A health and wellness service robot interaction method based on sound recognition

Abstract

The application relates to the technical field of robots and discloses a health-care service robot interaction method based on sound recognition, which comprises the following steps: fixing a double-ring vector cuff composed of an outer ring and an inner ring at the elbow of a target patient; preprocessing an acoustic and electric signal of the double-ring vector cuff, deducting a baseline acoustic and electric signal, and obtaining an analytic acoustic and electric signal; constructing a sampling value of a boundary complex value function based on the analytic acoustic and electric signal; substituting the boundary complex value function into a boundary integral model of a complex potential acoustic potential to obtain a complex potential acoustic potential of a rotation center of an elbow joint of the target patient; extracting a stress amplitude, a main direction angle and a radial gradient from the complex potential acoustic potential, and respectively determining virtual stiffness, damping and a fractional order based on the stress amplitude, the main direction angle and the radial gradient; generating an output torque by applying a fractional order impedance law based on the virtual stiffness, the damping and the fractional order, and driving a target health-care robot to perform an elbow auxiliary stretching operation based on the output torque.

Description

Technical Field

[0001] This invention relates to the field of robotics, and more specifically, to an interaction method for health and wellness service robots based on voice recognition. Background Technology

[0002] In recent years, with the increasing trend of population aging and the rapid growth of rehabilitation needs, rehabilitation service robots have been gradually promoted and applied in areas such as intelligent assisted rehabilitation, limb function reconstruction, and home health monitoring. Among them, the use of robots for passive joint stretching and training for patients with elbow flexion and extension dysfunction has become an important research direction in the field of rehabilitation engineering.

[0003] Existing elbow rehabilitation robots mostly use angle control or force control for assisted stretching, and their control parameters usually rely on fixed preset trajectories or limited electromyographic signal input. However, electromyographic signals are greatly affected by the patient's physiological condition and are not applicable to some severely disabled or poorly cooperative patients.

[0004] Therefore, existing technologies suffer from problems such as complex signal processing, high dependence on physical modeling, and weak real-time performance. Thus, there is an urgent need for a sensing and interaction method that can directly acquire acoustic and electrical response signals from the joint region and dynamically couple them with motion control parameters to improve the adaptability and intelligent control level of robots in elbow stretching rehabilitation. Summary of the Invention

[0005] This invention provides an interaction method for health and wellness service robots based on voice recognition, which solves the technical problems mentioned in the background art.

[0006] This invention provides a voice recognition-based interaction method for elderly care service robots, comprising:

[0007] Step 1: Fix a double-ring vector cuff consisting of an outer ring and an inner ring to the elbow of the target patient. The double-ring vector cuff is a double-ring signal acquisition device that can combine multiple piezoelectric pads in a concentric double-ring structure to simultaneously acquire the amplitude and spatial orientation characteristics of the acoustic signal.

[0008] Step 2: Preprocess the acoustic-electric signal of the double-ring vector cuff and subtract the baseline acoustic-electric signal to obtain the analytical acoustic-electric signal;

[0009] Step 3: Construct sampled values ​​of the boundary complex-valued function based on the analytical acoustic-electric signal;

[0010] Step 4: Substitute the boundary complex-valued function into the boundary integral model of the complex acoustic potential to obtain the complex acoustic potential at the center of rotation of the target patient's elbow joint. The complex acoustic potential is a complex numerical field function constructed by mapping the acoustic signals distributed on the boundary to the complex plane and using the integration of complex variable functions to characterize the acoustic energy concentration intensity and spatial phase characteristics at the center node.

[0011] Step 5: Extract the stress-like amplitude, principal direction angle and radial gradient from the complex acoustic potential, and determine the virtual stiffness, damping and fractional order based on the stress-like amplitude, principal direction angle and radial gradient respectively. The stress-like amplitude is the modulus of the complex acoustic potential used to reflect the overall strength of the acoustic field at the elbow joint rotation center.

[0012] Step 6: Based on the virtual stiffness, damping and fractional order, the fractional order impedance law is applied to generate the output torque, and the target health care robot is driven to perform elbow-assisted stretching operation based on the output torque.

