Hybrid control method for bionic upper limb prostheses

Abstract

The invention relates to a hybrid control method capable of mapping the muscle energy range of a person to the average speed range of the prosthesis used with a much lower average speed error.

Description

[0001] HYBRID CONTROL METHOD FOR BIONIC UPPER LIMB PROSTHESES

[0002] The technical field related to the invention:

[0003] The invention relates to a hybrid control method capable of mapping a person's muscle energy range to the prosthesis's average speed range with a much lower average speed error.

[0004] State of the art:

[0005] The desire of the amputee to move the missing limb can be obtained by "surface electromyography" (sEMG) signal without the need for surgical intervention. Therefore, these signals are frequently preferred in prosthesis control. Current sEMG-controlled upper limb prostheses are divided into two types: pattern recognition-based and conventional (non-pattern recognition-based).

[0006] Pattern recognition based prostheses work based on patterns (datasets) obtained during training, which are easily affected by environmental factors (electrode drift, muscle fatigue, sweating, etc.). Changes in sEMG signals due to environmental conditions may cause the relationship between the signal and the patterns to fail to be obtained with high reliability. In conventional (non-pattern recognition-based) prostheses, the sEMG signal is processed in various steps (amplification, bandpass filtering, rectification, etc.) to generate the control signal and the envelope of the signal is obtained by taking the root mean square or mean absolute value of the signal. The processed and enveloped sEMG signal (cEMG) becomes applicable to three different control methods (on-off, myo-pulse and proportional control methods) known to be used commercially in prosthesis control.

[0007] An academic study shared how three commercially available control methods reflect the user's muscle energy range to the average speed of the prosthesis (Qelik, 2023). Table 1 shows the results obtained in the study. Table 1 . Average absolute error values produced by the three controllers when the scaled sEMG RMS value is less and greater than 0.5 and for all data

[0008] E E

[0009] E (yEMG<0.5) (yEMG>0.5)

[0010] On-Off 0.5578±0.024 0.684±0.113 0.208±0.143

[0011] Myo-Pulse 0.1224±0.074 0.094±0.385 0.198±0.097

[0012] Proportional 0.045±0.042 0.023±0.015 0.106±0.03

[0013] As indicated by the results in the table, on-off, myo-pulse and proportional control methods were analysed in the study, and it showed that the proportional control method produced the least average speed error (Table 1 and Figure 1 ). The on-off control method produced the highest average speed error for the whole dataset. The myo-pulse method produced less average speed error than on-off control but more than proportional control. The experimental results (for the whole data set) show that the proportional control method produces the least error. However, it was observed that the mechanical speed outputs produced for medium and high muscle energy remained below the reference speed, even with proportional control. Therefore, although all three control methods move the muscle energy range to different speed ranges, none can do this mapping perfectly. Compared to the three existing methods, a hybrid control method is needed.

[0014] The only known state of the art is the three control methods (on-off, myo-pulse and proportional), which are only available in academic studies and commercial products. Although various proposals and applications have been developed for controlling artificial limbs, these developments remain insufficient. Some applications of inventions designed for this purpose are given below.

[0015] The invention subject to the application numbered "CN114897012" in the state of the art relates to an intelligent artificial limb arm control method based on a liveness interface. The invention comprises a task analysis phase: collecting sEMG signals of a subject in the process of a hand or elbow movement, performing muscle collaborative extraction and predicting a corresponding hand movement or elbow movement result through model mapping calculation; a task analysis phase: collecting a group of sEMG signals in the process of hand or elbow movement, respectively evaluating the myoelectric data quality of the current patient. The invention subject to the application numbered "CN113952093" in the state of the art discloses an unsupervised calibration method for myoelectric artificial limb control. The invention focuses on a regression-based proportional control system, the unsupervised calibration being performed on an artificial limb control model using a data weighting criterion based on myoelectric characteristic spatial correlation.

