Motion recognition method based on sensor cluster and related equipment

Abstract

The invention provides a motion recognition method based on a sensor cluster and related equipment, and is applied to the technical field of data processing. The method comprises the following steps: acquiring angle data of an inertial sensor arranged in a controller, an original angle of a left thigh outer inertial sensor and an original angle of a right thigh outer inertial sensor; zero calibration processing is performed on the original angle of the left thigh outer inertial sensor and the original angle of the right thigh outer inertial sensor based on the angle data of the inertial sensors, and the corrected initial angle of the inertial sensors is generated; smoothing the corrected three-axis angle, angular velocity and acceleration of the inertial sensor through moving average filtering to generate real-time motion data of the target user; processing the real-time motion data of the target user based on the motion recognition model to generate motion mode information; and processing the motion mode information and the real-time motion data of the target user based on the motion recognition mode to generate assistance triggering information.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a motion recognition method based on a sensor cluster and related equipment. Background Art

[0002] Inertial sensors are one of the common data sources for human activity monitoring and motion classification. Numerical methods are usually used to analyze and process the time series data generated by sensors. Machine learning techniques such as random forests and support vector machines are often used for motion classification, but they require manual extraction of various potential features from a large amount of raw data, and the signal preprocessing and feature extraction steps require professional and complex digital signal processing technology.

[0003] In the field of wearable exoskeletons, monitoring of human movement intentions is particularly important. Waist-lifting exoskeletons are often used in logistics warehouses, logistics warehousing and other scenarios. At present, lower limb motion recognition is one of the key issues in the integration of intelligent prostheses and humans. If there is no effective information exchange and coordinated control between humans, prostheses and the environment, it is impossible to achieve the ideal control effect under various road conditions, different behaviors and different walking stages, which makes it difficult for the prosthesis to maintain good stability and comfort. At present, the mainstream waist-lifting exoskeletons can only provide lifting assistance for the human body. Based on the motor output torque method, the exoskeleton can also provide walking assistance for the human body. How to accurately provide different assist torques for the human body depends on the exoskeleton's perception of various motion states in common scenarios, which is a technical problem that technicians in this field urgently need to solve.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide a motion recognition method and related equipment based on a sensor cluster, which at least overcomes the problems existing in the prior art to a certain extent, and realizes motion mode recognition and switching of walking mode and lifting mode by acquiring real-time data collected by a small inertial sensor group and processing it, thereby realizing differentiated gait judgment and assistance trigger point recognition in different motion modes.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a motion recognition method based on a sensor cluster is provided, comprising: obtaining angle data of an inertial sensor disposed inside a controller, an original angle of an inertial sensor on the outside of a left thigh, and an original angle of an inertial sensor on the outside of a right thigh; performing zero-position calibration processing on the original angle of the inertial sensor on the outside of the left thigh and the original angle of the inertial sensor on the outside of the right thigh based on the angle data of the inertial sensor, respectively, to generate an initial angle of the inertial sensor after correction, wherein the initial angle of the inertial sensor after correction includes the initial angle of the inertial sensor on the outside of the left thigh and the initial angle of the inertial sensor on the outside of the right thigh. degrees; real-time acquisition of the three-axis angle, angular velocity and acceleration of the corrected inertial sensor; smoothing the three-axis angle, angular velocity and acceleration of the corrected inertial sensor through sliding average filtering to generate real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the right thigh external inertial sensor around the Z axis and the angle of rotation of the left thigh external inertial sensor around the Z axis; processing the real-time motion data of the target user based on the motion recognition model to generate motion pattern information; processing the motion pattern information and the real-time motion data of the target user based on the motion recognition mode to generate power-assist triggering information.

[0008] Another aspect of the present application is a motion recognition device based on a sensor cluster, characterized in that the device includes: an acquisition module, used to acquire the angle data of an inertial sensor arranged inside a controller, the original angle of an inertial sensor on the outside of a left thigh, and the original angle of an inertial sensor on the outside of a right thigh; real-time acquisition of the three-axis angle, angular velocity, and acceleration of the corrected inertial sensor; a processing module, used to perform zero-position calibration processing on the original angle of the inertial sensor on the outside of the left thigh and the original angle of the inertial sensor on the outside of the right thigh based on the angle data of the inertial sensor, to generate an initial angle of the corrected inertial sensor, wherein the inertial sensor on the outside of the left thigh is calibrated to generate an initial angle of the corrected inertial sensor. The initial angle includes the initial angle of the inertial sensor on the outside of the left thigh and the initial angle of the inertial sensor on the right thigh; the three-axis angle, angular velocity and acceleration of the corrected inertial sensor are smoothed by sliding average filtering to generate real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the external inertial sensor of the right thigh around the Z axis and the angle of rotation of the external inertial sensor of the left thigh around the Z axis; the real-time motion data of the target user is processed based on the motion recognition model to generate motion pattern information; the motion pattern information and the real-time motion data of the target user are processed based on the motion recognition mode to generate power assist trigger information.

[0009] According to another aspect of the present application, an electronic device is characterized in that it includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-mentioned sensor cluster-based motion recognition method by executing the executable instructions.

[0010] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the above-mentioned motion recognition method based on a sensor cluster is implemented.

[0011] According to another aspect of the present application, a computer program product is provided, including a computer program, characterized in that when the computer program is executed by a third processor, the above-mentioned motion recognition method based on sensor cluster is implemented.