[0013] Furthermore, the double-ring vector cuff includes:

[0014] With the center of elbow joint rotation as the center C, and radii R1 and R2 respectively, construct an outer ring and an inner ring, where R1 > R2;

[0015] The outer ring consists of six piezoelectric pads with a radius of R1 and polar angles of [missing information]. , , , , , It fits the bony ring-shaped area formed by the olecranon process and the medial and lateral epicondyles;

[0016] The inner ring consists of six piezoelectric pads with a radius of R2. The polar angles of each pad are offset clockwise by 30° relative to the polar angles of the corresponding piezoelectric pads of the outer ring, fitting the surface of the joint capsule in the elbow fossa region.

[0017] Furthermore, the acoustic-electric signals of the double-ring vector cuff are preprocessed, and the baseline acoustic-electric signals are subtracted to obtain the analytical acoustic-electric signals, including:

[0018] Step 21, preprocessing, includes: analog-to-digital conversion, bandpass filtering, and moving average;

[0019] Step 22: Perform Hilbert transform on the differential signal formed by subtracting the baseline acoustic signal from the preprocessed acoustic signal to obtain the analytical acoustic signal. , ;

[0020] Step 23, the baseline acoustic-electric signal is acquired as follows:

[0021] At each sampling moment within the preset time period, the target patient was controlled to keep their elbow at a certain position. In the flexed state, acquire the acoustic-electric signal of the double-loop vector cuff, and repeat step 21;

[0022] The average value of the preprocessed acoustic-electric signals corresponding to all sampling times within a preset time period is taken as the baseline acoustic-electric signal.

[0023] Furthermore, the sampled values ​​of the boundary complex-valued function are constructed based on the analytical acoustic-electric signal, including:

[0024] Step 31, define the coordinates of the center C as: Where i represents the imaginary unit, Construct a complex plane coordinate system with center C.

[0025] Step 32: In the complex plane coordinate system, the coordinates of the 12 piezoelectric pads are numbered to form a piezoelectric pad coordinate cyclic sequence. , , This represents the coordinates of the k-th piezoelectric pad. This represents the polar angle of the k-th piezoelectric pad. Represents the natural base;

[0026] Step 33: Calculate the difference vector for adjacent piezoelectric pad coordinates in the piezoelectric pad coordinate cyclic sequence. , ;

[0027] Step 34: Determine if the sum of all difference vectors is 0, then... and Assigned to This forms a set of sampled values ​​for the boundary complex-valued function. Otherwise, return to step 32 to renumber.

[0028] Furthermore, by substituting the boundary complex-valued function into the boundary integral model of the complex acoustic potential, the complex acoustic potential at the target patient's elbow joint rotation center is obtained, including:

[0029] Sampled values ​​based on boundary complex functions The discrete integral kernel is calculated;

[0030] The complex acoustic potential is obtained by discretized Cauchy integral.

[0031] Furthermore, stress-like amplitude, principal direction angle, and radial gradient are extracted from the complex acoustic potential, including:

[0032] The magnitude of the complex acoustic potential is obtained by calculating the stress-like amplitude. ;

[0033] The argument of the complex acoustic potential is obtained to get the principal direction angle. ;

[0034] The radial gradient is obtained as follows:

[0035] Constructing a quadratic kernel function , ;

[0036] Calculate the radial gradient , .

[0037] Furthermore, the virtual stiffness, damping, and fractional order are determined based on the stress amplitude, principal direction angle, and radial gradient, respectively, including:

[0038] Calculate virtual stiffness , ;

[0039] Calculate damping parameters , ;

[0040] Calculate fractional order , ;

[0041] in, , , and These represent the preset minimum virtual stiffness, preset minimum damping, preset lower limit of fractional order, and preset upper limit of fractional order, respectively. and These represent the stiffness gain coefficient and the damping gain coefficient, respectively. and This represents the maximum stress amplitude and maximum radial gradient of the target patient;

[0042] The maximum stress amplitude and maximum radial gradient of the target patient were obtained as follows:

[0043] The target rehabilitation robot drives the target patient's elbow from a constant angular velocity. Flex slowly extend to Bending, then returning Complete the flexion-extension cycle;

[0044] The maximum stress amplitude and maximum radial gradient within the flexure-extension cycle are obtained by marking.