[0016] In the known state of the art, the on-off method gives maximum speed output for low muscle energy. Generally, on-off control always outputs maximum speed regardless of muscle energy and naturally produces significant speed error. When all the available data are considered together, the myo-pulse method can produce speeds below the reference speed and less error compared to the on-off control. However, in the presence of low amplitude sEMG signals, the myo-pulse control method sometimes fails to produce speed output, and when it does, it produces a very low average speed compared to the input signal. Proportional control follows the reference speed with little error for low- and medium-amplitude contractions but remains below the reference speed for medium- and high-amplitude contractions, producing relatively more error. Therefore, there is a need for a hybrid control method that maps the individual's muscle energy range to the average speed range of the prosthesis with a much lower average speed error by using a hybrid method with on-off, myo-pulse and proportional controllers.

[0017] As a result, due to the problems mentioned above and the inadequacy of the existing solutions on the subject, it has become necessary to make a development in the relevant technical field.

[0018] The objective of the invention:

[0019] The invention relates to a hybrid control method that can map a person's muscle energy range to the prosthesis's average speed range with a much lower average speed error.

[0020] The most important aim of the invention is to produce an average prosthesis speed output proportional to the amount of muscle energy so that the control of the prosthesis is closer to natural (more human-like). In other words, it uses less muscle energy for low prosthesis speed and (medium or) more muscle energy for (medium or) high prosthesis speed output. In other words, the output average prosthesis speed is an (approximately) linear function of the input sEMG amplitude.

[0021] A further aim of the invention is to enable the output prosthesis speed to follow the reference speed with very little error for all muscle energy amplitudes.

[0022] Description of the figures:

[0023] FIGURE-1 : This figure shows the scaled sEMG RMS value versus the scaled average speed outputs for three different controllers.

[0024] FIGURE -2: This figure shows the scaled sEMG RMS value versus the scaled average speed outputs for the hybrid controller and the other three controllers.

[0025] FIGURE -3: p is a plot of the scaled average speed outputs against the scaled sEMG RMS value divided into i equal parts to determine the deviation threshold value.

[0026] Description of the invention

[0027] The invention relates to a hybrid control method that can map a person's muscle energy range to the prosthesis's average speed range with a much lower average speed error.

[0028] For a reliable experiment, the sample size was determined as 22 by statistical methods (one-way ANOVA, p<0.05, power(1 -|3)>0.8). Since the proportional controller design will be different for each subject (the maximum contraction level will vary according to the user), only 30 data from one subject were used in the study. The subject was asked to sit on a chair and place his arm comfortably on the chair arm. Electrodes were placed on the extensor digitorium muscle of the subject. With the command "open", the subject performed a wrist extension movement and surface EMG data was recorded for 3 seconds by the muscle tester with a sampling frequency of 1000 Hz.

[0029] In order to obtain meaningful contraction data, 30 pieces of data obtained from the subject were first filtered with a bandpass filter with cut-off frequencies of 20-500 Hz, and then the negative components were rectified on the rectifier stage. The signal (rEMG) obtained at the output of the rectification stage is ready to be applied to the myo-pulse controller. For on-off and proportional controllers, the rEMG signal was passed through a low-pass filter with a cut-off frequency of 2 Hz, and the signal envelope was obtained. The enveloped signal (cEMG) was thresholded with a userspecific threshold value (5) to generate the control signals. These control signals for on-off, myo-pulse and proportional controllers were applied to a commercial prosthetic hand, and the position / speed changes in the prosthetic hand were recorded. A commercial prosthetic hand was used in the study. With an additional apparatus placed on the prosthetic hand, position changes in the prosthetic hand were recorded.

[0030] Ideally, assuming a linear (proportional) relationship between the sEMG signal energy (user muscle energy consumption) and the average prosthetic hand speed output, this relationship is expressed by the mathematical equation v(t)=sEMG. In this equation, sEMG is the normalised sEMG signal and v(t) is the normalised average speed signal. The average absolute distance (deviation-error) of the speed outputs obtained for all three control methods from this ideal curve is denoted by E. Table 1 and Figure 1 show the scaled average speed outputs obtained against the scaled sEMG signal for all three methods.

