A data fusion processing system and method for wireless attitude sensor

Abstract

The present invention relates to the field of intelligent sensing technology, and specifically to a data fusion processing system and method for a wireless attitude sensor. The system includes a sequence construction module, a dominant identification module, a trend calculation module, an inertia analysis module, and a fusion output module. During attitude information processing, the present invention constructs a displacement direction sequence based on the linear velocity vector direction derived from three-axis acceleration, and generates a dynamic feature set based on the multi-cycle three-axis angular velocity change amplitude. The overall dominant motion direction is extracted through the cosine relationship of the angles between multiple parts, and the motion direction update index is determined based on the time-varying trend, thereby achieving dynamic correction of the attitude reference direction. On this basis, the projection analysis of the three-axis acceleration and three-axis angular velocity in the dominant direction is introduced. The inertia concentration of each part in the unified motion direction is reflected through weighted combination, and a more coordinated and accurate fusion output is achieved when the information of multiple parts is significantly different.

Description

Technical Field

[0001] The present invention relates to the field of intelligent sensing technology, and in particular to a data fusion processing system and method for a wireless attitude sensor. Background Art

[0002] The field of intelligent perception technology involves obtaining environmental or object status information through various sensors, and combining processors, data fusion technology and other means to conduct real-time analysis and understanding of multi-source data, including sensor data collection, information fusion, status identification and judgment, and decision feedback.

[0003] Among them, the data fusion processing system of wireless attitude sensors refers to a system used on wearable devices such as exoskeleton manipulators to collect and fuse attitude data output by multiple sensors through wireless communication to estimate and reconstruct the spatial attitude of the target object.

[0004] Existing technologies only directly fuse data collected by multiple wireless attitude sensors to estimate spatial attitude, lacking recognition and trend analysis of the dominant motion direction. As a result, when there are directional conflicts or inertia differences among sensors in multiple locations, the main motion trend cannot be clearly determined, leading to attitude estimation direction drift or unstable output. In complex movements or rapidly changing states, such as when an exoskeleton manipulator swings sharply or switches between continuous movements, the overall estimation result is easily affected by local outliers, and the importance of key parts cannot be dynamically distinguished in data fusion, resulting in response lag, direction misjudgment, or fusion imbalance in the attitude output, limiting the adaptability and stability of the system in actual wearable devices. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a data fusion processing system and method for a wireless attitude sensor.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A data fusion processing system for a wireless attitude sensor includes:

[0007] The sequence construction module collects the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. It calculates the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence, and calculates the change amplitude of the three-axis angular velocity vector of each location in consecutive periods to form an angular velocity change amplitude sequence.

[0008] A dominant identification module identifies the dominant motion direction sequence of the entire exoskeleton manipulator based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part;

[0009] a trend calculation module, which calculates a motion trend index of the exoskeleton manipulator according to the temporal trend of the dominant motion direction sequence, and updates a posture reference direction set according to the index;

[0010] an inertial analysis module, which projects the three-axis angular velocity and the three-axis velocity vector onto the dominant motion direction sequence based on the posture reference direction set, calculates the rotational inertia response value and the motion inertia response value of each part to the dominant direction, and weights and combines them to form an inertia concentration index for each part;

[0011] The fusion output module analyzes the trend stability of the inertia concentration index, fuses the corresponding information of each part, and obtains the posture data fusion result.

[0012] As a further solution of the present invention, the displacement direction vector sequence includes the linear velocity direction information of each part, the spatial direction relationship, and the direction continuity within the period; the dominant motion direction sequence is specifically the overall motion direction vector, the unified direction attribute, and the cross-part direction fusion feature; the posture reference direction set includes the current reference direction, the historical direction update record, and the dominant direction replacement basis; the inertia concentration index includes the rotational response value, the linear response value, and the weighted response strength; the posture data fusion result specifically refers to the fused acceleration information, the fused angular velocity information, and the unified output posture direction.

[0013] As a further solution of the present invention, the steps for obtaining the displacement direction vector sequence and the angular velocity change amplitude sequence are specifically as follows:

[0014] The three-axis acceleration calculation submodule collects the three-axis acceleration of the wireless posture sensors in multiple parts of the exoskeleton manipulator within a specified period, performs bandpass filtering on the three-axis acceleration collected in each period of the wrist, elbow, and shoulder, integrates the acceleration of each axis in time sequence, calculates the three-axis velocity vector, and generates a three-axis velocity vector set;

[0015] The spatial direction acquisition submodule calculates the modulus of each three-axis velocity vector in the three-axis velocity vector set and performs normalization processing, extracts the spatial direction of the vector and arranges it in chronological order to generate a displacement direction vector sequence;

[0016] The angular velocity variation amplitude calculation submodule collects the three-axis angular velocities of wireless posture sensors at multiple locations of the exoskeleton manipulator in continuous cycles, extracts the three-axis angular velocity vectors of the wrist, elbow, and shoulder in two adjacent cycles, calculates the Euclidean distance of the three-axis angular velocity vectors between adjacent cycles as the variation amplitude of each location, and generates an angular velocity variation amplitude sequence.

[0017] As a further solution of the present invention, the step of acquiring the dominant motion direction sequence is specifically as follows:

[0018] A directional feature construction submodule extracts the displacement direction vector and angular velocity change amplitude within the corresponding period according to the displacement direction vector sequence and angular velocity change amplitude sequence of each part, and generates a directional feature set;

[0019] The angle cosine value calculation submodule calls the displacement direction vector of each part in the directional feature set, calculates the cosine values ​​of the angles between the wrist and elbow, wrist and shoulder, and elbow and shoulder, and calculates the average value of the three sets of angle cosine values ​​in each cycle to generate an angle cosine average value sequence;

[0020] The dominant motion direction sequence extraction submodule selects the direction vector associated with the maximum average value of each cycle in the angle cosine average value sequence as the dominant motion direction sequence of the current cycle, arranges the dominant directions of all cycles in chronological order, and generates a dominant motion direction sequence.

