A split hip-ankle assistive exoskeleton robot and an intention recognition control method thereof

Abstract

The present application relates to the technical field of exoskeleton robot equipment, in particular to a separated hip-ankle assisting exoskeleton robot and an intention recognition control method thereof, the robot comprises a hip joint exoskeleton assembly and an ankle joint exoskeleton assembly; the hip joint exoskeleton assembly is worn between the waist and the thigh of a human body; the ankle joint exoskeleton assembly is worn on the foot of a human body; the ankle joint exoskeleton assembly comprises a top plate, a bottom plate, an energy storage deformation element and a displacement sensor. The method comprises intention recognition of each gait phase under flat ground conditions, going upstairs conditions and going downstairs conditions, and placing corresponding gait data models into the hip joint exoskeleton assembly to assist the hip joint exoskeleton assembly in making upper intention recognition control and the like. The present application uses the ankle joint exoskeleton to make human gait phase intention recognition, and assists the hip joint exoskeleton to make upper intention recognition control, thereby providing more assisting force benefits to the human body under the premise of ensuring the freedom degree of human movement.

Description

Technical Field

[0001] This invention relates to the field of exoskeleton robot equipment technology, and in particular to a detachable hip and ankle-assisted exoskeleton robot and its intention recognition and control method. Background Technology

[0002] Currently, exoskeletons have garnered significant attention in the medical, industrial, and military fields. In the medical field, exoskeletons can aid patient recovery. In the industrial field, they can assist workers, reducing injuries and chronic strain. In the military field, they can provide assistance and reduce metabolic burdens for soldiers, enhancing their combat capabilities. They hold great potential in military, rehabilitation, and industrial applications. However, the number of truly practical exoskeletons, especially assistive exoskeletons, is currently limited. The main reason for the poor practicality of most exoskeletons is the human-machine integration problem, a major challenge in the field. This problem arises from a mismatch between human and machine force application, causing the exoskeleton to act as resistance in many scenarios, particularly for exoskeletons requiring high-speed response assistance. To enable exoskeletons to truly provide assistance in various states, the core is to develop a robust human-machine integration control model.

[0003] Exoskeletons are typically divided into upper-level control, middle-level control, and lower-level control. Upper-level control involves the exoskeleton's intent recognition, while middle-level control is the dynamic model. Only by effectively controlling the upper-level intent recognition, allowing the exoskeleton to accurately identify the human's intent, can the correct dynamic model be output. A crucial step in upper-level control is gait phase recognition, which is further subdivided into the support phase and the swing phase. Providing dynamic control models for the support phase when both legs are in the support phase and for the swing phase when both legs are in the swing phase are essential for ensuring human-machine integration. Furthermore, exoskeletons involve prior control and posterior control.

[0004] Currently, commonly used intention recognition sensor methods include prior and posterior signals. Prior signals refer to signals from the subject's musculoskeletal and nervous systems, such as muscle flexibility, musculoskeletal signals, and EEG signals. The existence of prior signals means that each subject's human signals often vary greatly, and this individual difference makes it difficult to establish a unified model for intention recognition and interpret the signals. Posterior signals refer to signals that do not contain the subject's direct intention but can be used to infer the subject's movement intention through algorithms and models, such as inertial measurement units (IMUs) and plantar pressure sensors. Generally, an IMU is used in conjunction with a plantar pressure sensor to measure the gait phase. Currently, commonly used plantar pressure sensors include FSR plantar pressure sensors and resistive strain gauge pressure sensors. However, FSR plantar pressure sensors suffer from insufficient sensitivity and accuracy, making them unsuitable for practical engineering applications. Resistive strain gauge pressure sensors offer higher accuracy but often require complex foot fixture interfaces and are prone to damage and malfunction due to prolonged contact with the soles of the feet.