[0012] The present application provides a motion recognition method based on a sensor cluster and related equipment, wherein a server obtains angle data of an inertial sensor disposed inside a controller, an original angle of an inertial sensor on the outside of a left thigh, and an original angle of an inertial sensor on the outside of a right thigh; based on the angle data of the inertial sensor, the original angle of the inertial sensor on the outside of the left thigh and the original angle of the inertial sensor on the outside of the right thigh are zero-calibrated to generate an initial angle of the inertial sensor after correction, wherein the initial angle of the inertial sensor after correction includes the initial angle of the inertial sensor on the outside of the left thigh and the initial angle of the inertial sensor on the outside of the right thigh. degrees; real-time acquisition of the three-axis angle, angular velocity and acceleration of the corrected inertial sensor; smoothing the three-axis angle, angular velocity and acceleration of the corrected inertial sensor through sliding average filtering to generate real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the right thigh external inertial sensor around the Z axis and the angle of rotation of the left thigh external inertial sensor around the Z axis; processing the real-time motion data of the target user based on the motion recognition model to generate motion pattern information; processing the motion pattern information and the real-time motion data of the target user based on the motion recognition mode to generate power-assist triggering information.

[0013] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flow chart of a motion recognition method based on a sensor cluster provided by an embodiment of the present application is shown;

[0015] Figure 2A schematic diagram of the structure of a motion recognition device based on a sensor cluster provided in an embodiment of the present application is shown;

[0016] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present application is shown;

[0017] Figure 4 A schematic diagram of a storage medium provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0019] Combine the following Figure 1 To describe the motion recognition method based on the sensor cluster according to the exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principle of the present application, and the embodiments of the present application are not limited in this respect. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0020] In one implementation, the present application also proposes a motion recognition method based on a sensor cluster and related equipment. Figure 1 The flowchart of a motion recognition method based on a sensor cluster according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to a server, comprising:

[0021] S101, obtaining angle data of an inertial sensor disposed inside a controller, an original angle of an inertial sensor disposed outside the left thigh, and an original angle of an inertial sensor disposed outside the right thigh.

[0022] In one embodiment, after the target user wears the waist lifting exoskeleton (or related equipment) equipped with the system of the present invention, the user adopts an upright standing posture to perform power-on operation. After the system is powered on, the controller starts to work, and it will collect the first 10 groups of data (such as angle, angular velocity, acceleration, etc.) of various types of data from each inertial sensor (including the inertial sensor inside the controller, the inertial sensor on the outside of the left thigh, and the inertial sensor on the outside of the right thigh) at a frequency of once every 10ms. For example, for angle data, the first 10 groups of angle data of the inertial sensor inside the controller, the first 10 groups of raw angles of the inertial sensor on the outside of the left thigh, and the first 10 groups of raw angles of the inertial sensor on the outside of the right thigh will be collected.

[0023] After collecting the first 10 sets of data, calculate the average value of these data. Taking the angle data of the inertial sensor inside the controller as an example, assume that the first 10 angle values ​​collected are θ1, θ2, ..., θ 10 , then its average value The purpose of averaging is to eliminate individual data errors or data loss that may occur in the initial state of the system. For example, in the actual acquisition process, factors such as the instability of the sensor at the moment of startup may cause one or several data to deviate from the normal range. By averaging, a value that better represents the actual initial state of the system can be obtained.

[0024] Finally, the calculated average angle data of inertial sensor 2 (inertial sensor inside the controller) is used to calibrate the initial angles of inertial sensor 3 (inertial sensor on the outside of the left thigh) and inertial sensor 4 (inertial sensor on the outside of the right thigh) to correct their initial angles. For example, if the calculated average angle of the inertial sensor inside the controller is The initial angles of the left thigh outer inertial sensor and the right thigh outer inertial sensor are adjusted to a state where the deviation from the average value is zero, which is used as an accurate benchmark for subsequent motion state monitoring and calculation. In this way, in the subsequent monitoring of the movement process, the changes in the movement state of both legs can be more accurately judged, such as the degree of difference in the movement of both legs (monitoring parameter D) and the degree of consistency in the movement state of both legs (monitoring parameter C), and then the movement mode (walking mode or lifting mode) and the power-assistance trigger point in the corresponding movement mode can be accurately identified.

[0025] S102: performing zero-position calibration processing on the original angle of the left thigh outer side inertial sensor and the original angle of the right thigh outer side inertial sensor based on the angle data of the inertial sensor, respectively, to generate the initial angle of the inertial sensor after correction.

[0026] In one embodiment, the original angles of the left thigh outer inertial sensor and the right thigh outer inertial sensor are zero-calibrated using the angle data of the inertial sensor disposed inside the controller, and the initial angles of the left thigh outer inertial sensor and the right thigh outer inertial sensor are corrected. The corrected initial angles of the inertial sensor include the initial angles of the left thigh outer inertial sensor and the initial angles of the right thigh outer inertial sensor.

[0027] Zero calibration principle and data preparation

[0028] After the target user wears the device and completes the system power-on, and executes step S101 to collect and process data, zero calibration begins.

[0029] Assume that the average angle data of the inertial sensor inside the controller has been obtained through step S101 (Assume that the average value of the angle of rotation around the Z axis is ), and the raw angle data θ of the left thigh outer inertial sensor L1 ,θ L2 , ..., θ L10 (The rotation angles around the Z axis are ) and the raw angle data θ of the right thigh outer inertial sensor R1 ,θ R2 , ..., θ R10 (The rotation angles around the Z axis are ).

[0030] Zero calibration of the left thigh outer inertial sensor, and calculation of the average original angle of the left thigh outer inertial sensor around the Z axis

[0031] Then use the average angle around the Z axis of the inertial sensor inside the controller The initial angle of the left thigh outer inertial sensor is zero-calibrated to obtain the initial angle of the left thigh outer inertial sensor after correction. For example, if but This 5° is the initial angle (around the Z axis) of the inertial sensor on the outside of the left thigh after calibration with the inertial sensor inside the controller as the reference.

[0032] The zero position calibration of the inertial sensor on the right thigh is also performed, and the average value of the original angle of the inertial sensor on the right thigh around the Z axis is calculated.

[0033] Use the average value of the controller's internal inertial sensor's angle around the Z axis The initial angle of the right thigh outer side inertial sensor is zero-calibrated to obtain the initial angle of the right thigh outer side inertial sensor after correction. for example but Then, this 3° is the initial angle (around the Z axis) of the calibrated right thigh outer inertial sensor.