[0021] As a further solution of the present invention, the steps of acquiring the attitude reference direction set are specifically as follows:

[0022] The dominant direction angle calculation submodule extracts the dominant motion direction vectors of consecutive periods of the dominant motion direction sequence, calculates the spatial angle between the dominant motion direction vectors of two adjacent periods in chronological order, and generates a dominant direction angle sequence;

[0023] The change rate generation submodule calls the dominant direction angle sequence, performs first-order difference on each cycle angle in chronological order, regards it as the increment of the angle change between each cycle, and generates a direction change rate sequence;

[0024] The reference direction update submodule determines whether the angle change in the current cycle shows a continuous upward trend based on the direction change rate sequence. If the change value continues to increase within two consecutive cycles, it is confirmed that the dominant motion direction offset state is established, and the dominant motion direction vector of the current cycle is called as the update basis to generate the posture reference direction set.

[0025] As a further solution of the present invention, the step of obtaining the inertia concentration index is specifically as follows:

[0026] The angular velocity projection calculation submodule projects the three-axis angular velocity vectors onto the dominant motion direction of the corresponding period in the dominant motion direction sequence according to the posture reference direction set, extracts the angular velocity components of the wrist, elbow, and shoulder in the target direction as the rotational inertia response value of each part, and generates a rotational inertia response value sequence;

[0027] The linear velocity projection calculation submodule calls the posture reference direction set, projects the three-axis velocity vectors onto the dominant motion direction of the corresponding period in the dominant motion direction sequence according to the body part, extracts the linear velocity change rate of the wrist, elbow, and shoulder in the target direction as the motion inertia response value of each body part, and generates a motion inertia response value sequence;

[0028] The inertia concentration generation submodule calls the rotational inertia response value sequence and the motion inertia response value sequence, performs a weighted combination of the response values ​​of the same period in the two sequences by part, synthesizes the inertia response aggregation result of each part in the dominant motion direction, and generates an inertia concentration index.

[0029] As a further solution of the present invention, the steps of obtaining the posture data fusion result are specifically as follows:

[0030] The trend stability analysis submodule constructs the inertia concentration index of each part into a time series according to the period, uses the Hurst exponent analysis method to analyze the fluctuation trend of each time series, determines the persistence and stability of the sequence value over time, and generates a trend stability judgment result;

[0031] The posture fusion weight calculation submodule calls the trend stability judgment result. If the current cycle is judged to be in a stable state, the dominant motion direction of the corresponding cycle in the dominant motion direction sequence is used as the fusion direction. The inertia concentration index of each part in the current cycle is extracted and normalized, and converted into a weighted coefficient for the current cycle to generate a posture fusion weight set.

[0032] The posture data fusion generation submodule calls the posture fusion weight set, performs superposition processing according to the current cycle value of the three-axis acceleration vector and the three-axis angular velocity vector of each part according to the fusion weight, and generates a posture data fusion result.

[0033] A data fusion processing method for a wireless attitude sensor, the data fusion processing method for the wireless attitude sensor being executed based on the data fusion processing system of the wireless attitude sensor, comprises the following steps:

[0034] S1: Collect the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. Calculate the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence. Also, calculate the change amplitude of the three-axis angular velocity vector at each location over consecutive periods to form an angular velocity change amplitude sequence.

[0035] S2: Identifying the dominant motion direction sequence of the entire exoskeleton manipulator based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part;

[0036] S3: Calculating a motion trend index of the exoskeleton manipulator according to the temporal trend of the dominant motion direction sequence, and updating a posture reference direction set according to the index;

[0037] S4: Based on the posture reference direction set, project the three-axis angular velocity and the three-axis velocity vector to the dominant motion direction sequence, calculate the rotational inertia response value and the motion inertia response value of each part to the dominant direction, and weightedly combine them to form an inertia concentration index for each part;

[0038] S5: Analyze the trend stability of the inertia concentration index, fuse the corresponding information of each part, and obtain the posture data fusion result.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are:

[0040] During the posture information processing process, the present invention constructs a displacement direction sequence based on the linear velocity vector direction derived from the three-axis acceleration, and generates a dynamic feature set in combination with the multi-cycle three-axis angular velocity change amplitude. The overall dominant motion direction is extracted through the cosine relationship of the angle between multiple parts, and the motion direction update index is determined as the time change trend, thereby realizing the dynamic correction of the posture reference direction. On this basis, the projection analysis of the three-axis acceleration and the three-axis angular velocity in the dominant direction is introduced to form rotational and linear inertial response values ​​respectively. The inertia concentration of each part in the unified motion direction is reflected through weighted combination. The time series trend analysis method is further introduced to improve the accuracy of the response index stability judgment. Finally, based on the inertia concentration change trend, the fusion weight of the posture data is dynamically adjusted, and a more coordinated and accurate fusion output is achieved when the information of multiple parts is greatly different, thereby improving the continuity, direction uniformity and multi-source data coordination of spatial posture estimation in complex dynamic scenes as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a system flow chart of the present invention;

[0042] Figure 2 A flow chart of the sequence building blocks of the present invention;

[0043] Figure 3 This is a flow chart of the leading identification module of the present invention;

[0044] Figure 4 This is a flow chart of the trend calculation module of the present invention;

[0045] Figure 5 is a flow chart of the inertia analysis module of the present invention;

[0046] Figure 6 This is a flow chart of the fusion output module of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0049] See also Figure 1 The present invention provides a technical solution: a data fusion processing system for a wireless attitude sensor includes:

[0050] The sequence construction module collects the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. It calculates the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence, and calculates the change amplitude of the three-axis angular velocity vector of each location in consecutive periods to form an angular velocity change amplitude sequence.