[0045] Furthermore, based on virtual stiffness, damping, and fractional order, the fractional-order impedance law is applied to generate an output torque, and this output torque drives the target rehabilitation robot to perform elbow-assisted stretching operations, including:

[0046] Determine the actual and expected angles of the target patient's elbow, and calculate the error angle;

[0047] The error angle, as well as the preset truncation order, fractional order, and sampling period of the acoustic signal are used as input variables for the Grünwald–Letnikov algorithm to calculate the fractional derivative.

[0048] The output torque of the target health and wellness robot is determined by combining the fractional derivative with virtual stiffness and damping, based on the fractional impedance law.

[0049] The target health and wellness robot is driven by output torque to perform elbow-assisted stretching operations.

[0050] The beneficial effects of this invention are as follows: By constructing a double-ring vector cuff structure with the elbow joint rotation center as the geometric center, and combining the acoustic-electric signals sampled from twelve piezoelectric pads at multiple points, a boundary complex-valued function is constructed in a complex plane coordinate system. Furthermore, a complex potential acoustic potential modeling method is introduced, achieving for the first time the dynamic coupling relationship between acoustic signals and virtual mechanical parameters (including virtual stiffness, damping, and fractional order) during joint movement. Further, by employing a fractional-order impedance control law, individualized acoustic-electric responses are mapped into dynamic torque outputs, enabling the rehabilitation service robot to achieve compliant adjustment and dynamic adaptation for elbow-assisted stretching operations, effectively improving comfort, safety, and personalized precision during the rehabilitation process. Attached Figure Description

[0051] Figure 1 This is a flowchart of an interactive method for a health and wellness service robot based on voice recognition, according to the present invention. Detailed Implementation

[0052] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0053] like Figure 1 As shown, a voice recognition-based interaction method for elderly care service robots includes:

[0054] Step 1: Fix a double-ring vector cuff consisting of an outer ring and an inner ring at the elbow of the target patient;

[0055] Step 2: Preprocess the acoustic-electric signal of the double-ring vector cuff and subtract the baseline acoustic-electric signal to obtain the analytical acoustic-electric signal;

[0056] Step 3: Construct sampled values ​​of the boundary complex-valued function based on the analytical acoustic-electric signal;

[0057] Step 4: Substitute the boundary complex-valued function into the boundary integral model of the complex acoustic potential to obtain the complex acoustic potential of the target patient's elbow joint rotation center.

[0058] Step 5: Extract the stress-like amplitude, principal direction angle and radial gradient from the complex acoustic potential, and determine the virtual stiffness, damping and fractional order based on the stress-like amplitude, principal direction angle and radial gradient respectively.

[0059] Step 6: Based on the virtual stiffness, damping and fractional order, the fractional order impedance law is applied to generate the output torque, and the target health care robot is driven to perform elbow-assisted stretching operation based on the output torque.

[0060] In one embodiment of the present invention, the double-ring vector cuff includes:

[0061] With the center of elbow joint rotation as the center C, and radii R1 and R2 respectively, construct an outer ring and an inner ring, where R1 > R2;

[0062] It should be noted that by using the elbow joint rotation center as a common positioning reference, all subsequent acoustic and electrical signal sampling is ensured to be symmetrically distributed around the same motion axis; the outer ring and the inner ring form a concentric double-ring structure, providing a basis for multipath acoustic comparison between the outer ring and the inner ring in space; the difference in size between R1 and R2 enables layered acquisition of vibrations of tissues at different depths; the double-ring vector cuff here refers to a double-ring signal acquisition device that can combine multiple piezoelectric pads in a specific spatial distribution to simultaneously acquire the amplitude and spatial orientation characteristics of acoustic signals.

