Wearable elbow joint angle acquisition device and elbow joint angle prediction method

Abstract

The invention relates to a wearable elbow joint angle acquisition device and an elbow joint angle prediction method.The elbow joint angle acquisition device integrates the synchronous acquisition function of electromyographic signals and elbow joint angles, is easy to operate and high in synchronism, has a wearable mode and a fixed mode and is wide in application range; according to the elbow joint angle prediction method provided by the invention, band-pass filtering, power frequency interference suppression and normalization processing are performed on the electromyographic signals after data acquisition, and the time domain, frequency domain and time-frequency characteristics are extracted, so that the sensitivity and prediction precision of the model to the angle change trend are improved; the regression model is adopted to integrate the electromyographic signals and the angle of the elbow joint, real-time prediction of the motion angle of the elbow joint can be achieved, and therefore the motion intention of a patient is decoded, and a basis is provided for personalized rehabilitation training.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation devices, specifically to a wearable elbow joint angle acquisition device and an elbow joint angle prediction method. Background Technology

[0002] Cerebrovascular diseases, severe traumatic brain injury, and other neurological diseases often lead to upper limb motor dysfunction in patients. Rehabilitation training is an important way to promote patient recovery. Surface electromyography signals can reflect the neural drive and activity state of muscles, while elbow joint angle reflects the kinematic characteristics of the upper limb. Combining the two can provide key evidence for motor intention recognition, rehabilitation assessment, and rehabilitation robot control.

[0003] Existing rehabilitation training devices have the following main shortcomings in signal acquisition: First, electromyography (EMG) acquisition devices and angle measurement devices are mostly used separately, which is complicated to operate and lacks synchronization, making it difficult to accurately correspond EMG signals and kinematic data; Second, some devices rely on laboratory supports, are bulky, and lack portability, limiting their application in home rehabilitation and daily monitoring; Third, existing methods often lack effective modeling of the mapping relationship between EMG signals and angles, making it impossible to accurately predict joint angles based on EMG signals. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a wearable elbow joint angle acquisition device and an elbow joint angle prediction method. It can simultaneously acquire electromyographic signals and elbow joint angles, model the mapping relationship between electromyographic signals and angles, and realize real-time prediction of joint angles. This enables the decoding of patients' movement intentions and provides a basis for personalized rehabilitation training.

[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: a wearable elbow joint angle acquisition device, comprising an upper arm fixation unit, a forearm fixation unit, an angle sensing unit, a connecting unit, and an electromyography (EMG) acquisition unit. The upper arm fixation unit is composed of an inner upper arm plate and an outer upper arm plate connected together, used to fix the user's upper arm; the forearm fixation unit is composed of an inner forearm plate and an outer forearm plate connected together, used to fix the user's forearm, and the end of the forearm fixation unit is provided with a grip for the user to grasp; the two ends of the connecting unit are respectively connected to the outer upper arm plate and the inner forearm plate, so that the upper arm fixation unit and the forearm fixation unit form a hinge structure, and the hinge axis is aligned with the rotation axis of the user's elbow joint; the angle sensing unit includes a single-axis angle sensor, which is coaxially arranged with the connecting unit, used to measure the user's elbow joint flexion and extension angle in real time; the EMG acquisition unit is worn on the user's upper arm and used to acquire the user's electromyography (EMG) signals.

[0006] Furthermore, the inner plate of the upper arm is designed to conform to the curved surface of the human upper arm, and the outer plate of the upper arm is provided with multiple sets of positioning holes. By selecting different hole positions to connect with the outer plate of the upper arm, the effective length of the upper arm fixing unit can be adjusted. The inner plate of the forearm has a sliding groove, and the fastener at one end of the outer plate of the forearm can slide and lock in the groove, realizing stepless adjustment of the length of the forearm fixing unit.

[0007] Furthermore, the outer plate of the upper arm is provided with strap holes for threading elastic straps, thereby directly fixing the angle acquisition device to the human upper limb, or installing the angle acquisition device on a fixed frame or rehabilitation robot platform by fasteners.

[0008] Furthermore, the upper arm fixation unit and the lower arm fixation unit are made of lightweight polylactic acid material using 3D printing technology.