[0034] The significance of this zero-position calibration is to ensure that the three sensors (the inertial sensor inside the controller and the inertial sensors on the outside of the left and right thighs) can use the same inertial sensor with the same accuracy to monitor the motion state when rotating in the sagittal plane of the human body.

[0035] When judging the consistency of the movement state of the two legs (monitoring parameter C) and the difference of the movement state of the two legs (monitoring parameter D), the movement mode can be more accurately identified based on the initial angle after zero calibration. (K is the proportional magnification factor), the accurate initial angle can obtain a more reliable monitoring parameter C value, so as to accurately judge whether the human body leaves the walking state and enters the lifting assistance mode (when the monitoring parameter C is greater than the threshold T and the monitoring parameter D is in the range of ±0.2), or judge whether the human body is in the walking assistance mode (when the monitoring parameter C is less than the threshold T and the monitoring parameter D leaves the interval near 0), etc., thereby realizing the accurate recognition of the human body's movement state and the accurate triggering of the assistance function by the entire system.

[0036] S103, collecting the three-axis angle, angular velocity and acceleration of the corrected inertial sensor in real time.

[0037] After completing the zero position calibration of the left thigh outer inertial sensor and the right thigh outer inertial sensor (step S102), the system starts to collect data in real time. The collection equipment is three inertial sensors that have been zero-calibrated, namely, the inertial sensor inside the controller, the left thigh outer inertial sensor (after correction), and the right thigh outer inertial sensor (after correction).

[0038] The data acquisition frequency is set to 100 Hz, which means that the relevant data of each inertial sensor will be collected 100 times per second. For example, in one second, data from the three sensors will be collected at 0.01s, 0.02s, 0.03s, etc.

[0039] For each inertial sensor, its three-axis angle, angular velocity and acceleration data need to be collected. Taking the inertial sensor on the outer side of the left thigh as an example, the data collected at a certain time t may be:

[0040] Three-axis angle: θ Lx (t) (rotation angle around the X axis), θ Ly (t) (rotation angle around the Y axis), θ Lz (t) (rotation angle around the Z axis, here θ Lz (t) is the angle value after zero position calibration, and its initial value has been calibrated in S102).

[0041] Angular velocity: ω Lx (t) (angular velocity around the X axis), ω Ly (t) (angular velocity around the Y axis), ω Lz (t) (angular velocity around the Z axis).

[0042] Acceleration: a Lx(t) (acceleration along the X-axis), a Ly (t) (acceleration along the Y axis), a Lz (t) (acceleration along the Z axis).

[0043] Similarly, for the inertial sensor on the outside of the right thigh and the inertial sensor inside the controller, similar three-axis angle, angular velocity and acceleration data are collected at the same time, which are recorded as θ Rx (t),θ Ry (t),θ Rz (t),ω Rx (t),ω Ry (t),ω Rz (t), a Rx (t), a Ry (t), a Rz (t) and θ Cy (t),θ Cz (t),θ Cx (t),ω Cx (t),ω Cy (t),ω Cz (t), a Cx (t), a Cy (t), a Cz (t).

[0044] These real-time collected data will be used in subsequent algorithm steps such as motion pattern recognition and assist trigger point recognition. For example, when judging walking mode and lifting mode (upper layer perception algorithm), the angle θ of the inertial sensors of both legs rotating around the Z axis will be used. Lx (t),θ Rz (t) and the monitoring parameters calculated from these angles (in is the average value of the inertial sensor angle around the Z axis inside the controller, K is the proportional magnification factor) and monitoring parameter D (negatively correlated with the consistency of leg movement, and its calculation is also based on data such as the angle of the inertial sensors of the legs). When assisting in starting point identification in the lifting mode, the angular velocity of the legs ωL is used z (t),ωR z (t) and other data to calculate the consistency of the angular velocity of both legs AVC, and determine the assist time point when AVC meets certain conditions. When identifying the assist trigger point in walking mode, the thigh Z-axis angle θ is used Lz (t),θ Rz (t) and angular velocity ωL z (t),ωR z(t) and other data to determine whether it has entered the pre-assistance interval and whether it has finally reached the assistance starting point. In short, the real-time collected data is the basis for the accurate operation of the entire motion state recognition system and provides key information basis for the implementation of various subsequent algorithms.

[0045] S104, smoothing the three-axis angle, angular velocity and acceleration of the corrected inertial sensor by means of a sliding average filter to generate real-time motion data of the target user.

[0046] In one implementation, the real-time motion data of the target user includes a rotation angle of the external inertial sensor of the right thigh around the Z axis and a rotation angle of the external inertial sensor of the left thigh around the Z axis.

[0047] Assume that we have collected a series of corrected inertial sensors (including the left thigh outer inertial sensor and the right thigh outer inertial sensor) in real time, such as t1, t2, t3, ..., t n The collected data.

[0048] Taking the rotation angle of the right thigh outer side inertial sensor around the Z axis as an example, let the collected rotation angle sequence of the right thigh outer side inertial sensor around the Z axis be θ Rz (t1), θ Rz (t2),θ Rz (t2), ..., θ Rz (t n ).

[0049] The sliding average filter selects a window of fixed length (for example, the window length is m), and then calculates the average value of the data in the window as the smoothing value of the data at the center of the window. Rz (t i )(i≥m / 2 and i≤nm / 2), the angle value after smoothing The calculation is as follows:

[0050] For example, when m=5, i=3,

[0051] Through such calculation, for each time point within the window range, a smoothed rotation angle value of the right thigh outer side inertial sensor around the Z axis can be obtained.

[0052] For the left thigh outer inertial sensor rotation angle sequence θ around the Z axis Lz (t1), θ Lz (t2),θ Lz (t2), ..., θ Lz (tn ), and the above-mentioned sliding average filtering method is also used for smoothing to obtain a smoothed angle value sequence.