[0051] The dominant recognition module identifies the dominant motion direction sequence of the exoskeleton manipulator as a whole based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part;

[0052] The trend calculation module calculates the motion trend index of the exoskeleton manipulator according to the temporal trend of the dominant motion direction sequence, and updates the posture reference direction set according to the index;

[0053] The inertial analysis module projects the three-axis angular velocity and the three-axis velocity vectors onto the dominant motion direction sequence based on the attitude reference direction set. It calculates the rotational inertial response value and the motion inertial response value of each part to the dominant direction and weights them together to form the inertia concentration index of each part.

[0054] The fusion output module analyzes the trend stability of the inertia concentration index, integrates the corresponding information of each part, and obtains the posture data fusion result;

[0055] The displacement direction vector sequence includes the linear velocity direction information of each part, the spatial direction relationship, and the direction continuity within the cycle. The dominant motion direction sequence specifically includes the overall motion direction vector, the unified direction attribute, and the cross-part direction fusion feature. The posture reference direction set includes the current reference direction, the historical direction update record, and the dominant direction replacement basis. The inertia concentration index includes the rotational response value, the linear response value, and the weighted response strength. The posture data fusion result specifically refers to the fused acceleration information, the fused angular velocity information, and the unified output posture direction.

[0056] See also Figure 2 , the steps for obtaining the displacement direction vector sequence and the angular velocity change amplitude sequence are as follows:

[0057] The three-axis acceleration calculation submodule collects the three-axis acceleration of the wireless posture sensors in multiple parts of the exoskeleton manipulator within a specified period, performs bandpass filtering on the three-axis acceleration collected in each period of the wrist, elbow, and shoulder, integrates the acceleration of each axis in time sequence, calculates the three-axis velocity vector, and generates a three-axis velocity vector set;

[0058] First, within a fixed sampling period, the acceleration values ​​in the X, Y, and Z directions are collected in sequence from the wireless attitude sensors on the wrist, elbow, and shoulder of the exoskeleton manipulator. Each direction forms its own acceleration sequence that changes with time. If the sampling period is set to once every 0.05 seconds, 20 sets of data points can be obtained within 1 second. Subsequently, the acceleration data in these three directions are band-pass filtered separately, and the band-pass range is set to, for example, 0.5 Hz to 5 Hz. The high-frequency and low-frequency components are removed by digital filtering algorithms such as Butterworth or moving average filtering, and only the signal within the main motion frequency band is retained. After the filtering is completed, the acceleration sequence in each direction is time-converted. Time integration processing, that is, the speed at each time point is equal to the speed at the previous moment plus the current acceleration multiplied by the time interval. The integration process is expressed as: speed at the first moment + current acceleration × time interval = speed at the second moment. If integration continues, then speed at the second moment + acceleration at the next moment × time interval = speed at the third moment, and this is repeated iteratively. After completing the integration for each of the X, Y, and Z directions, a linear velocity sequence in the three directions is obtained. The three directional velocities at the same time point are then combined to form a three-axis velocity vector, which represents the velocity state in space at that moment. Finally, the linear velocity vectors of all time points form a linear velocity vector set in sequence.

[0059] The spatial direction acquisition submodule calculates the modulus of each three-axis velocity vector in the three-axis velocity vector set and performs normalization processing, extracts the spatial direction of the vector and arranges it in chronological order to generate a displacement direction vector sequence;

[0060] Extract the direction information of each time point from the three-axis velocity vector set. The processing steps begin by obtaining the linear velocity vector at a certain time point. This vector contains the velocity values ​​in three directions. First, the modulus length of this vector needs to be calculated using the modulus length calculation formula: , expressed as the square root of the sum of the squared velocities along the three axes. The modulus represents the length of the vector in three-dimensional space, i.e., the overall velocity magnitude at that moment. After obtaining the modulus, the velocity components in each direction are normalized as follows: X-direction velocity ÷ modulus = unit X-direction component, Y-direction velocity ÷ modulus = unit Y-direction component, and Z-direction velocity ÷ modulus = unit Z-direction component. This process scales the vectors to unit vectors of length 1, retaining only the direction information without considering the magnitude. After processing all time points, all unit vectors are arranged in chronological order to form a sequence of displacement direction vectors.

[0061] The angular velocity variation calculation submodule collects the three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator in continuous cycles, extracts the three-axis angular velocity vectors of the wrist, elbow, and shoulder in two adjacent cycles, and calculates the Euclidean distance between the three-axis angular velocity vectors between adjacent cycles as the variation amplitude of each location, generating an angular velocity variation amplitude sequence.

[0062] The exoskeleton collects three-axis angular velocity data in continuous cycles, corresponding to the wrist, elbow, and shoulder. The angular velocity is collected once in each cycle, including the X, Y, and Z directions. The data of two adjacent cycles are combined into a difference pair to calculate the angular velocity change amplitude. The calculation method is:

[0063] ;

[0064] in, : Indicates the angular velocity in the X direction of the current cycle, : Indicates the angular velocity in the Y direction of the current cycle, : Indicates the angular velocity in the Z direction of the current cycle, : represents the angular velocity in the X direction of the previous cycle, : represents the angular velocity in the Y direction of the previous cycle, : Indicates the angular velocity in the Z direction of the previous cycle.