[0031] Figure 1 shows that the on-off control method provides similar speed outputs (constant speed) at all sEMG signal energies. The fact that the prosthesis outputs at almost full speed for a relatively low energy sEMG signal causes the error in speed to be the largest, as shown in Table 1. The analysis of myo-pulse and proportional control methods shows that both offer variable speeds depending on the energy of the input sEMG signal. When the whole data set is analysed, it is observed that the lowest error is in the proportional control method. When the data are analysed in regions, it is seen in Table 1 that in the area where the normalised sEMG RMS value is less than 0.5, the proportional control follows the reference speed well, but in the region where the RMS value is greater than 0.5, the speed produced by the proportional controller is below the reference speed and the relative speed error increases.

[0032] It is clear from the results of the experiments that each control method works effectively within a certain speed range and that no control method can cover the entire speed range (and, therefore, the sEMG energy range). The implication of not covering the entire sEMG energy range for the user in everyday use is that more muscle contraction (effort) results in less work (average mechanical speed) and less efficient use of the prosthesis. In order to use the prosthesis more efficiently, a hybrid model was designed to cover the entire speed and sEMG range. According to this model (for this data set), a proportional controller will be used in the region where the normalised sEMG RMS value is less than 0.5, while an on-off controller will be used in the region where it is greater. Table 2 and Figure 2 show the output obtained when using a hybrid controller designed in this way.

[0033] Table 2. Average absolute error values produced by the three controllers and the hybrid controller when the scaled sEMG RMS value is less than 0.5, greater than 0.5 and for all data

[0034] E _E_E

[0035] (SEMGRMS<0.5) (SEMGRMS>0.5)

[0036] On-Off 0.5578±0.024 0.684±0.113 0.208±0.143

[0037] Myo-Pu 0.1224±0.074 0.094±0.385 0.198±0.097

[0038] Proporti 0.045±0.042 0.023±0.015 0.106±0.03

[0039] Hybrid 0.0279±0.0176 0.0266±0.0155 0.037±0.0208

[0040] When Table 2 and Figure 2 are analysed together, the hybrid control method gives better results than other controllers for the whole speed and sEMG range. This result is achieved by using the on-off controller in the region where the normalised sEMG RMS value is greater than 0.5 for this data set. As can be seen from Table 2, in the region where the normalised sEMG RMS value is greater than 0.5, the output error of the hybrid controller is considerably lower than the other three controllers.

[0041] This study has shown that the three traditional control methods (on-off, myo-pulse and proportional) cannot provide a solution that covers all sEMG and prosthesis speed ranges. The maximum contraction amplitude and threshold value of the amputee who will use the prosthesis will be determined with sEMG samples from the muscles that will control the prosthesis. With this data, the curve of the ideal speed output curve given in Figure 1 and Figure 2 can be determined. When the deviation value (i.e. error value) from the curve is below a specified deviation threshold (p) value, the control output will be generated using a proportional controller, and when this error value is exceeded, the control output will be generated using an on-off controller. of the Hybrid Control Method

[0042] - The sEMG signals obtained from the user are transmitted wired or wirelessly to the signal processing and amplification circuit. The surface EMG signal from the user is amplified in this circuit (with active or passive amplifier circuits), bandpass filtered (active or passive bandpass filters can be used; in the literature, a bandpass filter with cut-off frequencies between 20 and 500 Hz is generally preferred), rectified and the envelope of the signal is obtained with the help of a low-pass filter (with a cut-off frequency between 2-10 Hz, which can be created with active or passive circuit elements). Thus, the cEMG signal is obtained. This signal can now be applied to controllers such as microprocessors / microcontrollers.

[0043] - A proportional controller is designed using any embedded system controller (microprocessor, microcontroller, FPGA, etc.).

[0044] - With the comparison function to be performed in the controller, it is checked whether the cEMG signal amplitude is above the threshold value 5.

[0045] - If the cEMG amplitude is not above the threshold value, no speed output is generated in the prosthesis.

[0046] - If the cEMG amplitude is above the threshold value 5 and below the deviation threshold value p, the prosthesis speed output is generated using the proportional controller.