[0005] Currently, full-body exoskeletons are more practical for rehabilitation applications. However, assistive exoskeletons, due to the added weight and restrictions on the body's freedom of movement, as well as their high complexity of control, often bring more disadvantages than benefits to assisting normal walking. Existing ankle-assistive exoskeletons often use laces to connect to the user's shoe, resulting in poor connectivity and practicality. Furthermore, no ankle exoskeletons capable of gait phase recognition have been reported. Summary of the Invention

[0006] To address the aforementioned problems, the present invention aims to provide a separate hip and ankle-assisted exoskeleton robot and its intention recognition and control method. The exoskeleton of the ankle joint is used to perform human gait phase intention recognition, and the exoskeleton of the hip joint is used to perform upper-level intention recognition and control, thereby providing greater assistance to the human body while ensuring the human body's freedom of movement.

[0007] The technical solution adopted in this invention is as follows:

[0008] This invention proposes a split hip and ankle-assisted exoskeleton robot, comprising a hip exoskeleton assembly and an ankle exoskeleton assembly. The hip exoskeleton assembly is worn between the lower back and thigh of a human body; the ankle exoskeleton assembly is worn on the foot of a human body. The ankle exoskeleton assembly includes a top plate, a bottom plate, energy storage deformation elements, and a displacement sensor. Both the top plate and the bottom plate have shapes corresponding to the foot. The bottom plate is positioned directly below the top plate, and the four corner areas of the top plate and the bottom plate are connected by energy storage deformation elements. A toe limiting plate is provided on the front side of the upper top surface of the top plate; a heel limiting plate is provided on the rear side of the upper top surface of the top plate. The displacement sensor is fixed to the upper rear surface of the heel limiting plate, with its working end pointing vertically downward. It is used to measure the distance between the upper and lower ankle exoskeletons, and gait phase prediction is achieved through the measurement data of the displacement sensor. The gait phase prediction results assist the hip exoskeleton assembly in performing upper-level intention recognition and control.

[0009] Furthermore, the energy storage deformation element is a spring.

[0010] Furthermore, a hollow screw is provided inside the spring; the upper and lower ends of the hollow screw are respectively fixed vertically to the top plate and the bottom plate by bolts.

[0011] An intention recognition and control method for a split hip-ankle-assisted exoskeleton robot, the method comprising the following steps:

[0012] S1. Displacement sensor data acquisition and processing: This includes the acquisition and processing of displacement sensor data under three conditions: flat ground, going upstairs, and going downstairs.

[0013] S2. After data processing, a gait data model containing various working conditions is obtained;

[0014] S3. The gait data model is put into the control chip of the hip joint component, so that the hip joint exoskeleton component can make corresponding intention recognition control based on the data of the foot displacement sensors.

[0015] Furthermore, under flat ground conditions, displacement sensor data of the heels of both feet were collected by a data acquisition device under three gait phases: "right foot supporting left foot swinging", "both feet supporting", and "left foot supporting right foot swinging". After normalizing the data, an SVM model was used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the three gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions in the training set and the test set for the three gait phases.

[0016] Furthermore, under the condition of going upstairs, displacement sensor data of the heels of both feet were collected by the acquisition device under four gait phases: "right foot forward with both feet supporting", "right foot supporting with left foot swinging", "left foot forward with both feet supporting", and "left foot supporting with right foot swinging". After normalizing the data, the SVM model was used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the four gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions in the training set and the test set for the four gait phases.

[0017] Furthermore, under the condition of going downstairs, displacement sensor data of the heels of both feet were collected by the acquisition device under four gait phases: "right foot forward with both feet supporting", "right foot supporting with left foot swinging", "left foot forward with both feet supporting", and "left foot supporting with right foot swinging". After normalizing the data, the SVM model was used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the four gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions in the training set and the test set for the four gait phases.

[0018] Furthermore, step S3 includes: placing the SVM model obtained in step S2, which includes three working conditions, into the control chip of the hip exoskeleton component; communicating the real-time data collected by the foot displacement sensors to the control chip of the hip exoskeleton component; and matching the gait phase data in the SVM model with the displacement sensor data to make corresponding upper-level intention recognition control actions.

[0019] Furthermore, the acquisition device includes a power supply, a signal conversion element, and a host computer; the host computer is connected to the displacement sensor in sequence through the signal conversion element and the power supply, and is used to acquire and read displacement sensor data; the power supply is connected to the displacement sensor and is used to supply power to it.