[0065] The degree of angular velocity change between two cycles is obtained. This calculation operation is repeated for all consecutive cycles. For example, if 10 cycles are sampled, 9 pairs of periodic data are formed. The above calculation is performed once for each pair, and finally 9 angular velocity change amplitude values ​​are obtained, forming a change amplitude sequence.

[0066] See also Figure 3 , the steps for obtaining the dominant motion direction sequence are as follows:

[0067] The directional feature construction submodule extracts the displacement direction vector and angular velocity change amplitude in the corresponding period according to the displacement direction vector sequence and angular velocity change amplitude sequence of each part, and generates a directional feature set;

[0068] Combining the sequence of displacement direction vectors and angular velocity change amplitudes for each part, a representative set of directional features is extracted for each cycle. The processing begins by iterating through each cycle. Within each motion cycle, the unit displacement direction vector and the corresponding angular velocity change amplitude for the corresponding part are obtained. The direction vector is the unit vector calculated in the previous module and represents the main direction of the motion during that cycle; the angular velocity change amplitude indicates the degree of posture adjustment of the part during that cycle. Each pair of direction vector and change amplitude corresponds to a directional feature combination at a specific time point. For example, in the fifth cycle of the elbow, the fifth unit direction vector of the elbow and its corresponding angular velocity change amplitude are extracted and combined into a structure or vector pair. This process is repeated multiple times for each part throughout the sampling process, constructing a directional feature unit in each time cycle, ultimately forming a directional feature set for that part.

[0069] The angle cosine calculation submodule calls the displacement direction vector of each part in the directional feature set, calculates the cosine values ​​of the angles between the wrist and elbow, wrist and shoulder, and elbow and shoulder, and calculates the average value of the three sets of angle cosine values ​​in each cycle to generate a sequence of angle cosine average values;

[0070] The consistency of the movement direction between different parts is measured by the cosine value of the angle between the direction vectors. In the specific processing, the unit displacement direction vectors of three key parts (wrist, elbow, shoulder) in each cycle are selected, and the cosine value of the angle between each two combinations of these three is calculated. The formula for calculating the cosine of the angle is: , where the numerator is the dot product of two direction vectors, calculated as: The denominator is the product of the two vector moduli: the square root of the sum of the squared directional components, then multiplication. This formula is used to calculate the cosine of the directional angles between the wrist and elbow, wrist and shoulder, and elbow and shoulder, yielding three cosine values ​​for each cycle. After calculating these three sets of values, sum them up (÷3 = average). This average is the overall quantitative expression of the directional consistency between the three locations in the current cycle. The average values ​​across all cycles form a sequence of angle cosine averages.

[0071] The dominant motion direction sequence extraction submodule selects the direction vector associated with the maximum average value in each cycle of the angle cosine average value sequence as the dominant motion direction sequence of the current cycle, and arranges the dominant directions of all cycles in chronological order to generate the dominant motion direction sequence;

[0072] Based on the sequence of average cosine angle values, the dominant direction of movement in each cycle is determined. This process involves iterating through the average cosine angle values ​​for each cycle, examining the corresponding average values ​​for the three sets of direction vectors, and finding the set with the largest value. For example, if the average cosine angle value between the wrist and elbow is the largest in a given cycle, this indicates that the directions of these two parts are most consistent. In this case, the direction vector with the largest change in direction in either the wrist or elbow is selected as the dominant direction. The selection method is to compare the direction corresponding to the maximum value among the three sets of average values; the direction corresponding to the maximum value is the dominant direction for the current cycle. The dominant directions of each cycle are arranged in chronological order to construct a sequence of dominant movement directions. This sequence can be used to identify the primary force-generating or guiding element at each stage of a complete movement. For example, in a continuous throw or pass, the shoulder may dominate in the early stages, the elbow gradually intervenes, and the wrist direction becomes the dominant direction in the later stages. This sequence has significant application value in scenarios such as movement decomposition, physical modeling, and rehabilitation assessment, providing a stable timeline reference for motion control strategies.

[0073] See also Figure 4 , the steps for obtaining the attitude reference direction set are as follows:

[0074] The dominant direction angle calculation submodule extracts the dominant motion direction vectors of consecutive periods of the dominant motion direction sequence, calculates the spatial angle between the dominant motion direction vectors of two adjacent periods in time sequence, and generates a dominant direction angle sequence;

[0075] The direction vectors of each consecutive cycle in the dominant motion direction sequence are extracted and used to calculate the spatial angle between the dominant directions of two adjacent cycles, generating a time-ordered dominant direction angle sequence. The process first traverses the entire dominant direction sequence, extracting the dominant direction vectors of the current cycle and the previous cycle one by one as a set of inputs. In each set, both vectors are unit vectors with normalized directions, representing the dominant direction of human motion in that cycle. Next, the angle calculation method is used, with the formula:

[0076] , because the vector has been normalized, the modulus is 1, the denominator product is 1, and the angle cosine This is the vector dot product value. The dot product calculation method is: The cosine value is directly obtained from the dot product, and then the inverse cosine function is used to convert the cosine value into a spatial angle value. After performing this operation on each pair of periodic vectors, the resulting angle values ​​are recorded and arranged in chronological order to form a dominant direction angle sequence.