[0063] The outer ring consists of six piezoelectric pads with a radius of R1 and polar angles of [missing information]. , , , , , It fits the bony ring-shaped area formed by the olecranon process and the medial and lateral epicondyles;

[0064] It should be noted that placing piezoelectric pads in the bony annular region can utilize the high conduction characteristics of bone tissue to acoustic vibrations to improve the coupling efficiency of bony acoustic wave signals; the high-energy bony vibration acoustic signals collected by the outer ring provide clear characteristics of the joint surface and bony friction.

[0065] The inner ring consists of six piezoelectric pads with a radius of R2. The polar angles of each pad are offset clockwise by 30° relative to the polar angles of the corresponding piezoelectric pads of the outer ring, fitting the surface of the joint capsule in the elbow fossa region.

[0066] It should be noted that the piezoelectric pads of the inner ring are offset by 30° to avoid the bony area, specifically capturing the vibrational acoustic signals of the elbow joint capsule and deep synovium, supplementing the blind spots of the bony signals of the outer ring.

[0067] In one embodiment of the present invention, the acoustic-electric signal of the double-loop vector cuff is preprocessed and the baseline acoustic-electric signal is subtracted to obtain the analytical acoustic-electric signal, including:

[0068] Step 21, preprocessing, includes: analog-to-digital conversion, bandpass filtering, and moving average;

[0069] It should be noted that the continuous acoustic and electrical signals of the dual-ring vector cuff are digitized through analog-to-digital conversion in order to perform subsequent processing in the control system; bandpass filtering is used to limit the signal to the key acoustic frequency band to remove low-frequency motion artifacts and high-frequency spurious noise; and then moving average is used to smooth instantaneous fluctuations, suppress residual spikes, and achieve signal amplitude stability.

[0070] Step 22: Perform Hilbert transform on the differential signal formed by subtracting the baseline acoustic signal from the preprocessed acoustic signal to obtain the analytical acoustic signal. , ;

[0071] By subtracting the pre-calculated baseline acoustic signal, a differential signal reflecting only the dynamic changes of the joint is obtained, and the static bias is removed. Then, a Hilbert transform is performed on this differential signal to construct an analytical acoustic signal containing instantaneous amplitude and phase information. .

[0072] Step 23, the baseline acoustic-electric signal is acquired as follows:

[0073] At each sampling moment within the preset time period, the target patient was controlled to keep their elbow at a certain position. In the flexed state, acquire the acoustic-electric signal of the double-loop vector cuff, and repeat step 21;

[0074] The average value of the preprocessed acoustic-electric signals corresponding to all sampling times within a preset time period is taken as the baseline acoustic-electric signal.

[0075] It should be noted that the designated patient is located at the elbow. Acquiring data while the device is in a flexed position ensures that the recorded acoustic-electric signals contain only environmental noise and static tissue coupling. Repeating step 21 and averaging all preprocessing results can eliminate occasional fluctuations and obtain a stable baseline acoustic-electric signal.

[0076] In one embodiment of the present invention, the sampling values ​​of the boundary complex-valued function are constructed based on the analytical acoustic-electric signal, including:

[0077] Step 31, define the coordinates of the center C as: Where i represents the imaginary unit, Construct a complex plane coordinate system with center C.

[0078] It should be noted that mapping the boundary sampling problem of the double-loop vector cuff to a complex plane coordinate system is necessary for subsequent boundary integration using the complex potential acoustic potential model; this is achieved by defining the center C as a complex number. And using this as the origin, a unified coordinate reference is established to ensure that all subsequent coordinate and signal mappings based on complex number operations are centered on the same reference point;

[0079] Step 32: In the complex plane coordinate system, the coordinates of the 12 piezoelectric pads are numbered to form a piezoelectric pad coordinate cyclic sequence. , , This represents the coordinates of the k-th piezoelectric pad. This represents the polar angle of the k-th piezoelectric pad. Represents the natural base;

[0080] It should be noted that discrete boundary integration requires the boundary points to be arranged in sequence to ensure the continuity and directionality of the integration path. The center coordinates of the 12 piezoelectric pads are sequentially numbered in the complex plane according to their polar angles. It completes the one-to-one mapping from boundary points to complex coordinates and generates a closed cyclic sequence, providing an explicit point order for discrete integration.