[0047] - If the cEMG amplitude is above the threshold value 5 and the deviation threshold value p, an on-off controller is used, and the prosthesis is moved with the speed v_max. Control Method Parameters Finding Operations

[0048] In order to use the hybrid control method, the threshold value 5 (the threshold value used to generate the control signal), the deviation threshold value p, which determines after which surface EMG amplitude value the control method is changed, and the maximum contraction intensity amplitude of the user must be determined. The following steps are applied to determine these variables:

[0049] - Firstly, surface EMG data is recorded from the user during a certain number of strong contractions.

[0050] - The main purpose of the thresholding function, whose value is expressed by the symbol 5 and used in generating the control signal, is to eliminate low and unstable data in the processed surface EMG. However, since the amplitude of the surface EMG signal varies depending on age, gender and muscle group, there is no optimal algorithm in the literature for thresholding. For this reason, a threshold value that separates noise and involuntary contraction from voluntary contraction is usually determined by studying a series of repetitions of signals from the user. Choosing a large amplitude threshold will require too much effort from the user and cause fatigue while choosing a low amplitude threshold will cause the prosthesis to behave erratically due to small involuntary contractions. A 5 value is determined by paying attention to these principles.

[0051] - According to the data obtained from the user, the maximum contraction amplitude value (EMGmax) of the user is obtained by finding the highest contraction intensity value.

[0052] - The RMS values of the surface EMG signals received from the user are obtained, and the values are scaled to the range 0-1. The RMS value of the signal can be calculated by the following formula: In the formula, represents the first moment when the EMG data crosses the threshold and T2represents the first moment when the EMG signal data falls below the threshold after the contraction. This value, obtained by taking RMS, is an indicator of the energy of the user contraction intensity.

[0053] - cEMG Signal Acquisition: The sEMG signals obtained from the user are applied to the signal processing and amplification circuit.. The surface EMG signal from the user is amplified in this circuit (with active or passive amplifier circuits), bandpass filtered (active or passive bandpass filters can be used; in the literature, a bandpass filter with cut-off frequencies between 20 and 500 Hz is generally preferred), rectified and the envelope of the signal is obtained with the help of a low-pass filter (with a cut-off frequency between 2-10 Hz, which can be created with active or passive circuit elements). Bbylece cEMG i§areti elde edilir. This signal can now be applied to controllers such as microprocessors / microcontrollers.

[0054] - A proportional controller is designed using any embedded system controller (microprocessor, microcontroller, FPGA, etc.). The proportional controller used in the study works as follows: o With the comparison function to be performed in the controller, it is checked whether the cEMG signal amplitude is above the threshold value 5. o If the cEMG amplitude is not above the threshold value, no speed output is generated in the prosthesis, and a new surface EMG signal is obtained from the user, returning to the "cEMG signal acquisition" process step. o If the cEMG amplitude exceeds the threshold value 5, the prosthesis speed output is generated using the proportional controller. Proportional controller speed output v(t) is calculated by the following formula. In the formula, cEMGmaxdefines the maximum processed sEMG amplitude determined for the user, vmaxdefines the maximum speed at which the prosthesis can move, redefines the minimum speed at which the prosthesis can move, and 5 defines the threshold value. z,Vmaxvmin , v(t)=cEMG -max- X 8X cEMG(t) +vmin

[0055] - Using a device such as a data card (DAQ) or a digital analogue converter, the pre-recorded and processed cEMG signals are generated one by one (simulating the user) and applied to the prosthesis controlled by a proportional controller. The prosthesis-generated speed outputs of each data are recorded.

[0056] - The RMS values of the recorded speed outputs are calculated (formula on the previous page) and scaled from 0 to 1 .

[0057] - The scaled RMS values of the obtained surface EMG signals and speed outputs are mapped to each other (an example data set is shown in Figure 1 and Figure 2). A linear curve v=sEMG is fitted to the graph, and the distances of each data to this curve with respect to the y-axis are calculated and summed. This curve represents the ideal projection of the surface EMG signal energy onto the prosthesis speed output. The sum of the distances of the data from this curve represents the total speed deviation error.

[0058] - After the total speed deviation error is determined, the x-axis, i.e. the scaled sEMG interval shown in Figure 1 and Figure 2, is divided into i equal intervals for a preferred number i, where i>2. When this number is large, a smaller speed deviation error value is obtained, and when it is small, a larger speed deviation error value is obtained. The example where the graph is divided into i parts is shown in Figure 3. In the application, the speed deviation error of each data with sEMG values in the range [m n2] is determined and summed iteratively. The total value is divided by the total number of data to determine the average total speed deviation error for this area.