[0053] Not only the rotation angle around the Z axis, but also the rotation angle around the X axis and the Y axis in the three-axis angle, as well as the angular velocity and acceleration data, are also smoothed using the same sliding average filtering method. For example, the rotation angle θ of the inertial sensor around the X axis on the outside of the right thigh Rz (t1), θ Rz (t2), θ Rz (t3), ..., θ Rz (t n ) After filtering, we get

[0054] The real-time motion data of the target user generated after the sliding average filter processing (such as the smoothed rotation angle of the inertial sensors of both legs around the Z axis, etc.) is of great significance in the subsequent motion pattern recognition and power-assistance trigger point recognition algorithm. When judging the walking mode and lifting mode, the monitoring parameters C and D are calculated using the smoothed rotation angle of the inertial sensors of both legs around the Z axis, which can reduce the misjudgment caused by data fluctuations and make the motion pattern recognition more accurate and reliable. In the power-assistance trigger point recognition algorithm, for example, when judging the pre-assistance interval and the starting point of the power-assistance in the walking mode, the smoothed thigh Z-axis angle and angular velocity data can more stably reflect the changes in the human body's motion state, thereby more accurately determining the timing of the power-assistance and improving the coordination and effectiveness of the exoskeleton power-assistance system and human body motion.

[0055] S105: Process the real-time motion data of the target user based on a motion recognition model to generate motion pattern information.

[0056] In one embodiment, the motion recognition model includes a calculation formula for calculating the monitoring parameters of the double leg motion, and the calculation formula is: C = K c (sin(θ Rx )+sin(θ Lx )) and D=K d (sin(θ Rx )-sin(θ Lx )); where C represents the consistency of the movement state of the two legs, D represents the difference of the movement of the two legs, θ Rx Indicates the angle of rotation of the right inertial sensor around the Z axis, θ Lx Indicates the rotation angle of the left inertial sensor around the Z axis, K c and K d is the proportional magnification factor.

[0057] For example, suppose that at a certain time t, the smoothed angle value of the right thigh outer side inertial sensor rotating around the Z axis obtained after processing in step S104 is (It is roughly considered here ), the smoothed angle value of the left thigh outer inertial sensor rotating around the Z axis (same approximate treatment), K c =2, K d =3.

[0058] For monitoring parameter C: first convert the angle value into radians.

[0059] Then substitute the formula to calculate

[0060] calculate (Keep three decimal places).

[0061] Then C = 2×(0.5+0.707) = 2.414.

[0062] For monitoring parameter D:

[0063] Similarly, convert the angle to radians first, and then substitute it into the formula

[0064] Calculation shows that D = 3 × (0.5-0.707) = -0.621.

[0065] The motion mode information is generated based on the calculated values ​​of monitoring parameters C and D. Generally, a threshold value T is set in advance. When c is greater than the threshold value t and d is in the range of ±0.2, it is considered that the human body has left the walking state, and the perception system enters the lifting and assisting mode; conversely, when C is less than the threshold value T and D leaves the interval near 0, the perception system will enter the walking and assisting mode. For example, assuming that the threshold value T = 2, since C = 2.414>T and D = -0.621 is in the range of ±0.2, the system determines that the human body is in the lifting and assisting mode and generates the corresponding motion mode information. If the calculation result does not meet the above conditions, it may be determined as a walking and assisting mode or further data analysis is required to determine the motion mode. In this way, through the processing of real-time motion data by the motion recognition model, the current motion mode of the target user can be accurately identified, providing a basis for subsequent assisting trigger point identification, etc.

[0066] In another embodiment, when the monitoring parameter C is greater than the preset threshold value T, and the monitoring parameter D is in the interval of ±0.2, the real-time motion mode of the target user is generated, wherein the real-time motion mode of the target user is the lifting mode; when the monitoring parameter C is less than the preset threshold value T, or the monitoring parameter D is greater than 0.2 or less than -0.2, the real-time motion mode of the target user is generated, wherein the real-time motion mode of the target user is the walking mode. Since D represents the difference between the movements of the legs, when the absolute value of D is greater than 0.2, it means that the difference between the legs is very large, and when the absolute value of D is less than 0.2, it is considered that the difference between the legs is reduced; C represents the consistency of the movement of the legs, and the threshold value T is a positive value. When C is greater than T and the absolute value of D is less than 0.2, it means that the thigh movement state of the legs is consistent, and the angle between the left and right legs and the human body trunk gradually decreases from 180°, which means that the human body is in a bent state, so it enters the lifting mode. In addition, when the monitoring parameters do not meet these requirements, they are all considered to be walking modes.

[0067] S106: Process the motion pattern information and the real-time motion data of the target user based on the motion recognition mode to generate power assist trigger information.

[0068] In one embodiment, if the current motion model information indicates that the real-time motion mode of the target user is a lifting mode, the motion recognition model further includes a calculation formula for calculating the angular velocity consistency monitoring parameter of the two legs, and the calculation formula is: AVC = K a (sin(ω RZ )+sin(ω LZ )); wherein θ represents angular velocity, the present invention uses Z-axis angular velocity, and AVC represents the consistency of angular velocity of the two legs; when AVC is lower than a preset threshold, power-assistance trigger information is generated. The threshold of AVC is less than 0. When AVC is lower than the threshold, it means that the angular velocity of the two legs increases in the negative direction, indicating that the angle between the human body trunk and the two legs gradually increases, which means that the user gradually enters the upright state from the bent state.

[0069] After determining that the current motion mode is the lifting mode, the motion recognition model uses the formula AVC = K a (sin(ω RZ )+sin(ω LZ )) to calculate the monitoring parameters of the angular velocity consistency of the two legs. RZ Represents the angular velocity of the right inertial sensor around the Z axis, ω LZ Represents the angular velocity of the left inertial sensor around the Z axis, K a is the proportional magnification factor (determined by experiment).