[0081] Step 33: Calculate the difference vector for adjacent piezoelectric pad coordinates in the piezoelectric pad coordinate cyclic sequence. , ;in, Indicates the first Coordinates of each piezoelectric pad;

[0082] It should be noted that discretizing the Cauchy boundary integral requires dividing the continuous boundary into a series of tiny line segments and calculating the weight of each segment; through Calculate the difference vector between adjacent boundary points to obtain the length and direction of each boundary line segment.

[0083] Step 34: Determine if the sum of all difference vectors is 0, then... and Assigned to This forms a set of sampled values ​​for the boundary complex-valued function. Otherwise, return to step 32 to renumber.

[0084] It should be noted that only when all difference vectors Only when the sum of the vectors is zero can we ensure that the boundary sequence is strictly closed in the complex plane, without interlacing or breaking; otherwise, the integration result will be distorted. Verify that the sum of all difference vectors is 0. If the verification passes, the acoustic-electric signal can be analyzed. and corresponding Assigned to This forms a set of sampled values ​​for the boundary complex-valued function. This completes the discretization and encapsulation of boundary conditions; if it fails, it returns to step 32 to rearrange the numbers until closure is achieved, ensuring the accuracy of the integration path and the consistency of the data.

[0085] In one embodiment of the present invention, the boundary complex-valued function is substituted into the boundary integral model of the complex acoustic potential to obtain the complex acoustic potential at the center of rotation of the target patient's elbow joint, including:

[0086] Sampled values ​​based on boundary complex functions The discrete integral kernel is calculated;

[0087] Wherein, the discrete integrator kernel is , ;

[0088] The complex acoustic potential is obtained by discretized Cauchy integral.

[0089] It should be noted that, in order to obtain the analytical acoustic-electric signal at the k-th piezoelectric pad... To introduce the correct weights into the boundary integral of the complex acoustic potential, the analytic acoustic signal must first be geometrically normalized, i.e., divided by the complex distance from the point to the center C. The result It also includes the signal amplitude. and corresponding boundary position The geometric information becomes the core kernel function for each boundary segment in the subsequent discretized Cauchy integration operation.

[0090] The complex acoustic potential is calculated using discretized Cauchy integrals, as follows: ;

[0091] in, The complex acoustic potential representing the center of rotation of the target patient's elbow joint.

[0092] It should be noted that continuous Cauchy integrals are difficult to implement directly at discrete sampling points; therefore, they are achieved by applying weights to each boundary segment. Its difference vector The weighted summation can efficiently approximate the complex acoustic potential at the elbow joint rotation center in discrete space. The final result is the complex acoustic potential at the target patient's elbow joint rotation center, which is the overall complex acoustic potential resulting from the superposition of signals from multiple boundary sampling points. This fully reflects the comprehensive influence of the boundary acoustic field on the elbow joint rotation center. The complex acoustic potential refers to the complex numerical field function constructed using the integral of complex variable functions after mapping the acoustic signals distributed on the boundary onto the complex plane, used to characterize the acoustic energy concentration intensity and spatial phase characteristics at the central node.

[0093] In one embodiment of the present invention, extracting stress-like amplitude, principal direction angle, and radial gradient from the complex acoustic potential includes:

[0094] The magnitude of the complex acoustic potential is obtained by calculating the stress-like amplitude. ;

[0095] It should be noted that the complex acoustic potential consists of two parts: a real part and an imaginary part. The modulus can be used to fuse these two parts of energy into a scalar energy level, which intuitively reflects the overall intensity of the acoustic field at the center of elbow joint rotation. Since this modulus plays a role similar to driving stress in a mechanical system in the subsequent fractional impedance control model, it is defined as a stress-like amplitude.

[0096] The argument of the complex acoustic potential is obtained to get the principal direction angle. ;

[0097] It should be noted that the argument reflects the principal phase direction of the complex acoustic potential in the complex plane, corresponding to the main propagation path of the acoustic energy. The obtained principal direction angle Y is used to adjust the directional weighting of the damping parameters, enhancing the ability to suppress components of the acoustic field that deviate from the principal axis.