[0009] The present invention also provides a method for predicting elbow joint angle using the above-mentioned wearable elbow joint angle acquisition device, comprising the following steps: Step 1: Device wearing. Secure the elbow joint angle acquisition device to the user's upper arm with a strap. At the same time, wear the MYO armband, which contains an 8-channel sEMG sensor, on the same user's upper arm. Step 2: Dual-modal data acquisition. The user's upper arm performs flexion and extension movements from 0° to 150° in a natural state. In the electromyography channel, multi-channel electromyography signals on the surface of the upper arm are acquired through an sEMG sensor. In the kinematics channel, the elbow joint movement angle is acquired through a single-axis angle sensor of an elbow joint angle acquisition device. Step 3: Data preprocessing. In the electromyography channel, the acquired electromyography signals are preprocessed by filtering, denoising, and alignment. In the kinematics channel, the acquired joint angle data are cropped and aligned to align the joint angle data with the time axis of the electromyography signals, ensuring data synchronization. Step 4: Feature extraction. Short-time Fourier transform is used to extract the time-frequency feature spectrum of the preprocessed electromyography signal. Step 5: Model training and prediction. The extracted time-frequency feature spectrum is enhanced by increasing the numerical distribution of the data with a fixed ratio to improve feature discriminability. Then, the time-frequency features and the processed joint angle signal are input into the Scale-Spatiotemporal Channel Cross Encoder (STCCE) for training and prediction. Finally, the predicted elbow joint angle is output.

[0010] Furthermore, in step three, the preprocessing of the electromyographic signal includes using a 50Hz notch filter to eliminate power frequency interference and using a fourth-order Butterworth high-pass filter to filter out baseline drift.

[0011] Furthermore, in step five, the spatiotemporal channel cross encoder performs end-to-end training by minimizing the error between the predicted angle and the true angle.

[0012] Furthermore, in step five, the fixed-proportion enhanced numerical distribution includes normalization and standardization processes.

[0013] Furthermore, in step five, the spatiotemporal channel cross encoder first scales the input features, then extracts the temporal features of each channel through a temporal attention mechanism, and then uses a cross-channel attention mechanism to fuse multi-channel information, and inputs the fused multi-channel features into a fully convolutional neural network (FCN).

[0014] Furthermore, the fully convolutional neural network (FCN) extracts spatial features through convolutional layers, compresses dimensions through pooling layers, and finally outputs continuous elbow angle predictions through regression layers.

[0015] Beneficial effects: 1. The elbow joint angle acquisition device of the present invention is designed with an exoskeleton structure suitable for the upper limb and integrates the synchronous acquisition function of electromyographic signals and elbow joint angle. It avoids the signal misalignment problem caused by inconsistent timestamps when the two signals are acquired separately. It is simple to operate, has strong synchronization, and ensures the accurate correspondence between the input and output of the subsequent angle prediction model.

[0016] 2. The elbow joint angle acquisition device of the present invention uses a 3D printed integrated exoskeleton-like support with an adjustable length design, which can be adaptively adjusted according to user needs to ensure that the user's elbow joint rotation center is strictly aligned with the angle sensor, thereby reducing the angle measurement error caused by sensor position offset.

[0017] 3. The elbow joint angle acquisition device of the present invention has a simple and lightweight structure and has two usage modes: fixed and wearable. It can be installed on a bracket in the laboratory to realize standardized motion acquisition, and can also be worn directly on the user's upper limb in the daily environment by means of a strap, which can meet the needs of home rehabilitation and long-term monitoring. It has a wide range of applications and low cost.

[0018] 4. Based on the aforementioned elbow joint angle acquisition device, this invention proposes an angle prediction method based on electromyographic signal feature input and regression model output. After data acquisition, the electromyographic signals are subjected to bandpass filtering, power frequency interference suppression, and normalization processing, and time-domain, frequency-domain, and time-frequency features are extracted, improving the model's sensitivity to angle change trends and prediction accuracy. The regression model integrates electromyographic signals and elbow joint angles, enabling real-time prediction of elbow joint movement angles, thereby decoding the patient's movement intentions and providing a basis for personalized rehabilitation training. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall structure of the elbow joint angle acquisition device; Figure 2This is a schematic diagram of the outer plate structure of the boom; Figure 3 This is a schematic diagram of the inner plate structure of the forearm; Figure 4 This is a flowchart of the data acquisition process of the present invention; Figure 5 This is a flowchart of the elbow joint angle prediction method of the present invention.