[0070] For example, suppose that at a certain time t, the angular velocity ω of the inertial sensor on the outside of the right thigh around the Z axis is RZ(t) = -0.3rad / s (the negative sign here indicates the direction. According to the setting, the swing to the front of the human leg is positive, and it may be swinging backward at this time). The angular velocity of the inertial sensor on the outer side of the left thigh around the Z axis is ω LZ (t) = -0.4 rad / s, K a =5.

[0071] Calculate AVC and determine the assist triggering conditions, specifically:

[0072] First, substitute the angular velocity value into the formula to calculate AVC:

[0073] AVC=5×(sin(-0.3)+sin(-0.4)).

[0074] Calculate sin(-0.3)=-0.295 (keep three decimal places), sin(-0.4)=-0.389 (keep three decimal places).

[0075] Then AVC=5×(-0.295-0.389)=-3.42.

[0076] Assume that the preset threshold is -2 (because the threshold of AVC is less than 0), since AVC=-3.42<-2, the condition that AVC is lower than the threshold is satisfied.

[0077] When it is lower than the threshold, according to the setting, this means that the angular velocity of the legs increases in the negative direction, that is, the angle between the human torso and the legs gradually increases. At this time, the user gradually enters the upright state from the bent state, and it is the time when power assistance is needed. Therefore, at this time, the system generates power assistance trigger information and sends the power assistance signal to the control system. The exoskeleton can begin to provide power assistance to the human body, assist the human hip joint to open, reduce the human body's force, and make the lifting action easier. In this way, the power assistance trigger point can be accurately identified in the lifting mode, improving the timeliness and effectiveness of the exoskeleton power assistance, better assisting the human body to complete the lifting action, and reducing the burden on the human body.

[0078] In another implementation, if the current motion model information indicates that the real-time motion mode of the target user is a walking mode, the human body posture information and pre-assistance interval information are obtained; the human body posture information is processed to generate human body posture pre-determination information; if the human body posture pre-determination information is in the pre-assistance interval, the human body posture pre-determination information is processed based on a preset processing rule to generate assistance triggering information. For example, when the Z-axis angle of the thigh enters the positive half zone, the angular velocity is monitored, and when the angular velocity enters the negative half zone, the pre-assistance interval is entered. In the pre-assistance interval, it is determined whether the assistance starting point is reached by judging a variety of data. Taking the left leg as an example, the sliding window method is used to screen the peak values ​​of the data adjacent interval. When the Z-axis angular velocity of the left thigh external inertial sensor installed on the left leg reaches the peak value, if the Z-axis angle of the right thigh external inertial sensor installed on the right leg enters the negative half zone, and at this moment the Z-axis angle of the left thigh external inertial sensor is greater than the minimum assistance angle set by the system, it enters the assistance interval and sends an assistance signal to the control system; if the Z-axis angle of the left thigh external inertial sensor enters the negative half zone or the Z-axis angles of the left thigh external inertial sensor and the right thigh external inertial sensor both enter the positive half zone, the pre-assistance interval is exited and the pre-assistance interval is re-determined.

[0079] In another embodiment, when the Z-axis angular velocity of the right thigh outer inertial sensor reaches a peak value and the Z-axis angular velocity of the left thigh outer inertial sensor enters the negative half zone; if the Z-axis angle of the right thigh outer inertial sensor is greater than the preset power-assistance angle, then the power-assistance trigger information is generated. Taking the right leg as an example, the sliding window method is used to filter the peak values ​​of the data adjacent interval. When the Z-axis angular velocity of the right thigh external inertial sensor installed on the right leg reaches a peak value, if the Z-axis angle of the left thigh external inertial sensor installed on the left leg enters the negative half zone, and at this moment the Z-axis angle of the right thigh external inertial sensor is greater than the minimum power-assistance angle set by the system, then the power-assistance interval is entered, and a power-assistance signal is sent to the control system; if the Z-axis angle of the right thigh external inertial sensor enters the negative half zone or the Z-axis angles of the right thigh external inertial sensor and the left thigh outer inertial sensor both enter the positive half zone, then the pre-assistance interval is exited and the pre-assistance interval is re-determined.

[0080] The present invention realizes adaptive switching between lifting assistance and walking assistance modes by real-time monitoring of the angle and angular velocity data of the inertial sensor at the thigh, and by cross-validation of the inertial sensor data change rules, motion monitoring parameter design, specific data conditions and double-leg inertial sensors, and simultaneously realizes the positioning of the assistance starting point and assistance end point during lifting, walking, climbing stairs and climbing slopes. During the process of going down stairs and down slopes, these movements are screened out by the perception algorithm to achieve no assistance. In the lifting mode, the assistance starting point is based on the trend of the human hip joint opening, which assists the human hip joint to open and reduces the human body's force; in the walking mode, the assistance starting point is based on the time point when the human leg touches the ground, and the assistance process is to apply a torque to open the hip joint, so that the human body's pushing torque increases, thereby increasing the human body's walking speed and reducing the human body's own walking force torque.

[0081] After waiting for the relevant control system to complete the corresponding assistance, the perception system determines the current state. If the assistance is terminated, it will re-enter the next motion recognition cycle. The perception system's recognition of the end of motion is different in different motion modes, and it is also related to the control system. If the control system's assistance torque is based on the length of time, the control system will send a signal to the perception system after completing the assistance, which serves as a sign of the end of the motion assistance. The motion assistance can also be terminated by the perception system's monitoring of the human body's posture. In the lifting mode, the human body is fully opened as the standard, and when the average Z-axis angle difference between the inertial sensor in the controller and the external inertial sensor on the left thigh and the external inertial sensor on the right thigh is less than the threshold, it is considered that the human body has reached an upright state and the assistance interval is terminated. In the walking mode, taking the left leg as an example, when the Z-axis angular velocity of the external inertial sensor on the left thigh reaches 0, the motion assistance of the left leg is stopped, or when the Z-axis angle of the external inertial sensor on the right thigh is greater than 0, the motion assistance of the left leg is stopped.