[0098] The radial gradient is obtained as follows:

[0099] Constructing a quadratic kernel function , ;

[0100] It should be noted that dividing the discrete integrator kernel by the square of the distance can emphasize the influence of near-field depth on the acoustic signal and simulate the second-order effect of sound wave attenuation and phase change with radial direction.

[0101] Calculate the radial gradient , .

[0102] It should be noted that each quadratic kernel function With the corresponding difference vector Multiplying and summing these yields an approximation of the derivative of the acoustic potential with respect to the radial direction at the discrete boundary. The resulting radial gradient quantifies the rate of change of the acoustic field along the radial direction, providing a depth-sensitive indicator for real-time fractional-order adjustments.

[0103] In one embodiment of the present invention, the virtual stiffness, damping, and fractional order are determined based on the stress amplitude, principal direction angle, and radial gradient, respectively, including:

[0104] Calculate virtual stiffness , ;

[0105] It should be noted that the stress amplitude is... Hardness parameters mapped to mechanical resistance; by applying stress amplitude analogous to stress amplitude. and Normalization plus preset minimum stiffness And multiplied by stiffness gain factor virtual stiffness It can linearly increase with the increase of acoustic energy level, ensuring the system's adaptive stiffness adjustment for different stress levels.

[0106] Calculate damping parameters , ;

[0107] It should be noted that, based on the principal direction angle To adjust the lateral vibration suppression force. Utilizing The lateral component deviating from the primary flexion-extension axis is characterized by multiplying it by the damping gain coefficient. And add preset minimum damping , so that the damping parameters The acoustic field automatically increases when it deviates from the principal axis, thereby actively suppressing lateral sway and improving system stability.

[0108] Calculate fractional order , ;

[0109] It should be noted that the radial gradient along the radial direction of the acoustic field is required. Depth information is extracted to dynamically adjust the differential order of the impedance model. The radial gradient is then used. and Normalization, then mapping to , Interval, making the fractional order As the penetration depth of the sound wave increases, it rises and smoothly transitions between elastic and viscoelastic models, achieving impedance behavior that is more consistent with the characteristics of biological tissues.

[0110] in, , , and These represent the preset minimum virtual stiffness, preset minimum damping, preset lower limit of fractional order, and preset upper limit of fractional order, respectively. and These represent the stiffness gain coefficient and the damping gain coefficient, respectively. and This represents the maximum stress amplitude and maximum radial gradient of the target patient;

[0111] It should be noted that, , , and The stiffness gain coefficient and damping gain coefficient are both set by experts in the field of health and wellness.

[0112] The maximum stress amplitude and maximum radial gradient of the target patient were obtained as follows:

[0113] The target rehabilitation robot drives the target patient's elbow from a constant angular velocity. Flex slowly extend to Bending, then returning Complete the flexion-extension cycle;

[0114] The maximum stress amplitude and maximum radial gradient within the flexure-extension cycle are obtained by marking.

[0115] It should be noted that data is collected under dynamic motion conditions through one complete flexion-extension cycle. and The extreme values ​​are used to determine and This ensures that subsequent mapping functions are personalized for each patient.

[0116] In one embodiment of the present invention, based on virtual stiffness, damping, and fractional order, an output torque is generated using the fractional-order impedance law, and the target rehabilitation robot is driven to perform elbow-assisted stretching operations based on the output torque, including:

[0117] Determine the actual and expected angles of the target patient's elbow, and calculate the error angle;

[0118] The error angle, as well as the preset truncation order, fractional order, and sampling period of the acoustic signal are used as input variables for the Grünwald–Letnikov algorithm to calculate the fractional derivative.

[0119] The output torque of the target health and wellness robot is determined by combining the fractional derivative with virtual stiffness and damping, based on the fractional impedance law.

[0120] The target health and wellness robot is driven by output torque to perform elbow-assisted stretching operations.

[0121] It should be noted that the expected perspective is as follows:

[0122] Determine the starting time, angular velocity, and the range of flexion-extension angles for one complete flexion-extension cycle. Based on the time difference between the current time and the starting time, determine the expected angle within the flexion-extension angle range at the current time.

