Introduction to the information source mode used in upper limb prosthesis control

The human limb is a complex and delicate dynamic system. How to recreate this structure is the main content of prosthetic research. With the development of technologies such as rehabilitation medicine and rehabilitation engineering, materials science, electronics and control theory, revolutionary research has also taken place in prosthetic research. Prosthetics are no longer just an ornament. People put forward higher requirements for comfort, practicability, accuracy and flexibility. Prosthetics are also moving towards intelligent, aesthetic and humanoid. At present, there are no more than a thousand kinds of commercialized prostheses worldwide.

In prosthetic control, the upper limb prosthesis has received much attention due to its high complexity and difficulty in controlling the error. From the current situation, mechanical prostheses and myoelectric prostheses have matured and commercialized. Reconstruction of prosthetic limbs and voice-activated prostheses have also taken the experimental stage, while upper limb prostheses controlled by biological information such as EEG and brain magnetics are still under study. in. In addition, in order to improve the function of the prosthesis, there has also been a form of closed-loop control that adds feedback to the open-loop control, and some progress has been made. For example, a prosthetic system constructed by Mabuchi et al. can not only transmit the stimulus to the subject to produce a somatosensory sensation corresponding to the original stimulus, but also has a corresponding relationship in magnitude.

In the control method, it is an ideal solution to control the prosthetic hand according to the human will. At this time, the control signal source may be taken out in three ways: (I) directly taking out instructions from the human central nervous system; (II) taking out signals from the motor nervous system; (III) taking the motor nervous system The myoelectric potential generated by the pulse reaching the muscle is taken out. At present, the myoelectric control prosthesis studied by the third method is dominant. If specifically classified, the current information source modes used in various upper limb prosthesis control are mainly the following:

l, using the mechanical movement of the body as a source of information

The control idea of ​​the upper limb prosthesis is mainly to use the residual motion function of the patient to trigger the corresponding switch through the transmission device for control. The contraction/diastolic and bulging of the stump muscles of the amputated patient, the expansion of the limbs such as the chest, the lifting/sagging of the shoulders, and the abduction/adduction can be used as information sources for the prosthetic drive. This kind of artificial limb control structure is simple and low in cost, so it is welcomed by patients. In the first chapter, the introduction will still have a certain market in the future.

From the current point of view, the development of this kind of upper limb prosthesis mainly depends on improving the detection means of information sources and improving the control strategy of prosthetic limbs. For example, Gu Yufufu uses a three-position positioning measuring device to detect the specific movement of the patient, and uses a neural network to determine the corresponding control command. According to Simpson's idea of ​​controlling the prosthesis by measuring the angle of the patient's undamaged joint, Aghili et al. establish a map of the relationship between the closed-loop curve of the shoulder joint motion angle and the action, and use the discriminant function to determine the angle of the elbow joint and the wrist joint. Vector. Rochel et al. used surgical procedures to connect the tendon to the prosthetic drive and directly drive the prosthetic movement by sensing the movement of the tendon.

Although the use of thousands of prosthetic patients over decades has shown that this type of upper limb prosthesis is highly reliable and robust, the control tasks it can achieve are very simple, combined with the residual limb response to the human brain. The sluggishness makes it less flexible and is not suitable for use as a source of information with high accuracy and combination. For example, the prosthesis designed in the literature can only complete the predetermined three steps of combing hair, taking the phone and drinking water, and the accuracy of the prosthetic control is only 85%.

2. Using myoelectric (EMG) as a source of information

The information source of the myoelectric prosthesis comes from the action potential of the muscles of the residual limb. Because it can reflect people's limb movement information, it is superior to mechanically-actuated prosthesis. It is not only favored by patients, but also has a broad market, and it has become a hot spot in the study of upper limb prosthesis.

The EMG signal is extremely weak and is often submerged in other biosignals and external noise, so a high quality EMG inspection system must be used. At present, the detection method mainly relies on the surface electrode and the needle electrode, and the EMG is taken out from the appropriate amputation site, and then filtered. A prototype of a radio probe has also been developed but is still in the experimental phase.