[0077] The change rate generation submodule calls the dominant direction angle sequence, performs first-order difference on each cycle angle in chronological order, regards it as the increment of the angle change between each cycle, and generates the direction change rate sequence;

[0078] Based on the dominant direction angle sequence, the change in angle between each cycle and the previous cycle, known as the angle increment, is calculated. This results in a sequence of directional change rates. This is accomplished by performing a first-order difference on the angle sequence: current cycle angle minus previous cycle angle = current cycle angle change. This operation is repeated for each pair of consecutive cycle angles, and the resulting result represents the degree of change in the dominant direction within each cycle. For example, in a series of sampled movements, if the angle in one cycle is a certain value and the angle in the next cycle is a larger value, the difference is positive, indicating a continuous shift in the dominant direction. A negative angle change indicates a regression or reduction in the dominant direction, while a zero angle indicates a stable dominant direction within the consecutive cycles. The angle increments from all cycles are arranged in sequence to form a sequence of directional change rates. This sequence is used in dynamic analysis to describe the intensity of directional change trends. For example, in movements such as continuous rotations or shoulder rotations, the angle change rate will exhibit large positive increments; whereas, during extended movements or static holds, the rate will approach zero.

[0079] The reference direction update submodule determines whether the angle change in the current cycle shows a continuous upward trend based on the direction change rate sequence. If the change value continues to increase within two consecutive cycles, it is confirmed that the dominant motion direction offset state is established. The dominant motion direction vector of the current cycle is used as the update basis to generate the attitude reference direction set.

[0080] The direction change rate sequence is used to determine the angle change trend within each cycle and update the reference direction set for attitude control. The basic approach is to determine whether the angle increments between two cycles show a sustained upward trend. This means that the angle change value of the current cycle > the angle change value of the previous cycle, which indicates an increasing trend. If the angle change values ​​in two consecutive cycles are both greater than those in the previous cycle (i.e., current change value minus previous change value > 0 and previous change value minus previous change value > 0), the direction deviation trend is considered stable. Under these conditions, the dominant direction vector of the current cycle is used as a reference for attitude updates and included in the attitude reference direction set. This indicates that the dominant direction of the system's motion has shifted or turned, and this direction needs to be reset as the reference for judgment or control. For example, in an arm-lifting and rotating combination, if the original dominant direction is shoulder forward and the angle change rate increases continuously to a certain threshold as the movement shifts to the side, the current direction vector is included in the reference direction set. This processing logic allows the system to dynamically adjust the reference standard to adapt to the deviation of the movement direction during continuous actions. Ultimately, the posture reference direction set records all key dominant directions that have been confirmed to have deviation trends.

[0081] See also Figure 5 , the specific steps for obtaining the inertia concentration index are:

[0082] The angular velocity projection calculation submodule projects the three-axis angular velocity vectors onto the dominant motion direction of the corresponding period in the dominant motion direction sequence according to the posture reference direction set. It extracts the angular velocity components of the wrist, elbow, and shoulder in the target direction as the rotational inertia response value of each part and generates a rotational inertia response value sequence.

[0083] The three-axis angular velocity vector of each body part is projected onto the dominant direction determined for each cycle in the dominant motion direction sequence based on the set of posture reference directions. This allows the angular velocity component of each body part in the dominant direction to be extracted. The process begins with each time period, obtaining the three-axis angular velocity values ​​of the wrist, elbow, and shoulder for that period, along with the dominant direction vector corresponding to that period. The three-axis angular velocity vectors are then projected onto that dominant direction. This process is implemented using a dot product, expressed as: X-axis angular velocity × dominant direction X component + Y-axis angular velocity × dominant direction Y component + Z-axis angular velocity × dominant direction Z component = projection value. The result is the rotational component of the current body part in that dominant direction, expressed as the rotational inertial response value in that direction. This operation is performed separately for the wrist, elbow, and shoulder, generating a value for each body part that reflects the strength of its rotational response in that direction. This process is repeated for all time periods, and these values ​​are arranged in sequence to form a sequence of rotational inertial response values ​​for each body part.

[0084] The linear velocity projection calculation submodule calls the posture reference direction set and projects the three-axis velocity vectors to the dominant motion direction of the corresponding cycle in the dominant motion direction sequence according to the body part. It extracts the linear velocity change rate of the wrist, elbow, and shoulder in the target direction as the motion inertia response value of each body part and generates a motion inertia response value sequence.

[0085] The three-axis velocity vectors are projected onto the dominant motion direction for each part of the movement, and the linear velocity component of each part in the target direction is extracted and used as the motion inertial response value. Specifically, the posture reference direction set and the dominant motion direction sequence are first called to process the linear velocity data for each part in the current cycle, obtaining linear velocity values ​​in the X, Y, and Z directions. These three components are then dot-producted with the dominant direction vector using the following calculation: X-axis linear velocity × dominant direction X component + Y-axis linear velocity × dominant direction Y component + Z-axis linear velocity × dominant direction Z component = projection value. This projection value is the velocity change component of the part in the dominant direction of the current cycle, also known as its motion inertial response value. This value represents the actual linear velocity trend of the part along the dominant direction during the movement and reveals the linear contribution of the part to the overall dominant motion. After completing the projection for each part, the results for the corresponding cycles are recorded sequentially to obtain the linear velocity projection sequence for the wrist, elbow, and shoulder, which is also known as the motion inertial response value sequence.