[0020] Figure reference numerals: 1. Inner plate of boom, 2. Outer plate of boom, 3. Angle sensor, 4. Coupling, 5. Inner plate of forearm, 6. Outer plate of forearm, 7. Hand grip, 8. Strap hole, 9. Multiple positioning holes, 10. Angle sensor mounting hole, 11. Coupling mounting hole, 12. Slide groove. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0022] This embodiment provides a wearable elbow joint angle acquisition device for simultaneously acquiring surface electromyography signals of the human upper limb and elbow joint angle signals; specifically, as shown in the example... Figure 1-3 As shown, the wearable elbow joint angle acquisition device includes an upper arm fixation unit, a forearm fixation unit, an angle sensing unit, a connection unit, and an electromyography signal acquisition unit.

[0023] The upper arm fixation unit consists of an inner upper arm plate 1 and an outer upper arm plate 2, used to fix the user's upper arm. The inner upper arm plate 1 is connected inside the outer upper arm plate 2 and is designed to conform to the curved surface of the user's upper arm. The forearm fixation unit consists of an inner forearm plate 5 and an outer forearm plate 6, used to fix the user's forearm. The end of the forearm fixation unit is equipped with a grip 7 for the user to grasp. The angle sensing unit includes a single-axis angle sensor 3, used to measure the elbow joint flexion and extension angle in real time. The connecting unit adopts a coupling 4, with both ends of the coupling 4 fixedly connected to the outer upper arm plate 2 and the inner forearm plate 5, so that the outer upper arm plate 2 and the inner forearm plate 5 form a hinge structure through the coupling 4, and the hinge axis is kept consistent with the rotation axis of the user's elbow joint. The single-axis angle sensor 3 is coaxial with the hinge axis of the outer upper arm plate 2 and the inner forearm plate 5, thereby ensuring accurate measurement of the elbow joint angle. The electromyography signal acquisition unit is a MYO arm ring, which realizes human-computer interaction by detecting the bioelectric signals generated by the user's arm muscle movement.

[0024] Furthermore, the device is also equipped with a length adjustment mechanism, and the outer plate 2 of the main arm has the following structure: Figure 2 The outer plate 2 of the boom has multiple sets of positioning holes 9. By selecting different hole positions to connect with the inner plate 1 of the boom, the effective length of the boom fixing unit can be adjusted. The bottom end of the outer plate 2 of the boom also has an angle sensor mounting hole 10. The structure of the inner plate 5 of the forearm is as follows: Figure 3The upper end of the inner plate 5 of the forearm is provided with a coupling mounting hole 11, and the lower end of the inner plate 5 of the forearm is provided with a sliding groove 12. The fastener at one end of the outer plate 6 of the forearm can slide and lock in the groove, so as to realize the stepless adjustment of the length of the forearm fixing unit.

[0025] In addition, the outer plate 2 of the upper arm is provided with strap holes 8 for threading elastic straps so that the device can be directly fixed to the upper limb of the human body. In the laboratory fixation mode, the device can also be installed on a fixed frame or rehabilitation robot platform by fasteners to enhance its applicability in different application scenarios.

[0026] The upper arm fixation unit and forearm fixation unit of the present invention are made of lightweight polylactic acid material using 3D printing technology. This lightweight exoskeleton structure is easy to wear and reduces the burden on the user. It also has an adjustable length design, which can be adapted to the user's needs to ensure that the user's elbow joint rotation center is strictly aligned with the angle sensor.

[0027] The upper arm fixation unit adopts a combination of an inner upper arm plate 1 and an outer upper arm plate 2, which can not only fit the user's upper arm for reliable fixation, but also ensure that there is enough space between the outer upper arm plate 2 and the user's upper arm to facilitate the wearing of the electromyography signal acquisition unit.