[0059] - If the average speed deviation error calculated for the interval [ni n2] is greater than the average speed deviation error obtained over the entire sEMG range, the deviation threshold value p is set to the smallest value of this range, n1 , and the iteration is terminated.

[0060] - If the average speed deviation error calculated for the interval [ni n2] is smaller than the average speed deviation error obtained over the entire sEMG interval, the average speed deviation error is determined for the data in the next interval [n2 ns].

[0061] - The iteration continues until the average speed deviation error value calculated for any interval is greater than the average speed deviation error value of all data. When this interval is found, the deviation threshold p is equal to the smallest value of the interval.

[0062] Determination of Threshold Value 6, Deviation Threshold Value u, and EMGmax:

[0063] In order to use the hybrid control method, the threshold value 5 (the threshold value used to generate the control signal), the deviation threshold value p, which determines after which surface EMG amplitude value the control method is changed, and the maximum contraction intensity amplitude of the user must be determined.

[0064] - To determine these constants, surface EMG data are first recorded from the user during a certain number of strong contractions.

[0065] - The main purpose of the thresholding function, which is determined by the value of 5 and used to generate the control signal, is to eliminate low and unstable data in the processed surface EMG. However, since the amplitude of the surface EMG signal varies depending on age, gender and muscle group, there is no optimal algorithm for threshold determination in the literature. For this reason, a threshold value that separates noise and involuntary contraction from voluntary contraction is usually determined by studying a series of repetitions of signals from the user. Choosing a large amplitude threshold will require too much effort from the user and cause fatigue while choosing a low amplitude threshold will cause the prosthesis to behave erratically due to involuntary small contractions. A 5 value is determined by considering these principles.

[0066] - According to the data obtained from the user, the maximum contraction amplitude value (EMGmax) of the user is obtained by finding the highest contraction intensity value. - The RMS values of the surface EMG signals received from the user are obtained, and the values are scaled to the range 0-1. The RMS value of the signal can be calculated by the following formula:

[0067] In the formula, represents the first moment when the EMG data crosses the threshold and T2represents the first moment when the EMG signal data falls below the threshold after the contraction. This value, obtained by taking the RMS, is an indicator of the energy of the user contraction intensity.

[0068] - cEMG Signal Acquisition: The sEMG signals obtained from the user are applied to the signal processing and amplification circuit. The surface EMG signal from the user is amplified in this circuit (with active or passive amplifier circuits), bandpass filtered (active or passive bandpass filters can be used, a bandpass filter with cut-off frequencies between 20 and 500 Hz is generally preferred in the literature), rectified and the envelope of the signal is obtained with the help of a low-pass filter (the cut-off frequency can vary between 2-10 Hz, which can be created with active or passive circuit elements). Thus, the cEMG signal is obtained. This signal can now be applied to controllers such as microprocessors / microcontrollers.

[0069] - A proportional controller is designed using any embedded system controller (microprocessor, microcontroller, FPGA, etc.). The proportional controller used in the study works as follows: o With the comparison function to be performed in the controller, it is checked whether the cEMG signal amplitude is above the threshold value 5. o If the cEMG amplitude is not above the threshold value, no speed output is generated in the prosthesis, and a new surface EMG signal is obtained from the user, returning to the "cEMG signal acquisition" process step. o If the cEMG amplitude exceeds the threshold value 5, the prosthesis speed output is generated using the proportional controller. Proportional controller speed output v(t) is calculated by the following formula. In the formula, cEMGmaxdefines the maximum processed sEMG amplitude determined for the user, vmaxdefines the maximum speed at which the prosthesis can move, redefines the minimum speed at which the prosthesis can move, and 5 defines the threshold value. Vmax— vmin , v(t)=cEMG -max- X 8X cEMG(t) +vmin

[0070] - Using a device such as a data card (DAQ) or a digital analogue converter, the pre-recorded and processed cEMG signals are generated one by one (simulating the user) and applied to the prosthesis controlled by a proportional controller. The prosthesis-generated speed outputs of each data are recorded.