Although the myoelectric prosthesis has achieved great success in practical applications, when the amputee's residual limb is too short, or muscle atrophy due to paralysis, it is impossible to provide sufficient control information for the myoelectric prosthesis. At the same time, muscle fatigue, changes in electrode position, training of myoelectric signals, and fluctuations in body weight can change the characteristic values ​​of myoelectric signals, making it difficult to improve the control accuracy of multi-degree-of-free myoelectric prostheses. In addition, since the user must learn unnatural movements to drive the movement of the hand, and also due to the ability to decode the EMG signals, the degree of freedom that the myoelectric prosthesis can control is also limited.

For these reasons, most of the myoelectric prostheses in practical applications are still single degrees of freedom, controlled by the contraction of the upper arm muscles. From the current research results, the improvement of myoelectric prosthesis is mainly from the following aspects in the first chapter:

1: Adopt a new strategy to increase the number of degrees of freedom that a single channel can control.

2: New technologies including signal filtering, spectral analysis and pattern recognition technology are used to obtain more independent channel numbers from EMG signals.

3: Introduce feedback technology into the myoelectric prosthesis so that the myoelectric prosthesis can perform more functions. But it also relies on the development of new electrodes that can stimulate large numbers of nerve fibers at the same time.

4: Develop an adaptive myoelectric prosthesis with adaptive ability to compensate for the defects of myoelectric signal changes during use.

5: Combine other sources of control information to compensate for the lack of EMG information. Such as the mechanical movement of the residual limbs and shoulders.

3. Reconstructing the source of prosthetic control

In order to break through the obstacles that electronic prosthetics use to improve the accuracy of myoelectric control, Professor Hu Tianpei and Academician Chen Zhongwei fundamentally changed the original model, starting with an alternative control signal source, and applying microsurgical techniques to the field of rehabilitation engineering. A "finger" is created on the stump of the disabled person as a source of information that can accurately transmit information about the movement of the human brain. The method is: after general anesthesia, the patient is applied with microsurgery to transplant the second toe with vascular pedicle to the stump of the right forearm, and then reconstruct the "finger" to survive and then perform rehabilitation function training, and then use physics. Methods (such as temperature, pressure, displacement) convert control information into operational commands to achieve accurate control of electronic prosthetic hands.

This kind of artificial limb with the re-creation of "referring to" as the information source has opened up a new concept of close integration of medicine and engineering, which has important guiding significance both in theory and in practice. Tracking the use of the patient shows that the prosthesis can not only correctly transmit the brain's motion information, but also reconstruct the weight perception ability; the control accuracy is high, and the false motion rate of the developed three-degree-of-freedom six-motion prosthetic hand is 0%.

However, the patient has to suffer from the pain of surgery when installing the prosthesis, and the necessary microsurgical techniques also limit its promotion and application, and the recovery period required after surgery is longer. In addition, the amount of information that a single reconstruction can express is very limited, and it is difficult to deliver complex instructions of the brain quickly and in time. When the degree of freedom increases, the number of codes is also increased, which not only brings difficulties to the patient's mastery, but also reduces the stability of the system and the accuracy of the action. In order to minimize the suffering of patients, Dr. Chen Zhongwei is currently working on the "localized vascular neuromuscular and / or sputum transplantation constitutes a signal source", so that it can replace the re-creation of "finger" accurate delivery of brain instructions.

4, with sound as a signal source

Voice-activated prostheses have unparalleled advantages in helping paraplegic patients recover their motor function. Using the patient's voice information and transforming it into corresponding control commands through digital processing technology (DSP), not only the line is simple, but also opens up new avenues for the development of multiple free vacations.

At present, voice control technology has been widely used in various fields. For example, the voice-activated robot developed by the American company can provide a stable and clear image for microsurgery by placing a slender surgical instrument such as an endoscope through a small incision according to the doctor's password. The voice-controlled wheelchair model designed by WalterR achieves almost 100% control accuracy at low and medium levels of ambient noise.

The application of voice control technology on artificial limbs has also taken the experimental stage and started to go to market. For example, the simple language instruction prosthetic control system designed in the literature not only has good control effect, but also can program the system to complete more functions when the number of instruction words increases. Document 1. The proposed language recognition The portable DSP system provides sufficient control commands to accurately control the electric prosthesis.