[0028] Based on the aforementioned elbow joint angle acquisition device, this embodiment provides an elbow joint angle prediction method, the data acquisition process of which is as follows: Figure 4 As shown, firstly, the MYO armband is worn on the user's upper arm. Simultaneously, the connection lengths of the upper arm fixing unit and the forearm fixing unit of the elbow joint angle acquisition device are adjusted according to the user's needs. The upper arm outer plate 2 is fixed to the user's upper arm via straps, and the user's hand is positioned to grip the hand 7. The MYO armband has a built-in 8-channel sEMG sensor, which acquires multi-channel surface electromyography signals at a sampling frequency of 200Hz. Simultaneously, a single-axis angle sensor 3 acquires the true value signal of the elbow joint angle at the same sampling frequency, establishing a synchronous data acquisition system. The acquired data is preprocessed and then... Figure 5 The elbow joint angle prediction method based on surface electromyography signals is shown in the figure.

[0029] The elbow joint angle prediction method mainly uses the STCCE network model mapping to achieve real-time prediction of elbow joint movement angle, and specifically includes the following steps.

[0030] Dual-modal data acquisition: After the device is worn, the user's upper limb performs flexion and extension movements from 0° to 150° in a natural state. In the electromyography channel, multi-channel electromyography signals on the surface of the upper arm are acquired through an sEMG sensor. In the kinematics channel, the elbow joint movement angle is acquired through a single-axis angle sensor of the elbow joint angle acquisition device.

[0031] Data preprocessing: In the electromyography (EMG) channel, the acquired EMG signals are preprocessed through filtering, denoising, and alignment. Specifically, the preprocessing of the EMG signals includes using a 50Hz notch filter to eliminate power frequency interference and using a fourth-order Butterworth high-pass filter to filter out baseline drift. In the kinematics channel, the acquired joint angle data are cropped and aligned to align the joint angle data with the time axis of the EMG signals, ensuring data synchronization.

[0032] Feature extraction: Short-time Fourier transform is used to extract time-frequency feature spectra of the preprocessed electromyographic signal. This method can simultaneously analyze the "time-frequency" distribution of the signal and capture the dynamic changes of muscle contraction.

[0033] For model training and prediction, the extracted time-frequency feature spectrum is enhanced by increasing the numerical distribution by a fixed ratio (such as normalization and standardization) to improve feature distinguishability. Then, the time-frequency features and the processed joint angle signal are input into the Scale-Spatiotemporal Channel Cross Encoder (STCCE) for training and prediction, and finally the predicted elbow joint angle is output.

[0034] This spatiotemporal channel cross-encoder network structure first scales the input features, then extracts the temporal features of each channel through a temporal attention mechanism, and then fuses the multi-channel information using a cross-channel attention mechanism. The fused multi-channel features are then input into a fully convolutional neural network (FCN). The FCN extracts spatial features through convolutional layers, compresses dimensions through pooling layers, and finally outputs continuous elbow angle prediction values ​​through a regression layer.

[0035] The model is trained end-to-end by minimizing the error between the predicted angle and the true angle, achieving an accurate mapping from electromyographic signals to joint angles.

[0036] The elbow joint angle prediction method of this invention achieves efficient fusion of multimodal information of "electromyography + kinematics" through a closed loop of "dual-modal data acquisition → preprocessing → feature extraction → enhancement and mapping → attention encoding → cross-channel fusion → deep learning prediction", and finally accurately predicts the elbow joint angle. It can be applied to rehabilitation training assessment, sports biomechanical analysis, human-computer interaction (such as motion intention recognition of MYO armband devices) and other scenarios.