[0071] - The RMS values of the recorded speed outputs are calculated (formula on the previous page) and scaled from 0-1 .

[0072] - The scaled RMS values of the obtained surface EMG signals and speed outputs are mapped to each other (an example data set is shown in Figure 1 and Figure 2). A linear curve v=sEMG is fitted to the graph, and the distances of each data to this curve with respect to the y-axis are calculated and summed. This curve represents the ideal projection of the surface EMG signal energy onto the prosthesis speed output. The sum of the distances of the data from this curve represents the total speed deviation error.

[0073] The hybrid control method includes the following process steps;

[0074] - cEMG Signal Acquisition: The sEMG signals obtained from the user are applied to the signal processing and amplification circuit. The surface EMG signal from the user is amplified in this circuit (with active or passive amplifier circuits), bandpass filtered (active or passive bandpass filters can be used, a bandpass filter with cut-off frequencies between 20 and 500 Hz is generally preferred in the literature), rectified and the envelope of the signal is obtained with the help of a low-pass filter (the cut-off frequency can vary between 2-10 Hz, which can be created with active or passive circuit elements). Thus, the cEMG signal is obtained. This signal can now be applied to controllers such as microprocessors / microcontrollers.

[0075] With the comparison function to be performed in the controller, it is checked whether the cEMG signal amplitude is above the threshold value 5.

[0076] - If the cEMG amplitude is not above the threshold value, no speed output is generated in the prosthesis, and a new surface EMG signal is obtained from the user and processed, returning to the "cEMG Signal Acquisition" process step.

[0077] - If the cEMG amplitude is above the threshold value 5 and the cEMG amplitude p is above the deviation threshold value (how it is found will be explained later), the use of the on-off controller, i.e. rotation of the motor at full speed and movement of the prosthesis at full speed (i.e. movement with vmax) is performed.

[0078] - If the cEMG amplitude exceeds the threshold value 5 and the cEMG amplitude p is below the deviation threshold value, the prosthesis speed output is generated using the proportional controller. Proportional controller speed output v(t) is calculated by the following formula. In the formula, cEMGmaxdefines the maximum processed sEMG amplitude determined for the user, Vmax defines the maximum speed at which the prosthesis can move, redefines the minimum speed at which the prosthesis can move, and 5 defines the threshold value.

Claims

CLAIMS1. It is a hybrid control method characterised by the following processing steps:- Acquisition of the processed sEMG (cEMG) signal,- Checking whether the cEMG signal amplitude is above the threshold amplitude value 5,- Returning to the process step of obtaining the processed sEMG (cEMG) signal if the cEMG amplitude is not above the threshold value,- Using the on-off controller, if the cEMG amplitude is above the threshold amplitude value 5 and the cEMG amplitude p is above the deviation threshold value,- Generation of prosthesis speed output using a proportional controller if cEMG amplitude is above the threshold value 5 and cEMG amplitude deviation is below the threshold value.

2. The hybrid control method according to claim 1 , characterised in that the speed output of the proportional controller is calculated by the formula v(t) = cvE™M?G~mVatxn—inocx cEMG(t) + vminin the process step of generating the prosthesis speed output using the proportional controller if the cEMG amplitude is above the threshold value 5 and the cEMG amplitude deviation is below the threshold value.

3. The hybrid control method according to claim 1 , characterised in that the process step of obtaining the processed sEMG (cEMG) signal comprises the following process steps:- Application of the sEMG signal obtained from the user to the signal processing and amplification circuit,- Amplification of the surface EMG signal from the user with active or passive amplifier circuits- Obtaining the cEMG signal by passing it through active or passive bandpass filters, rectifying it and obtaining the signal's envelope with the help of a low-pass filter.

4. The hybrid control method according to claim 1 , characterised in that the process step of checking whether the cEMG signal amplitude 5 is above the threshold value comprises the following process steps:- Recording of surface EMG data from the user- Determination of a threshold value that separates noise and involuntary contraction from voluntary contraction by studying a series of repetitions of signals from the user.

Images