[0086] The inertia concentration generation submodule calls the rotational inertia response value sequence and the motion inertia response value sequence, performs a weighted combination of the response values ​​of the same period in the two sequences by position, synthesizes the inertial response aggregation results of each position in the dominant motion direction, and generates the inertia concentration index;

[0087] The rotational inertia response value and the motion inertia response value of each part within a certain period are integrated to form a numerical index reflecting the responsiveness of the part to the dominant motion direction. This index is calculated using a normalized weighted combination to unify the dimensions of the two types of response values, and a harmonic term is introduced to improve the ability to identify nonlinear coupling. The calculation formula for inertia concentration is:

[0088] ;

[0089] in, :Indicates the Parts (such as wrist, elbow, shoulder) in the first The inertia concentration value of the cycle; :Indicates the The part in The normalized result of the rotational inertia response value of the cycle is calculated as the original rotation response value Divide by the maximum rotation response value of the part in the full cycle ,Right now: , :Indicates the The part in The normalized result of the motion inertia response value of each cycle is calculated as the original motion response value Divide by the maximum motion response value of the part in the full cycle ,Right now: , : is the weighting coefficient of the rotation response term, and its value is set to 0.45; : is the weighting coefficient of the motion response term, and its value is set to 0.45; : is the weighting coefficient of the harmonic term of rotation and motion response value, and its value is set to 0.10; :It is a very small constant to prevent the denominator from causing an error when it is zero. Its value is The weight setting is based on the fact that the rotation response and motion response have equal dimensions and are equally important, so , ensuring that the two items have a balanced weight when there is no significant deviation; the harmonic term is used to enhance the sensitivity of the joint amplification of the two, so a smaller The guidance is reflected when coupling occurs but does not dominate the overall value, and the sum of the weights is controlled to be 1.0 to maintain normalization.

[0090] Let the original rotation response value be: , the maximum value of rotation: , original motion response value: , maximum value of motion: , the normalized result is: .

[0091] Substitute into the main formula to calculate:

[0092] Item 1: ;

[0093] Item 2: ;

[0094] Item 3: ;

[0095] Final result: .

[0096] This value represents the elbow's inertia concentration fusion value in the dominant direction of the cycle, indicating the degree to which this part integrates the concentrated response of rotational and linear inertia in that direction. This entire process is repeated per part and per cycle to form a complete inertia concentration index.

[0097] First, normalization is performed by dividing both responses by their respective maximum values ​​over the full cycle, eliminating dimensional differences and unifying the numerical scale. The two normalized response values ​​reflect the relative rotational and linear movement capabilities of the part along the dominant motion direction within that cycle, respectively. Subsequently, the normalized rotational and motion responses are weighted and summed according to predefined weights to form the two main contribution components, describing the part's direct inertial contribution to the dominant direction. A harmonic term is also introduced, consisting of the product of the two responses divided by their sum, multiplied by a harmonic weight, to characterize the coupling strength between the rotational and linear responses. This term is significant when both are large and has little influence when they are inconsistent. The three terms are combined to form the final inertial concentration value, which represents the comprehensive inertial response of the part in the dominant motion direction of the current cycle. A larger value indicates a greater contribution to the part's contribution to the dominant motion direction during that cycle. This entire process is processed cycle by cycle, ultimately generating a time series of inertial concentration values ​​for each part, which can be used for subsequent identification of the dominant part, motion dominance analysis, or dynamic collaborative modeling.

[0098] See also Figure 6 ,The specific steps for obtaining the posture data fusion results are:

[0099] The trend stability analysis submodule constructs the inertia concentration index of each part into a time series according to the period, uses the Hurst exponent analysis method to analyze the fluctuation trend of each time series, judges the persistence and stability of the sequence value over time, and generates the trend stability judgment result;

[0100] The long-term dependence analysis of the inertia concentration time series of each part is performed. The Hurst index is calculated to evaluate the persistence of the fluctuation of the series in the time dimension, and to determine whether it shows a stable trend, a reversal trend or irregular fluctuations. The core calculation method of this submodule is Hurst index analysis. The Hurst index calculation formula is as follows:

[0101] ;

[0102] in, : represents Hurst exponent; : Indicates the length of each segment in the time series (window size); : is the range of the segment, that is, the difference between the maximum cumulative deviation and the minimum cumulative deviation; : is the standard deviation of the segment; : is the standardized value of the range of this segment, that is, the re-scaled range; : is the natural logarithm function.

[0103] The theoretical explanation of the Hurst exponent is:

[0104] like : Indicates that there is a positive correlation in the inertia concentration time series, that is, the current trend will continue and the series is stable; if : Indicates that the sequence shows a reversal trend, that is, the numerical fluctuations show self-regulation and the trend is unstable; if : Indicates that the sequence is a purely random process with no significant trend characteristics.

[0105] Assume that the normalized value of the elbow inertia concentration time series is: .

[0106] Step 1: Calculate the mean:

[0107] ;

[0108] Step 2: Calculate the deviation sequence:

[0109] ;

[0110] Step 3: Construct the deviation accumulation sequence:

[0111] ;

[0112] Step 4: Calculate the range:

[0113] ;

[0114] Step 5: Calculate the sample standard deviation:

[0115] ;

[0116] Step 6: Calculate the rescaled range:

[0117] ;

[0118] Step 7: Calculate the Hurst exponent:

[0119] .

[0120] The final calculated Hurst index is 0.2562, indicating that the time series of the inertial concentration of the elbow during this period lacks stability and persistence, and exhibits reversal fluctuation characteristics.

[0121] The posture fusion weight calculation submodule calls the trend stability judgment result. If the current cycle is judged to be in a stable state, the dominant motion direction of the corresponding cycle in the dominant motion direction sequence is used as the fusion direction. The inertia concentration index of each part in the current cycle is extracted and normalized, converted into a weighted coefficient for the current cycle, and a posture fusion weight set is generated.