[0082] In the present application, the server obtains the angle data of the inertial sensor disposed inside the controller, the original angle of the inertial sensor on the outside of the left thigh, and the original angle of the inertial sensor on the outside of the right thigh; based on the angle data of the inertial sensor, the original angle of the inertial sensor on the outside of the left thigh and the original angle of the inertial sensor on the outside of the right thigh are zero-calibrated to generate the initial angle of the corrected inertial sensor, wherein the initial angle of the inertial sensor on the outside of the left thigh and the initial angle of the inertial sensor on the outside of the right thigh are respectively generated; the three-axis angle, angular velocity and acceleration of the corrected inertial sensor are collected in real time; the three-axis angle, angular velocity and acceleration of the corrected inertial sensor are smoothed by sliding average filtering to generate the real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the external inertial sensor of the right thigh around the Z axis and the angle of rotation of the external inertial sensor of the left thigh around the Z axis; the real-time motion data of the target user is processed based on the motion recognition model to generate motion mode information; the motion mode information and the real-time motion data of the target user are processed based on the motion recognition mode to generate the power-assistance trigger information. By acquiring real-time data collected by a group of small inertial sensors and processing it, it is possible to realize motion mode recognition and switching between walking mode and lifting mode, and to achieve differentiated gait judgment and assistance trigger point recognition in different motion modes.

[0083] Optionally, in another embodiment of the method of the present application, the human body posture pre-determination information is processed based on a preset processing rule to generate assistance triggering information, including:

[0084] When the Z-axis angular velocity of the left thigh outer inertial sensor reaches a peak value and the Z-axis angular velocity of the right thigh outer inertial sensor changes from a positive value to a negative value, if the Z-axis angle of the left thigh outer inertial sensor is greater than the preset power assist angle, power assist trigger information is generated.

[0085] In one implementation, in the walking mode, in order to accurately determine the power-assistance trigger point, the human body posture is first pre-determined and the pre-assistance interval is set. When the Z-axis angle of the thigh enters the positive half zone, the angular velocity is monitored. Here, the inertial sensor on the outer side of the left thigh is taken as an example. Assume that its Z-axis angle θ Lz (t) enters the positive half zone (i.e. θ Lz (t)>0), then start monitoring its Z-axis angular velocity ω Lz (t).

[0086] For example, at time t1, θ Lz (t1) = 10° (entering the positive half zone), now start to focus on ω Lz (t) changes. When ω Lz (t) enters the negative half zone (i.e. ω Lz(t)<0), it enters the pre-assistance range. Assume that at time t2, ω Lz (t2) = -5° / s (entering the negative half zone), at which point the system determines that it has entered the pre-assist interval.

[0087] After entering the pre-assistance range, continue to monitor relevant data to determine whether the assistance starting point has been reached. Lz (t) reaches the peak value (assuming that at time t3, ω Lz (t3) reaches a peak value, such as ω Lz (t2) = -10° / s), and the Z-axis angular velocity of the right thigh outer inertial sensor changes from a positive value to a negative value and enters the negative half zone (assuming that at time t4, ω Lz (t4) When the speed changes from 3° / s to -2° / s and enters the negative half zone, further check the Z-axis angle θ of the inertial sensor on the outer side of the left thigh. Lz (t).

[0088] If θ Lz (t) is greater than the preset power assist angle (the preset angle is usually designed to be 40°), then the power assist trigger information is generated. Lz (t5)=45°>40°. At this time, the system believes that the user has raised his legs to a large extent and is not walking on flat ground but walking up stairs or a slope. It generates power-assist trigger information (power-assist torque parameter), and the power assist applied by the exoskeleton to the user will increase accordingly.

[0089] If during the monitoring process, the Z-axis angle θ of the external inertial sensor on the left thigh Lz (t) enters the negative half region (e.g. θ Lz (t6) = -5°), or the Z-axis angles of the left thigh external inertial sensor and the right thigh external inertial sensor both enter the positive half zone (assuming that at time t7, θ Rz (t7) = 20° and θ Lz (t7) = 15°), then exit the pre-assistance interval and re-determine the pre-assistance interval. This means that the current motion state may not meet the expected power-assistance triggering conditions, and it is necessary to re-monitor the changes in human posture and find the appropriate power-assistance triggering time again to ensure the accuracy and effectiveness of the power-assistance and avoid misjudgment and unnecessary power-assistance operations.

[0090] By applying the above technical solution, the server obtains the angle data of the inertial sensor arranged inside the controller, the original angle of the inertial sensor on the outside of the left thigh, and the original angle of the inertial sensor on the outside of the right thigh; based on the angle data of the inertial sensor, the original angle of the inertial sensor on the outside of the left thigh and the original angle of the inertial sensor on the outside of the right thigh are zero-calibrated to generate the initial angle of the corrected inertial sensor, wherein the initial angle of the corrected inertial sensor includes the initial angle of the inertial sensor on the outside of the left thigh and the initial angle of the inertial sensor on the outside of the right thigh; the three-axis angle, angular velocity and acceleration of the corrected inertial sensor are collected in real time; the three-axis angle, angular velocity and acceleration of the corrected inertial sensor are smoothed by sliding average filtering to generate the real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the external inertial sensor of the right thigh around the Z axis and the angle of rotation of the external inertial sensor of the left thigh around the Z axis.