Compared with myoelectric controlled prostheses, voice-activated prostheses can perform more control functions, more convenience, and higher precision. Its shortcoming is that the usual communication between the patient and others may also cause the prosthesis to malfunction. In addition, how to improve the ability of the voice-activated prosthesis to resist environmental noise. It is also an urgent problem to be solved for this type of prosthesis.

5. Using EEG as a signal source

EEG activity is essentially the electrical noise generated by nerves and suddenness in the work of the central nervous system. As early as 1875, British physiologist Richard Caton recorded a weak current from the rabbit brain and monkey brain. Brain science research shows that there is a connection between EEG activity and exercise information. If the EEG or some of its components can be converted into a new output channel, information exchange and control with the outside world can be used to control the prosthesis. To this end, many scholars have conducted a lot of research in this regard. The literature mocks the amplitude of the short latency visual evoked potential (VEP) through real-time analysis and transforms it into the input of the neuromuscular controller to achieve knee extension control with an accuracy of 95%. Pfurtscheller et al. designed a brain-computer interface device (BCIs) that can analyze finger movements based on the user's imaginary motion, with an accuracy of about 70%, and can distinguish between left-hand and right-hand imaginary movements. Roberts et al. presented a faster and more accurate real-time BCIs interface system based on single-channel differential EEG signals. The detected EEG signals were classified by an 8th-order AR model and a Bayesian logic classifier. Controls the up and down movement of the mouse, and the overall performance is up to 82%.

Using EEG activity recorded from the scalp as a source of information, neuromuscular control is not required, so that even the most severely disabled patients can use it. Moreover, electrical stimulation does not affect the recording of EEG signals. However, the EEG process is very complicated, and its research on it is only limited. There is still a long way to go before constructing a system that is completely controlled by EEG and relying on brain science research.

6, with brain magnetic as a signal source

Currents in muscles and nerves produce external magnetic fields, such as cardiac magnetism and brain magnetics. The first reliable experimental measurement of biomagnetism was done in 1963 by Baule and McFee, who recorded the magnetism for the first time using two side-by-side coils (gradient meters). A few years later, S. Williamson et al. confirmed that the use of a two-step meter can measure brain magnetic signals in a background noise-rich environment (such as urban areas) without magnetic shielding, thereby promoting the development of biology. .

To develop prostheses for brain magnetic control, the first thing that should be solved is how to construct a human-machine interface to establish a reasonable mapping relationship between brain magnetic signals and prosthetic control instructions. However, due to the complexity of brain magnetics, people's understanding of it has just started, and its research still remains in the determination of important functional areas in brain surgery and pathological evaluation of brain functional deficiency. Applying it to prosthetic control can only be in the exploratory phase.

The above-mentioned flaws in the control information source force rehabilitation workers to make unremitting efforts to find new sources of information. Neural signals are considered to be the most ideal mode. It is not only affected by the degree of human fatigue, but also has high reproducibility, and the neural information does not interfere with each other, and has excellent definition.

Many studies have also confirmed that the nervous system is malleable, not only as a means of compensating and adapting to various external stimuli, but more importantly, it has the above-mentioned characteristics of nerve energy information that has structural and functional repair or reconstruction after injury. The more realistic multi-degree of freedom of the bionic function lays the foundation for the upper limb prosthesis. Based on this, Wan et al. proposed the idea of ​​using neural networks to transform human upper limb nerve signals into control commands to control prostheses. Many scholars have also made useful discussions on the use of neural information to control prosthetic techniques, and have achieved a consensus:

1) If a silicon wafer with many micropores is placed on the severed nerve, the nerve cells will regenerate and make electrical connections with them;

2) The electrode can obtain a stable discharge record during a continuous neural cell response time period;

3) There is a strong relationship between the discharge rate of the nerve and the spatial direction of the upper arm movement.

The above results make it possible to apply neural information to control prosthetic technology. Using neural buried electrodes to guide out neural information, and establishing a mapping relationship between neural information and limb movement after pattern classification, thus controlling prosthetic movement, will be an ideal prosthetic control mode. We have reason to believe that neural prostheses will be developed for use in a wide range of patients in the next few years. Its success will surely bring a new leap to the improvement of human living conditions.

More related content Scan QR code to follow official WeChat