[0020] Furthermore, the signal conversion element is a 485 interface to USB interface element.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. This invention designs a practical split hip and ankle exoskeleton. While providing assistance, the ankle exoskeleton also performs gait behavior recognition, thus providing upper-level intention recognition and control for the hip exoskeleton. The designed ankle exoskeleton adopts an integrated design, adding a screw to the spring deformation energy storage component to fix the spring deformation direction, stabilizing the direction of the exoskeleton's energy storage rebound force and solving the problem of ankle sprains common with existing exoskeletons. The integrated design also allows for flexible and snug wear at the point of contact with the user. Furthermore, a displacement sensor is incorporated into the ankle exoskeleton to measure the distance between the upper and lower ankle exoskeletons. By collecting these features, gait phase prediction is achieved.

[0023] 2. The classification accuracy of key gait phases in three common working conditions—flat ground, going upstairs, and going downstairs—was verified. The accuracy of the training and test sets for each working condition can reach over 99%.

[0024] 3. Currently, posterior sensor signals based on physical signals often experience a time lag during the "acquisition → transmission → calculation → execution" process. This delay causes control commands to "lag behind the actual system state," resulting in slow system response. This invention utilizes the delay effect of the ankle exoskeleton's ground contact energy storage process to detect the foot's intention to touch the ground in advance. Through energy storage and rebound deformation, it can detect the intention to lift the foot in advance, achieving advanced prediction and recognition of the foot's ground contact and lifting intervals, thereby improving the accuracy of upper-level control based on intention recognition. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the structure of a split hip and ankle-assisted exoskeleton robot proposed in this invention;

[0026] Figure 2 for Figure 1 Schematic diagram of the mid-ankle joint assembly;

[0027] Figure 3 This is a schematic diagram of the connection structure between the ankle joint assembly and the acquisition device of the present invention;

[0028] Figure 4 This is a schematic diagram illustrating the accuracy of the training set classification for flat terrain conditions in this invention.

[0029] Figure 5 This is a schematic diagram illustrating the accuracy of the flat terrain test set classification in this invention.

[0030] Figure 6 This is a schematic diagram of the confusion matrix of the training set for flat terrain conditions in this invention;

[0031] Figure 7 This is a schematic diagram of the confusion matrix of the flat ground test set of the present invention;

[0032] Figure 8 This is a schematic diagram illustrating the accuracy of the training set classification for the upstairs working conditions in this invention.

[0033] Figure 9 This is a schematic diagram illustrating the accuracy of the upstairs working condition test set classification in this invention;

[0034] Figure 10 This is a schematic diagram of the confusion matrix of the training set for the upstairs working conditions of this invention;

[0035] Figure 11 This is a schematic diagram of the confusion matrix of the test set for the upstairs working conditions of this invention;

[0036] Figure 12 This is a schematic diagram illustrating the accuracy of the training set classification for the downstairs working conditions in this invention.

[0037] Figure 13 This is a schematic diagram illustrating the accuracy of the downstairs working condition test set classification in this invention;

[0038] Figure 14 This is a schematic diagram of the confusion matrix of the training set for the downstairs working conditions of this invention;

[0039] Figure 15 This is a schematic diagram of the confusion matrix of the downstairs working condition test set of the present invention.

[0040] The attached figures are labeled as follows: 1-1, hip exoskeleton assembly; 1-2, ankle exoskeleton assembly; 2-1, top plate; 2-2, bottom plate; 2-3, energy storage deformation element; 2-4, hollow screw; 2-5, bolt A; 2-6, displacement sensor; 2-7, nut; 2-8, bolt B; 2-9, toe limiting plate; 2-10, heel limiting plate; 3-1, power supply equipment; 3-2, signal conversion element; 3-3, host computer. Detailed Implementation

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] It should be noted that in the description of this invention, the terms "upper", "lower", "top", "bottom", "one side", "the other side", "left", "right", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not mean that the device or element must have a specific orientation, or be constructed and operated in a specific orientation.

[0043] See appendix Figure 1-2The present invention proposes a split hip and ankle-assisted exoskeleton robot, including a hip joint exoskeleton component 1-1 and an ankle joint exoskeleton component 1-2; the hip joint exoskeleton component 1-1 can be an existing hip joint exoskeleton, which is worn between the human waist and back and the thigh; the ankle joint exoskeleton component 1-2 is worn on the human foot.