[0091] The motion recognition model includes a calculation formula for calculating the monitoring parameters of the two-leg motion, which is: C = K c (sin(θ R )+sin(θ L )); D = K d (sin(θ R )-sin(θ L )); wherein, the monitoring parameter C represents the consistency of the movement state of both legs, the monitoring parameter D represents the difference of the movement of both legs, θ represents the angle of rotation of the inertial sensor around the Z axis, R represents the right side, L represents the left side, and K is the proportional magnification factor; when the monitoring parameter C is greater than the preset threshold value T, and the monitoring parameter D is in the ±0.2 interval, the real-time movement mode of the target user is generated, wherein the real-time movement mode of the target user is the lifting mode; when the monitoring parameter C is less than the preset threshold value T, and the monitoring parameter D is greater than or less than the 0 interval, the real-time movement mode of the target user is generated, wherein the real-time movement mode of the target user is the walking mode; if the current movement model information indicates that the real-time movement mode of the target user is the lifting mode; the movement recognition model also includes a calculation formula for calculating the monitoring parameters for the consistency of the angular velocity of the movement of both legs, and the calculation formula is: in, Represents angular velocity. This method uses the Z-axis angular velocity. AVC represents the consistency of the angular velocity of the two legs.

[0092] When AVC is lower than a preset threshold, an assist trigger information is generated; if the current motion model information indicates that the real-time motion mode of the target user is a walking mode, the human body posture information and pre-assistance interval information are obtained; the human body posture information is processed to generate human body posture pre-determination information; if the human body posture pre-determination information is in the pre-assistance interval, when the Z-axis angular velocity of the inertial sensor on the left outer thigh reaches a peak value and the Z-axis angular velocity of the inertial sensor on the right outer thigh enters the negative half zone; if the Z-axis angle of the inertial sensor on the left outer thigh is greater than the preset assist angle, an assist trigger information is generated; when the Z-axis angular velocity of the inertial sensor on the right outer thigh reaches a peak value and the Z-axis angular velocity of the inertial sensor on the left outer thigh enters the negative half zone; if the Z-axis angle of the inertial sensor on the right outer thigh is greater than the preset assist angle, an assist trigger information is generated. By real-time monitoring of the angle and angular velocity data of the inertial sensor at the thigh, and through the change rules of the inertial sensor data, the design of motion monitoring parameters, specific data conditions and cross-validation of the inertial sensors of both legs, adaptive switching between lifting assistance and walking assistance modes is achieved, and lifting and walking can be achieved at the same time. In the walking mode, it can be used for common actions such as walking on flat ground, climbing stairs, and climbing slopes, and can filter out actions such as going down stairs and slopes that do not require assistance. In the lifting mode, the assistance starting point is based on the opening trend of the human hip joint, which assists the opening of the human hip joint and reduces the force exerted by the human body.

[0093] In one embodiment, if Figure 2 As shown, the present application also provides a motion recognition device based on a sensor cluster, comprising:

[0094] The acquisition module 201 is used to acquire the angle data of the inertial sensor disposed inside the controller, the original angle of the inertial sensor on the outside of the left thigh, and the original angle of the inertial sensor on the outside of the right thigh; and to collect the three-axis angle, angular velocity and acceleration of the corrected inertial sensor in real time;

[0095] The processing module 202 is used to perform zero-position calibration processing on the original angle of the left thigh outer side inertial sensor and the original angle of the right thigh outer side inertial sensor respectively based on the angle data of the inertial sensor to generate the initial angle of the corrected inertial sensor, wherein the initial angle of the corrected inertial sensor includes the initial angle of the left thigh outer side inertial sensor and the initial angle of the right thigh outer side inertial sensor; smooth the three-axis angle, angular velocity and acceleration of the corrected inertial sensor through sliding average filtering to generate the real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the right thigh outer side inertial sensor around the Z axis and the angle of rotation of the left thigh outer side inertial sensor around the Z axis; process the real-time motion data of the target user based on the motion recognition model to generate motion pattern information; process the motion pattern information and the real-time motion data of the target user based on the motion recognition mode to generate the power-assistance trigger information.

[0096] The present application embodiment provides an electronic device, such as Figure 3 As shown, the electronic device 3 includes a first processor 300, a memory 301, a bus 302 and a communication interface 303, and the first processor 300, the communication interface 303 and the memory 301 are connected via the bus 302; the memory 301 stores a computer program that can be run on the first processor 300, and when the first processor 300 runs the computer program, it executes the motion recognition method based on the sensor cluster provided in any of the aforementioned embodiments of the present application.

[0097] The memory 301 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 303 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0098] The bus 302 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 301 is used to store programs, and the first processor 300 executes the programs after receiving the execution instruction. The motion recognition method based on the sensor cluster disclosed in any implementation of the embodiment of the present application may be applied to the first processor 300, or implemented by the first processor 300.

[0099] The first processor 300 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the first processor 300. The above-mentioned first processor 300 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be embodied as a hardware decoding processor to execute, or a combination of hardware and software modules in the decoding processor to execute. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 301, and the first processor 300 reads the information in the memory 301 and completes the steps of the above method in combination with its hardware.

[0100] The electronic device provided by the above-mentioned embodiments of the present application and the motion recognition method based on sensor cluster provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0101] The present application embodiment provides a computer-readable storage medium, such as Figure 4 As shown, the computer-readable storage medium stores 401 a computer program, and when the computer program is read and executed by the second processor 402, the motion recognition method based on the sensor cluster as described above is implemented.

[0102] The technical solution of the embodiment of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an electronic device (which may be an air conditioner, a refrigeration device, a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0103] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the motion recognition method based on a sensor cluster provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0104] The embodiment of the present application provides a computer program product, including a computer program, which is executed by a third processor to implement the method described above. The computer program product provided by the above embodiment of the present application and the motion recognition method based on sensor cluster provided by the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the method adopted, run or implemented by the application program stored therein.

[0105] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for evaluating the motion recognition method based on sensor clusters, electronic devices, electronic devices, and readable storage medium embodiments, since they are basically similar to the above-mentioned motion recognition method based on sensor clusters, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned motion recognition method based on sensor clusters.