[0122] Based on the trend stability assessment results, and provided the inertia concentration time series for the current cycle is stable, the dominant direction vector in the dominant motion direction sequence for the current cycle is selected as the unified direction for fusion of posture contributions from all parts. The inertia concentration value for each part in the current cycle is then extracted and normalized, with the normalized result serving as the weighting coefficient for posture fusion in the current cycle. The process begins by dividing the inertia concentration value for each part by the maximum inertia concentration value across all parts in the current cycle. For example, if the inertia concentration values ​​for three parts in the current cycle are different, the normalization is as follows: wrist inertia concentration divided by maximum value = wrist weighting coefficient, elbow inertia concentration divided by maximum value = elbow weighting coefficient, and shoulder inertia concentration divided by maximum value = shoulder weighting coefficient. This process ensures that the weight of the most contributing part is 1, while other parts are scaled relative to each other, maintaining a proportional structure and maintaining the effectiveness of weighted fusion. Finally, these three normalized weights are sequentially combined to form the posture fusion weight set for the current cycle.

[0123] The posture data fusion generation submodule calls the posture fusion weight set, and performs superposition processing based on the current cycle values ​​of the three-axis acceleration vector and the three-axis angular velocity vector of each part according to the fusion weight to generate the posture data fusion result;

[0124] The three-axis acceleration vector and three-axis angular velocity vector for each body part in the current cycle are superimposed according to the fusion weights to form a unified posture data fusion output. Specifically, the acceleration and angular velocity components are calculated in three directions. The X-axis acceleration of each body part is multiplied by its weighting coefficient. For example, elbow X-axis acceleration × elbow weighting coefficient + shoulder X-axis acceleration × shoulder weighting coefficient + wrist X-axis acceleration × wrist weighting coefficient = the fused X-axis acceleration component. The same applies to Y and Z directions, forming the fused three-axis acceleration vector. The same approach is applied to angular velocity processing: the angular velocity of each body part is multiplied by its weighting coefficient in each of the three directions and then added together. For example, elbow Y-axis angular velocity × elbow weighting coefficient + shoulder Y-axis angular velocity × shoulder weighting coefficient + wrist Y-axis angular velocity × wrist weighting coefficient = the fused Y-axis angular velocity component. This is then combined to form the fused three-axis angular velocity vector. This fusion result represents the overall posture performance of the motion state of all parts in the dominant direction in the current cycle, which can be used in subsequent trajectory fitting, posture recognition or state feedback control to form an integrated data output of multi-source posture information in the dominant direction.

[0125] A data fusion processing method for a wireless attitude sensor is provided. The data fusion processing method for a wireless attitude sensor is performed based on the data fusion processing system of the wireless attitude sensor, and includes the following steps:

[0126] S1: Collect the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. Calculate the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence. Also, calculate the change amplitude of the three-axis angular velocity vector at each location over consecutive periods to form an angular velocity change amplitude sequence.

[0127] S2: Identify the dominant motion direction sequence of the exoskeleton manipulator as a whole based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part;

[0128] S3: Calculate the motion trend index of the exoskeleton manipulator based on the temporal trend of the dominant motion direction sequence, and update the posture reference direction set based on the index;

[0129] S4: Based on the posture reference direction set, the three-axis angular velocity and the three-axis velocity vector are projected onto the dominant motion direction sequence, and the rotational inertia response value and motion inertia response value of each part to the dominant direction are calculated and weighted and combined to form the inertia concentration index of each part;

[0130] S5: Analyze the trend stability of the inertia concentration index, fuse the corresponding information of each part, and obtain the posture data fusion result.

[0131] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A data fusion processing system for a wireless attitude sensor, characterized in that: The system comprises: The sequence construction module collects the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. It calculates the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence, and calculates the change amplitude of the three-axis angular velocity vector of each location in consecutive periods to form an angular velocity change amplitude sequence. A dominant identification module identifies the dominant motion direction sequence of the entire exoskeleton manipulator based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part; a trend calculation module, which calculates a motion trend index of the exoskeleton manipulator according to the temporal trend of the dominant motion direction sequence, and updates a posture reference direction set according to the index; an inertial analysis module, which projects the three-axis angular velocity and the three-axis velocity vector onto the dominant motion direction sequence based on the posture reference direction set, calculates the rotational inertia response value and the motion inertia response value of each part to the dominant direction, and weights and combines them to form an inertia concentration index for each part; The fusion output module analyzes the trend stability of the inertia concentration index, fuses the corresponding information of each part, and obtains the posture data fusion result.

2. The data fusion processing system of the wireless attitude sensor according to claim 1, characterized in that: The displacement direction vector sequence includes the linear velocity direction information of each part, the spatial direction relationship, and the direction continuity within the cycle. The dominant motion direction sequence specifically includes the overall motion direction vector, the unified direction attribute, and the cross-part direction fusion feature. The posture reference direction set includes the current reference direction, the historical direction update record, and the dominant direction replacement basis. The inertia concentration index includes the rotational response value, the linear response value, and the weighted response strength. The posture data fusion result specifically refers to the fused acceleration information, the fused angular velocity information, and the unified output posture direction.

3. The data fusion processing system of the wireless attitude sensor according to claim 1, characterized in that: The steps for obtaining the displacement direction vector sequence and the angular velocity change amplitude sequence are specifically as follows: The three-axis acceleration calculation submodule collects the three-axis acceleration of the wireless posture sensors in multiple parts of the exoskeleton manipulator within a specified period, performs bandpass filtering on the three-axis acceleration collected in each period of the wrist, elbow, and shoulder, integrates the acceleration of each axis in time sequence, calculates the three-axis velocity vector, and generates a three-axis velocity vector set; The spatial direction acquisition submodule calculates the modulus of each three-axis velocity vector in the three-axis velocity vector set and performs normalization processing, extracts the spatial direction of the vector and arranges it in chronological order to generate a displacement direction vector sequence; The angular velocity variation amplitude calculation submodule collects the three-axis angular velocities of wireless posture sensors at multiple locations of the exoskeleton manipulator in continuous cycles, extracts the three-axis angular velocity vectors of the wrist, elbow, and shoulder in two adjacent cycles, calculates the Euclidean distance of the three-axis angular velocity vectors between adjacent cycles as the variation amplitude of each location, and generates an angular velocity variation amplitude sequence.