[0044] The ankle exoskeleton assembly 1-2 includes a top plate 2-1, a bottom plate 2-2, an energy storage deformation element 2-3, and a displacement sensor 2-6. The bottom plate 2-2 is positioned directly below the top plate 2-1. Both are shaped to correspond to the foot, and the four corner areas of the top plate 2-1 and the bottom plate 2-2 are connected by the energy storage deformation element 2-3, which is a spring. When the human body touches the ground, the top plate 2-1 moves downward, the energy storage deformation element 2-3 is compressed and deformed to store energy, and the value measured by the displacement sensor 2-6 also changes.

[0045] In this embodiment, a hollow screw 2-4 is provided inside the spring; the upper and lower ends of the hollow screw 2-4 are respectively fixed perpendicularly to the top plate 2-1 and the bottom plate 2-2 by bolts A2-5. The hollow screw 2-4 can ensure the stability of the direction of the rebound assist after energy storage.

[0046] A toe limiting plate 2-9 is provided on the front side of the upper end face of the top plate 2-1; a heel limiting plate 2-10 is provided on the rear side of the upper end face of the top plate 2-1; the displacement sensor 2-6 is fixed to the upper part of the rear end face of the heel limiting plate 2-10 by bolt B2-8 and nut 2-7, and its working end is vertically downward; the displacement sensor 2-6 is connected to the control chip of the hip joint exoskeleton component 1-1 through wired or Bluetooth communication, so that the data of the displacement sensor 2-6 can be fed back in real time to assist the hip joint exoskeleton component 1-1 in performing upper-level intention recognition and control.

[0047] An intention recognition and control method for a split hip-ankle-assisted exoskeleton robot, the method comprising the following steps:

[0048] S1. Displacement sensor data acquisition and processing: This includes the acquisition and processing of displacement sensor data under three conditions: flat ground, going upstairs, and going downstairs.

[0049] Under flat terrain conditions, displacement data from sensors 2-6 at the heel of both feet were collected in three gait phases: "right foot support with left foot swing," "both feet support," and "left foot support with right foot swing." 10,000 data points were collected for each gait phase, totaling 30,000 data points. These 30,000 data points were then categorized and labeled accordingly. The training set comprised 70% of the total data, and the test set comprised 30%. After data normalization, an SVM model was used for recognition and classification. The accuracy rates for the training and test sets were obtained, such as... Figure 4 and Figure 5 As shown, the confusion matrices of the training and test sets with respect to the three gait phases are obtained, as follows: Figure 6 and Figure 7 As shown, the confusion matrix will specifically yield the number of correct and incorrect recognitions for the training and test sets for the three gait phases.

[0050] The data acquisition for flat ground conditions includes phase acquisition with both feet supported (A1, C1), phase acquisition with the right foot supported and the left foot swinging (B1), and phase acquisition with the left foot supported and the right foot swinging (D1).

[0051] Under the condition of climbing stairs, displacement data of the heels of both feet were collected by a data acquisition device under four gait phases: "right foot forward, both feet supporting," "right foot supporting, left foot swinging," "left foot forward, both feet supporting," and "left foot supporting, right foot swinging." 10,000 data points were collected for each gait phase, for a total of 40,000 data points. These 40,000 data points were then categorized and labeled accordingly. The training set comprised 70% of the total data, and the test set comprised 30%. After data normalization, an SVM model was used for recognition and classification. The recognition accuracy of the training and test sets was obtained, such as... Figure 8 and Figure 9 As shown, the confusion matrices of the training and test sets with respect to the four gait phases are obtained, as follows: Figure 10 and Figure 11 As shown, the confusion matrix will specifically yield the number of correct and incorrect recognitions for the training and test sets for the four gait phases.

[0052] The data collection for the upstairs working conditions includes phase A3 with the right foot in front and both feet supporting, phase B3 with the right foot supporting and the left foot swinging, phase C3 with the left foot in front and both feet supporting, and phase D3 with the left foot supporting and the right foot swinging.