[0106] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims

1. A motion recognition method based on sensor cluster, characterized in that: include: Acquire angle data of an inertial sensor disposed inside the controller, an original angle of an inertial sensor on the outside of the left thigh, and an original angle of an inertial sensor on the outside of the right thigh; Based on the angle data of the inertial sensor, the original angle of the left thigh outer side inertial sensor and the original angle of the right thigh outer side inertial sensor are respectively calibrated to generate the initial angle of the inertial sensor after correction, wherein the initial angle of the inertial sensor after correction includes the initial angle of the left thigh outer side inertial sensor and the initial angle of the right thigh outer side inertial sensor; Collecting the three-axis angle, angular velocity and acceleration of the corrected inertial sensor in real time; Smoothing the three-axis angle, angular velocity and acceleration of the corrected inertial sensor by sliding average filtering to generate real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the right thigh external inertial sensor around the Z axis and the angle of rotation of the left thigh external inertial sensor around the Z axis; Processing the real-time motion data of the target user based on a motion recognition model to generate motion pattern information; The motion pattern information and the real-time motion data of the target user are processed based on the motion recognition mode to generate power assist triggering information.

2. The method according to claim 1, characterized in that The processing of the real-time motion data of the target user based on the motion recognition model to generate motion pattern information includes: The motion recognition model includes a calculation formula for calculating the monitoring parameters of the two-leg motion, and the calculation formula is: C=K c (sin(θ) Rx )+sin(θ Lx )); D=K d (sin(θ) Rx )-sin(θ Lx )); Among them, C represents the consistency of the movement state of the two legs, D represents the difference of the movement of the two legs, θ Rx Indicates the angle of rotation of the right inertial sensor around the Z axis, θ Lx Indicates the rotation angle of the left inertial sensor around the Z axis, K c and K d is the proportional magnification factor.

3. The method according to claim 2, characterized in that The processing of the motion pattern information and the real-time motion data of the target user based on the motion recognition model to generate the motion pattern information further includes: When the monitoring parameter C is greater than the preset threshold T and the monitoring parameter D is in the range of ±0.2, a real-time motion mode of the target user is generated, wherein the real-time motion mode of the target user is a lifting mode; When the monitoring parameter C is less than the preset threshold T, and the monitoring parameter D is greater than or less than the 0 interval, a real-time motion mode of the target user is generated, wherein the real-time motion mode of the target user is a walking mode.

4. The method according to claim 3, characterized in that The processing of the motion model information and the real-time motion data of the target user based on the motion recognition mode to generate the assist trigger information includes: If the current motion model information indicates that the real-time motion mode of the target user is a lifting mode; The motion recognition model also includes a calculation formula for calculating the monitoring parameters of the angular velocity consistency of the two-leg motion, and the calculation formula is: AVC=K a (sin(ω) RZ )+sin(ω LZ )); Among them, ω RZ Represents the angular velocity of the right inertial sensor around the Z axis, ω LZ Represents the angular velocity of the left inertial sensor around the Z axis, K a is the proportional amplification factor. When AVC is lower than the preset threshold, a power-assistance trigger information is generated.

5. The method according to claim 3, characterized in that The processing of the motion model information and the real-time motion data of the target user based on the motion recognition mode to generate the power-assistance triggering information further includes: If the current motion model information indicates that the real-time motion mode of the target user is a walking mode, obtaining human body posture information and pre-assistance interval information; Processing the human body posture information to generate human body posture pre-determination information; If the human body posture pre-determination information is in the pre-assistance interval, the human body posture pre-determination information is processed based on a preset processing rule to generate assistance triggering information.

6. The method according to claim 5, characterized in that The human body posture pre-determination information is processed based on a preset processing rule to generate power-assistance triggering information, including: When the Z-axis angular velocity of the inertial sensor on the left outer thigh reaches its peak value and the Z-axis angular velocity of the inertial sensor on the right outer thigh enters the negative half zone; If the Z-axis angle of the inertial sensor on the outside of the left thigh is greater than the preset power-assistance angle, power-assistance trigger information is generated.

7. The method according to claim 5, characterized in that The human body posture pre-determination information is processed based on a preset processing rule to generate power-assistance triggering information, further comprising: When the Z-axis angular velocity of the left thigh outer inertial sensor reaches a peak value and the Z-axis angular velocity of the right thigh outer inertial sensor changes from a positive value to a negative value, if the Z-axis angle of the left thigh outer inertial sensor is greater than the preset power assist angle, power assist trigger information is generated.

8. A motion recognition device based on a sensor cluster, characterized in that: The device comprises: An acquisition module is used to acquire angle data of an inertial sensor disposed inside the controller, an original angle of an inertial sensor on the outside of the left thigh, and an original angle of an inertial sensor on the outside of the right thigh; and to collect three-axis angles, angular velocities, and accelerations of the corrected inertial sensors in real time; The processing module is used to perform zero-position calibration processing on the original angle of the left thigh outer side inertial sensor and the original angle of the right thigh outer side inertial sensor respectively based on the angle data of the inertial sensor to generate the initial angle of the corrected inertial sensor, wherein the initial angle of the corrected inertial sensor includes the initial angle of the left thigh outer side inertial sensor and the initial angle of the right thigh outer side inertial sensor; smooth the three-axis angle, angular velocity and acceleration of the corrected inertial sensor through sliding average filtering to generate the real-time motion data of the target user, wherein the real-time motion data of the target user includes the angle of rotation of the right thigh outer side inertial sensor around the Z axis and the angle of rotation of the left thigh outer side inertial sensor around the Z axis; process the real-time motion data of the target user based on the motion recognition model to generate motion pattern information; process the motion pattern information and the real-time motion data of the target user based on the motion recognition mode to generate the power-assistance trigger information.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; The first processor is configured to execute the sensor cluster-based motion recognition method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the motion recognition method based on the sensor cluster described in any one of claims 1 to 7 is implemented.

Images