4. The data fusion processing system for wireless attitude sensors according to claim 3, characterized in that: The steps for obtaining the dominant motion direction sequence are specifically as follows: A directional feature construction submodule extracts the displacement direction vector and angular velocity change amplitude within the corresponding period according to the displacement direction vector sequence and angular velocity change amplitude sequence of each part, and generates a directional feature set; The angle cosine value calculation submodule calls the displacement direction vector of each part in the directional feature set, calculates the cosine values ​​of the angles between the wrist and elbow, wrist and shoulder, and elbow and shoulder, and calculates the average value of the three sets of angle cosine values ​​in each cycle to generate an angle cosine average value sequence; The dominant motion direction sequence extraction submodule selects the direction vector associated with the maximum average value of each cycle in the angle cosine average value sequence as the dominant motion direction sequence of the current cycle, arranges the dominant directions of all cycles in chronological order, and generates a dominant motion direction sequence.

5. The data fusion processing system for wireless attitude sensors according to claim 4, characterized in that: The steps for obtaining the attitude reference direction set are specifically as follows: The dominant direction angle calculation submodule extracts the dominant motion direction vectors of consecutive periods of the dominant motion direction sequence, calculates the spatial angle between the dominant motion direction vectors of two adjacent periods in chronological order, and generates a dominant direction angle sequence; The change rate generation submodule calls the dominant direction angle sequence, performs first-order difference on each cycle angle in chronological order, regards it as the increment of the angle change between each cycle, and generates a direction change rate sequence; The reference direction update submodule determines whether the angle change in the current cycle shows a continuous upward trend based on the direction change rate sequence. If the change value continues to increase within two consecutive cycles, it is confirmed that the dominant motion direction offset state is established, and the dominant motion direction vector of the current cycle is called as the update basis to generate the posture reference direction set.

6. The data fusion processing system for wireless attitude sensors according to claim 5, characterized in that: The steps for obtaining the inertia concentration index are specifically as follows: The angular velocity projection calculation submodule projects the three-axis angular velocity vectors onto the dominant motion direction of the corresponding period in the dominant motion direction sequence according to the posture reference direction set, extracts the angular velocity components of the wrist, elbow, and shoulder in the target direction as the rotational inertia response value of each part, and generates a rotational inertia response value sequence; The linear velocity projection calculation submodule calls the posture reference direction set, projects the three-axis velocity vectors onto the dominant motion direction of the corresponding period in the dominant motion direction sequence according to the body part, extracts the linear velocity change rate of the wrist, elbow, and shoulder in the target direction as the motion inertia response value of each body part, and generates a motion inertia response value sequence; The inertia concentration generation submodule calls the rotational inertia response value sequence and the motion inertia response value sequence, performs a weighted combination of the response values ​​of the same period in the two sequences by part, synthesizes the inertia response aggregation result of each part in the dominant motion direction, and generates an inertia concentration index.

7. The data fusion processing system for wireless attitude sensors according to claim 6, characterized in that: The steps for obtaining the posture data fusion result are specifically as follows: The trend stability analysis submodule constructs the inertia concentration index of each part into a time series according to the period, uses the Hurst exponent analysis method to analyze the fluctuation trend of each time series, determines the persistence and stability of the sequence value over time, and generates a trend stability judgment result; The posture fusion weight calculation submodule calls the trend stability judgment result. If the current cycle is judged to be in a stable state, the dominant motion direction of the corresponding cycle in the dominant motion direction sequence is used as the fusion direction. The inertia concentration index of each part in the current cycle is extracted and normalized, and converted into a weighted coefficient for the current cycle to generate a posture fusion weight set. The posture data fusion generation submodule calls the posture fusion weight set, performs superposition processing according to the current cycle value of the three-axis acceleration vector and the three-axis angular velocity vector of each part according to the fusion weight, and generates a posture data fusion result.

8. A data fusion processing method for a wireless attitude sensor, characterized in that: The data fusion processing system of the wireless attitude sensor according to any one of claims 1 to 7 is executed, comprising the following steps: S1: Collect the three-axis acceleration and three-axis angular velocity of the wireless posture sensors at multiple locations of the exoskeleton manipulator within a specified period. Calculate the spatial direction of the three-axis velocity vector based on the three-axis acceleration to obtain a displacement direction vector sequence. Also, calculate the change amplitude of the three-axis angular velocity vector at each location over consecutive periods to form an angular velocity change amplitude sequence. S2: Identifying the dominant motion direction sequence of the entire exoskeleton manipulator based on the displacement direction vector sequence and angular velocity change amplitude sequence of each part; S3: Calculating a motion trend index of the exoskeleton manipulator according to the temporal trend of the dominant motion direction sequence, and updating a posture reference direction set according to the index; S4: Based on the posture reference direction set, project the three-axis angular velocity and the three-axis velocity vector to the dominant motion direction sequence, calculate the rotational inertia response value and the motion inertia response value of each part to the dominant direction, and weightedly combine them to form an inertia concentration index for each part; S5: Analyze the trend stability of the inertia concentration index, fuse the corresponding information of each part, and obtain the posture data fusion result.

Images