[0053] Under the condition of descending stairs, displacement data of the heels of both feet were collected by a data acquisition device under four gait phases: "right foot forward, both feet supporting," "right foot supporting, left foot swinging," "left foot forward, both feet supporting," and "left foot supporting, right foot swinging." 10,000 data points were collected for each gait phase, totaling 40,000 data points. These 40,000 data points were then categorized and labeled accordingly. The training set comprised 70% of the total data, and the test set comprised 30%. After data normalization, an SVM model was used for recognition and classification. The recognition accuracy of the training and test sets was obtained, such as... Figure 12 and Figure 13 As shown, the confusion matrices of the training and test sets with respect to the four gait phases are obtained, as follows: Figure 14 and Figure 15 As shown, the confusion matrix will specifically yield the number of correct and incorrect recognitions for the training and test sets for the four gait phases.

[0054] The data collection for the downstairs operation includes phase A4 with the right foot forward and both feet supporting, phase B4 with the right foot supporting and the left foot swinging, phase C4 with the left foot forward and both feet supporting, and phase D4 with the left foot supporting and the right foot swinging.

[0055] Among them, such as Figure 3 As shown, the acquisition device includes a power supply device 3-1, a signal conversion element 3-2, and a host computer 3-3. The host computer 3-3 is connected to the displacement sensor 2-6 in sequence through the signal conversion element 3-2 and the power supply device 3-1, and is used to acquire and read data from the displacement sensor 2-6. The signal conversion element 3-2 is a 485 interface to USB interface element. The power supply device 3-1 is connected to the displacement sensor 2-6 and is used to supply power to it.

[0056] S2. After data processing, a gait data model containing various working conditions is obtained;

[0057] S3. The gait data model is placed into the control chip of the hip joint component, so that the hip joint exoskeleton component can make corresponding intention recognition control based on the data of the foot displacement sensors: The SVM model containing three working conditions obtained in step S2 is placed into the control chip of the hip joint exoskeleton component 1-1. The real-time data collected by the foot displacement sensors 2-6 is communicated to the control chip of the hip joint exoskeleton component 1-1. The control chip of the hip joint exoskeleton component 1-1 matches the gait phase data in the SVM model with the data collected by the displacement sensors 2-6, so as to make corresponding upper-level intention recognition control actions based on the gait phase.

[0058] In this invention, the energy storage deformation element 2-3 of the ankle exoskeleton component 1-2 is mounted on the hollow screw 2-4 of the designed ankle exoskeleton component 1-2. Compared with other ankle exoskeleton solutions, this solution has a stable buffering and assist direction and is equipped with a displacement sensor 2-6. When the human foot contacts the ground, due to the human body's weight and impact force, the energy storage deformation element 2-3 of the ankle exoskeleton component 1-2 deforms and stores energy, causing the top plate 2-1 to shift. At this time, the displacement characteristics of the top plate 2-1 can be detected, thereby detecting the design principle of foot contact and lifting. This realizes the beneficial effect of 1+1 greater than 2 by combining the displacement sensor 2-6 with the energy storage deformation characteristics. At the same time, the designed intention recognition control method and machine learning are used to solve and verify the exoskeleton gait phase recognition, with an accuracy rate of over 99%. Meanwhile, another innovation and benefit of this solution is that, since the ankle exoskeleton components 1-2 have a large time interval in the energy storage deformation stage and the energy release stage, the threshold method can be used to detect the intention of the human foot to touch the ground and lift up in advance, thereby achieving prior recognition by the physical sensor, solving the problem of signal delay of the physical sensor, and improving the error tolerance of recognition.

[0059] This solution effectively addresses the critical issue of accurate gait phase recognition in the exoskeleton robot industry, achieving, for the first time, a priori recognition method using physical sensors. Compared to priori recognition by biosensors such as EMG and EEG, this solution offers a significant accuracy advantage. It also demonstrates superior performance compared to piezoresistive (low precision) and capacitive (complex tooling leading to ankle sprains) foot pressure sensors, which are typically placed on the soles of the feet and are prone to wear and tear over time. Furthermore, it resolves the latency issues associated with IMU sensors that detect foot contact through vertical acceleration changes. This invention offers substantial advantages in practicality, reliability, quality, and accuracy, demonstrating promising application prospects. It solves a major problem in accurate gait phase recognition within the exoskeleton industry, providing precise recognition for future upper-level control of hip exoskeletons and other similar devices.