[0037] This invention collects multi-channel surface electromyography (EMG) signals from the upper arm and combines them with data synchronously collected by an elbow joint angle sensor. Through processing steps such as filtering, normalization, and feature extraction, a mapping relationship between EMG signals and joint angles is established, thereby enabling real-time prediction of elbow joint angles. It fully utilizes the time-frequency characteristics and spatial distribution features of EMG signals, and through the effective fusion of multi-channel information, achieves high-precision continuous prediction of elbow joint motion angles, providing an effective motion intent decoding solution for upper limb rehabilitation robots and intelligent prosthesis control.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A wearable elbow joint angle acquisition device, characterized in that, The device includes an upper arm fixation unit, a lower arm fixation unit, an angle sensing unit, a connecting unit, and an electromyography (EMG) acquisition unit. The upper arm fixation unit is composed of an inner upper arm plate (1) and an outer upper arm plate (2) connected together to fix the user's upper arm. The lower arm fixation unit is composed of an inner lower arm plate (5) and an outer lower arm plate (6) connected together to fix the user's lower arm. The lower arm fixation unit has a hand grip (7) at its end for the user to grasp. The two ends of the connecting unit are connected to the outer upper arm plate (2) and the inner lower arm plate (5) respectively, so that the upper arm fixation unit and the lower arm fixation unit form a hinge structure and the hinge axis is aligned with the user's elbow joint rotation axis. The angle sensing unit includes a single-axis angle sensor (3), which is coaxially arranged with the connecting unit and is used to measure the user's elbow joint flexion and extension angle in real time. The EMG acquisition unit is worn on the user's upper arm and is used to collect the user's upper arm EMG signals. 2.The wearable elbow joint angle acquisition device according to claim 1, characterized in that, The inner plate (1) of the upper arm is designed to fit the curved surface of the human upper arm. The outer plate (2) of the upper arm is provided with multiple sets of positioning holes (9). The inner plate (1) of the upper arm can be connected to the outer plate (2) of the upper arm by selecting different hole positions, so as to adjust the effective length of the upper arm fixing unit. The inner plate of the forearm has a sliding groove (12). The fastener at one end of the outer plate (6) of the forearm can slide and lock in the groove, so as to realize the stepless adjustment of the length of the forearm fixing unit. 3.The wearable elbow joint angle acquisition device according to claim 1, characterized in that, The outer plate (2) of the upper arm is provided with strap holes (8) for threading elastic straps, so as to directly fix the angle acquisition device to the upper limb of the human body, or install the angle acquisition device on a fixed frame or rehabilitation robot platform by fasteners.

4. The wearable elbow joint angle acquisition device according to claim 1, characterized in that, The upper arm fixation unit and the lower arm fixation unit are made of lightweight polylactic acid material using 3D printing technology.

5. A method for predicting the angle of the elbow joint using the wearable elbow joint angle acquisition device according to any one of claims 1 to 4, characterized in that, Includes the following steps: Step 1: Device wearing. Secure the elbow joint angle acquisition device to the user's upper arm with a strap. At the same time, wear the MYO armband, which contains an 8-channel sEMG sensor, on the same user's upper arm. Step 2: Dual-modal data acquisition. The user's upper arm performs flexion and extension movements from 0° to 150° in a natural state. In the electromyography channel, multi-channel electromyography signals on the surface of the upper arm are acquired through an sEMG sensor. In the kinematics channel, the elbow joint movement angle is acquired through a single-axis angle sensor (3) of the elbow joint angle acquisition device. Step 3: Data preprocessing. In the electromyography channel, the acquired electromyography signals are preprocessed by filtering, denoising, and alignment. In the kinematics channel, the acquired joint angle data are cropped and aligned to align the joint angle data with the time axis of the electromyography signals, ensuring data synchronization. Step 4: Feature extraction. Short-time Fourier transform is used to extract the time-frequency feature spectrum of the preprocessed electromyography signal. Step 5: Model training and prediction. The extracted time-frequency feature spectrum is enhanced by increasing the numerical distribution of the feature by a fixed ratio to improve the feature distinguishability. Then, the time-frequency features and the processed joint angle signal are input into the Scale-Time-Channel Cross Encoder (STCCE) for training and prediction. Finally, the predicted elbow joint angle is output.

6. The elbow joint angle prediction method according to claim 5, characterized by, In step three, the preprocessing of the electromyographic signal includes using a 50Hz notch filter to eliminate power frequency interference and using a fourth-order Butterworth high-pass filter to filter out baseline drift.

7. The elbow angle prediction method according to claim 5, characterized by, In step five, the spatiotemporal channel cross encoder performs end-to-end training by minimizing the error between the predicted angle and the true angle.

8. The elbow angle prediction method according to claim 5, characterized by, In step five, the fixed-proportion enhanced numerical distribution includes normalization and standardization processes.

9. The elbow angle prediction method according to claim 5, characterized by, In step five, the spatiotemporal channel cross encoder first scales the input features, then extracts the temporal features of each channel through a temporal attention mechanism, and then uses a cross-channel attention mechanism to fuse multi-channel information. The fused multi-channel features are then input into a fully convolutional neural network (FCN).

10. The elbow angle prediction method according to claim 9, characterized by, In step five, the fully convolutional neural network (FCN) extracts spatial features through convolutional layers, compresses dimensions through pooling layers, and finally outputs continuous elbow angle prediction values ​​through regression layers.

Images