipulator control system |
| CN107657964A (en) * | 2017-08-15 | 2018-02-02 | 西北大学 | Depression aided detection method and grader based on acoustic feature and sparse mathematics |
| CN108272537A (en) * | 2018-03-25 | 2018-07-13 | 唐山云时代网络科技有限公司 | A kind of modular multiple degrees of freedom under-actuated bionic prosthetic hand |
| CN108814778A (en) * | 2018-07-19 | 2018-11-16 | 郭伟超 | A kind of myoelectricity humanoid dexterous prosthetic hand cascade Mach-Zehnder interferometer method and system |
| CN110163049A (en) * | 2018-07-18 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of face character prediction technique, device and storage medium |
| CN110731839A (en) * | 2018-07-18 | 2020-01-31 | 北京理工大学 | kinds of artificial hands |
| CN110946683A (en) * | 2019-11-25 | 2020-04-03 | 西安交通大学 | Electric stimulation driving method and system for artificial finger tip pressure and joint angle feedback |
| CN111209857A (en) * | 2020-01-06 | 2020-05-29 | 武汉科技大学 | Two-way recursive myoelectric feature selection method based on myoelectric signals |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8826733B2 (en) * | 2009-06-30 | 2014-09-09 | Orthosensor Inc | Sensored prosthetic component and method |
| WO2015006235A1 (en) * | 2013-07-12 | 2015-01-15 | The State Of Oregon Acting By And Through The State Board Of Higher Education On Behalf Of The University Of Oregon | Powered prosthetic devices using emg-based locomotion state classifier |
-
2021
- 2021-05-27 CN CN202110584128.7A patent/CN113171214B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH08294503A (en) * | 1995-04-25 | 1996-11-12 | Kenzo Akazawa | Motor-driven artificial hand |
| CN107049570A (en) * | 2017-03-13 | 2017-08-18 | 日照若比邻机器人科技有限公司 | Manipulator control system |
| CN106890038A (en) * | 2017-03-21 | 2017-06-27 | 上海师范大学 | Prosthetic hand control system and its control method based on MYO armlets |
| CN107657964A (en) * | 2017-08-15 | 2018-02-02 | 西北大学 | Depression aided detection method and grader based on acoustic feature and sparse mathematics |
| CN108272537A (en) * | 2018-03-25 | 2018-07-13 | 唐山云时代网络科技有限公司 | A kind of modular multiple degrees of freedom under-actuated bionic prosthetic hand |
| CN110163049A (en) * | 2018-07-18 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of face character prediction technique, device and storage medium |
| CN110731839A (en) * | 2018-07-18 | 2020-01-31 | 北京理工大学 | kinds of artificial hands |
| CN108814778A (en) * | 2018-07-19 | 2018-11-16 | 郭伟超 | A kind of myoelectricity humanoid dexterous prosthetic hand cascade Mach-Zehnder interferometer method and system |
| CN110946683A (en) * | 2019-11-25 | 2020-04-03 | 西安交通大学 | Electric stimulation driving method and system for artificial finger tip pressure and joint angle feedback |
| CN111209857A (en) * | 2020-01-06 | 2020-05-29 | 武汉科技大学 | Two-way recursive myoelectric feature selection method based on myoelectric signals |
Non-Patent Citations (2)
| Title |
|---|
| 基于肌电信号层级分类的手部动作识别方法;赵漫丹等;《北京生物医学工程》;20141015;第33卷(第05期);全文 * |
| 基于表面肌电信号的时频组合特征融合识别;杨新亮等;《华中科技大学学报(自然科学版)》;20111115;第39卷;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113171214A (en) | 2021-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113171214B (en) | Multi-path feedback myoelectric control prosthetic hand based on self-adaptive enhancement classifier and method | |
| Meattini et al. | An sEMG-based human–robot interface for robotic hands using machine learning and synergies | |
| US10959863B2 (en) | Multi-dimensional surface electromyogram signal prosthetic hand control method based on principal component analysis | |
| US11069148B2 (en) | Visualization of reconstructed handstate information | |
| Wang et al. | Design and myoelectric control of an anthropomorphic prosthetic hand | |
| Bi et al. | A review on EMG-based motor intention prediction of continuous human upper limb motion for human-robot collaboration | |
| US10496168B2 (en) | Calibration techniques for handstate representation modeling using neuromuscular signals | |
| Hussain et al. | The soft-sixthfinger: a wearable emg controlled robotic extra-finger for grasp compensation in chronic stroke patients | |
| Gijsberts et al. | Stable myoelectric control of a hand prosthesis using non-linear incremental learning | |
| Hahne et al. | Linear and nonlinear regression techniques for simultaneous and proportional myoelectric control | |
| Ibrahimy et al. | Design and optimization of Levenberg-Marquardt based neural network classifier for EMG signals to identify hand motions | |
| CN112074225A (en) | Method and apparatus for reducing neuromuscular signaling artifacts | |
| Dwivedi et al. | A learning scheme for EMG based decoding of dexterous, in-hand manipulation motions | |
| Yamanoi et al. | EMG-based posture classification using a convolutional neural network for a myoelectric hand | |
| Patel et al. | A classification method for myoelectric control of hand prostheses inspired by muscle coordination | |
| Stival et al. | Online subject-independent modeling of semg signals for the motion of a single robot joint | |
| Bimbraw et al. | Towards sonomyography-based real-time control of powered prosthesis grasp synergies | |
| Ryait et al. | Study of issues in the development of surface EMG controlled human hand | |
| Fan et al. | Robust neural decoding for dexterous control of robotic hand kinematics | |
| KR100994408B1 (en) | Method and device for deducting pinch force, method and device for discriminating muscle to deduct pinch force | |
| Shair et al. | Finger movement discrimination of EMG signals towards improved prosthetic control using TFD | |
| Mohebbian et al. | A comprehensive review of myoelectric prosthesis control | |
| JP3816762B2 (en) | Neural network, neural network system, and neural network processing program | |
| Wang et al. | Multi-finger myoelectric signals for controlling a virtual robotic prosthetic hand | |
| Kim et al. | Using common spatial pattern algorithm for unsupervised real-time estimation of fingertip forces from sEMG signals |
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 |