[0123] The error angle, as well as the preset truncation order, fractional order, and sampling period of the acoustic signal are used as input variables for the Grünwald–Letnikov algorithm to calculate the fractional derivative.

[0124] It should be noted that conventional first derivatives are insufficient to capture the memory effect and viscoelastic properties of biological tissues. Therefore, non-integer derivatives are required to enhance the model's historical dependence. The calculation formula is as follows: ;

[0125] in, Indicates angular error. Denotes the fractional derivative. Indicates the sampling period. Indicates the preset truncation order. For the index of the preset truncation order, Indicates the fractional order. Denotes the coefficients of the generalized binomial. , Indicates alternating sign factors, This indicates the nth sampling period.

[0126] The fractional derivative of the output integrates the current rate of change of deviation with the cumulative effect of the past few cycles, providing a viscoelastic feedback component for the generation of the output torque.

[0127] The output torque of the target health and wellness robot is determined by combining the fractional derivative with virtual stiffness and damping parameters, based on the fractional impedance law.

[0128] It should be noted that the assisted stretching of the elbow by the rehabilitation robot needs to consider both elastic recovery and viscoelastic dissipation to achieve a force output that is both compliant and stable. The calculation formula is as follows: ;

[0129] in, Indicates the output torque. Indicates virtual stiffness, Indicates damping.

[0130] The resulting output torque will be used by the rehabilitation robot to execute joint stretching commands, so that the mechanical response is both stiffness-driven according to the deviation and viscosity-adjusted according to the fractional derivative, thereby achieving adaptive assisted stretching of the patient's elbow.

[0131] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A voice recognition-based elderly care service robot system, characterized in that, Including double-ring vector cuffs, the interaction methods for elderly care service robots include: Step 1: Fix a double-ring vector cuff consisting of an outer ring and an inner ring to the elbow of the target patient. The double-ring vector cuff is a double-ring signal acquisition device that can combine multiple piezoelectric pads in a concentric double-ring structure to simultaneously acquire the amplitude and spatial orientation characteristics of the acoustic signal. Step 2: Preprocess the acoustic-electric signal of the double-ring vector cuff and subtract the baseline acoustic-electric signal to obtain the analytical acoustic-electric signal; Step 3: Construct sampled values ​​of the boundary complex-valued function based on the analytical acoustic-electric signal; Step 4: Substitute the boundary complex-valued function into the boundary integral model of the complex acoustic potential to obtain the complex acoustic potential at the center of rotation of the target patient's elbow joint. The complex acoustic potential is a complex numerical field function constructed by mapping the acoustic signals distributed on the boundary to the complex plane and using the integration of complex variable functions to characterize the acoustic energy concentration intensity and spatial phase characteristics at the center node. Step 5: Extract the stress-like amplitude, principal direction angle and radial gradient from the complex acoustic potential, and determine the virtual stiffness, damping and fractional order based on the stress-like amplitude, principal direction angle and radial gradient respectively. The stress-like amplitude is the modulus of the complex acoustic potential used to reflect the overall strength of the acoustic field at the elbow joint rotation center. Step 6: Based on the virtual stiffness, damping and fractional order, the fractional order impedance law is applied to generate the output torque, and the target health care robot is driven to perform elbow-assisted stretching operation based on the output torque.

2. The health and wellness service robot system based on voice recognition according to claim 1, characterized in that, Double-ring vector cuffs, including: With the center of elbow joint rotation as the center C, and radii R1 and R2 respectively, construct an outer ring and an inner ring, where R1 > R2; The outer ring consists of six piezoelectric pads with a radius of R1 and polar angles of [missing information]. , , , , , It fits the bony ring-shaped area formed by the olecranon process and the medial and lateral epicondyles; The inner ring consists of six piezoelectric pads with a radius of R2. The polar angles of each pad are offset clockwise by 30° relative to the polar angles of the corresponding piezoelectric pads of the outer ring, fitting the surface of the joint capsule in the elbow fossa region.