[0060] Matters not covered in this invention are common knowledge.

[0061] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A detachable hip-ankle assisted exoskeleton robot, comprising a hip joint exoskeleton component and an ankle joint exoskeleton component; the hip joint exoskeleton component is worn between the lower back and thigh of a human body; the ankle joint exoskeleton component is worn in the foot of a human body; characterized in that: The ankle exoskeleton assembly includes a top plate, a bottom plate, energy storage deformation elements, and a displacement sensor. Both the top plate and the bottom plate are shaped to correspond to the foot. The bottom plate is positioned directly below the top plate, and the four corner areas of the top and bottom plates are connected by energy storage deformation elements. A toe-limiting plate is provided on the front side of the upper top surface of the top plate, and a heel-limiting plate is provided on the rear side of the upper top surface of the top plate. The displacement sensor is fixed to the upper rear surface of the heel-limiting plate, with its working end pointing vertically downwards. It is used to measure the distance between the upper and lower ankle exoskeleton layers, and gait phase prediction is achieved through the measurement data from the displacement sensor. The gait phase prediction results assist the hip exoskeleton assembly in performing upper-layer intention recognition and control. The energy storage deformation element is a spring; The spring has a hollow screw inside; the upper and lower ends of the hollow screw are respectively fixed vertically to the top plate and the bottom plate by bolts. 2.The intention recognition control method of the separable hip-ankle assist exoskeleton robot according to claim 1, wherein, The method includes the following steps: S1. Displacement sensor data acquisition and processing: This includes the acquisition and processing of displacement sensor data under three conditions: flat ground, going upstairs, and going downstairs. S2. After data processing, a gait data model containing various working conditions is obtained; S3. The gait data model is put into the control chip of the hip joint component, so that the hip joint exoskeleton component can make corresponding intention recognition control based on the data of the foot displacement sensors.

3. The intent recognition and control method according to claim 2, characterized in that: Under flat ground conditions, displacement sensor data of the heels of both feet were collected by the acquisition device under three gait phases: "right foot supporting left foot swinging", "both feet supporting", and "left foot supporting right foot swinging". After normalizing the data, an SVM model is used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the three gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions of the training set and the test set for the three gait phases.

4. The intention recognition control method according to claim 2, characterized by: Under the condition of going upstairs, displacement sensor data of the heels of both feet were collected by the acquisition device under four gait phases: "right foot in front, both feet supporting", "right foot supporting, left foot swinging", "left foot in front, both feet supporting", and "left foot supporting, right foot swinging". After normalizing the data, an SVM model is used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the four gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions of the training set and the test set for the four gait phases.

5. The intention recognition control method according to claim 2, characterized by: Under the condition of going downstairs, displacement sensor data of the heels of both feet were collected by the acquisition device under four gait phases: "right foot in front, both feet supporting", "right foot supporting, left foot swinging", "left foot in front, both feet supporting", and "left foot supporting, right foot swinging". After normalizing the data, an SVM model is used for recognition and classification to obtain the recognition accuracy of the training set and the test set, as well as the confusion matrix of the training set and the test set for the four gait phases. The confusion matrix includes the number of correct recognitions and the number of incorrect recognitions of the training set and the test set for the four gait phases.

6. The intention recognition control method according to claim 5, characterized by: Step S3 includes: putting the SVM model obtained in step S2, which includes three working conditions, into the control chip of the hip exoskeleton component; communicating the real-time data collected by the foot displacement sensors to the control chip of the hip exoskeleton component; and matching the gait phase data in the SVM model with the displacement sensor data to make corresponding upper-level intention recognition control actions.

7. The intention recognition control method according to claim 5, characterized by: The acquisition device includes a power supply, a signal conversion element, and a host computer; the host computer is connected to the displacement sensor in sequence through the signal conversion element and the power supply to acquire and read displacement sensor data; the power supply is connected to the displacement sensor to supply power to it.

8. The intention recognition control method according to claim 7, characterized by: The signal conversion element is a 485 interface to USB interface element.

Images