3. The health and wellness service robot system based on voice recognition according to claim 2, characterized in that, The acoustic-electric signal of the double-loop vector cuff is preprocessed and the baseline acoustic-electric signal is subtracted to obtain the analytical acoustic-electric signal, including: Step 21, preprocessing, includes: analog-to-digital conversion, bandpass filtering, and moving average; Step 22: Perform Hilbert transform on the differential signal formed by subtracting the baseline acoustic signal from the preprocessed acoustic signal to obtain the analytical acoustic signal. , ; Step 23, the baseline acoustic-electric signal is acquired as follows: At each sampling moment within the preset time period, the target patient was controlled to keep their elbow at a certain position. In the flexed state, acquire the acoustic-electric signal of the double-loop vector cuff, and repeat step 21; The average value of the preprocessed acoustic-electric signals corresponding to all sampling times within a preset time period is taken as the baseline acoustic-electric signal.

4. The health and wellness service robot system based on voice recognition according to claim 3, characterized in that, The sampled values ​​for constructing boundary complex-valued functions based on analytical acoustic-electric signals include: Step 31, define the coordinates of the center C as: Where i represents the imaginary unit, Construct a complex plane coordinate system with the center C of the circle. Step 32: In the complex plane coordinate system, the coordinates of the 12 piezoelectric pads are numbered to form a piezoelectric pad coordinate cyclic sequence. , , This represents the coordinates of the k-th piezoelectric pad. This represents the polar angle of the k-th piezoelectric pad. Represents the natural base; Step 33: Calculate the difference vector for adjacent piezoelectric pad coordinates in the piezoelectric pad coordinate cyclic sequence. , ; Step 34: Determine if the sum of all difference vectors is 0, then... and Assigned to This forms a set of sampled values ​​for the boundary complex-valued function. Otherwise, return to step 32 to renumber.

5. A voice recognition-based elderly care service robot system according to claim 4, characterized in that, Substituting the boundary complex-valued function into the boundary integral model of the complex acoustic potential, the complex acoustic potential at the center of rotation of the target patient's elbow joint is obtained, including: Sampled values ​​based on boundary complex functions The discrete integral kernel is calculated; The complex acoustic potential is obtained by discretized Cauchy integral.

6. A health and wellness service robot system based on voice recognition according to claim 5, characterized in that, Extracting stress-like amplitude, principal direction angle, and radial gradient from complex acoustic potentials, including: The magnitude of the complex acoustic potential is obtained by calculating the stress-like amplitude. ; The argument of the complex acoustic potential is obtained to get the principal direction angle. ; The radial gradient is obtained as follows: Constructing a quadratic kernel function , ; Calculate the radial gradient , .

7. A health and wellness service robot system based on voice recognition according to claim 6, characterized in that, The virtual stiffness, damping, and fractional order are determined based on the stress amplitude, principal direction angle, and radial gradient, including: Calculate virtual stiffness , ; Calculate damping parameters , ; Calculate fractional order , ; in, , , and These represent the preset minimum virtual stiffness, preset minimum damping, preset lower limit of fractional order, and preset upper limit of fractional order, respectively. and These represent the stiffness gain coefficient and the damping gain coefficient, respectively. and This represents the maximum stress amplitude and maximum radial gradient of the target patient; The maximum stress amplitude and maximum radial gradient of the target patient were obtained as follows: The target rehabilitation robot drives the target patient's elbow from a constant angular velocity. Flex slowly extend to Bending, then returning Complete the flexion-extension cycle; The maximum stress amplitude and maximum radial gradient within the flexure-extension cycle are obtained by marking.

8. A health and wellness service robot system based on voice recognition according to claim 7, characterized in that, Based on virtual stiffness, damping, and fractional order, the fractional-order impedance law is applied to generate an output torque. This output torque then drives the target rehabilitation robot to perform elbow-assisted stretching operations, including: Determine the actual and expected angles of the target patient's elbow, and calculate the error angle; The error angle, as well as the preset truncation order, fractional order, and sampling period of the acoustic signal are used as input variables for the Grünwald–Letnikov algorithm to calculate the fractional derivative. The output torque of the target health and wellness robot is determined by combining the fractional derivative with virtual stiffness and damping, based on the fractional impedance law. The target health and wellness robot is driven by output torque to perform elbow-assisted stretching operations.

Images