Sensor position sensing methods, devices, garments, and products for smart clothing

Abstract

This invention provides a sensor position sensing method, device, garment, and product for smart clothing. The method integrates an IMU sensor in the cuff of the smart garment. The method includes: acquiring measurement data in real time; determining the state of acceleration data in each frame of the measurement data; using measurement data with a predetermined number of consecutive frames of acceleration data indicating motion as motion data; detecting whether the number of cuff folds has changed; if so, estimating a first height change value; if not, extracting first data from the motion data and inputting it into a first LSTM model to obtain the motion type; when the motion type is cuff pulling, dividing the motion data into multiple second data, and inputting each of the multiple second data into a second LSTM model to obtain multiple second height change values; adding the first height change value to the multiple second height change values ​​to obtain the current height change value of the IMU sensor. This technical solution improves the measurement accuracy of the sensor in small-scale movements.

Description

Technical Field

[0001] This invention relates to the field of smart clothing, and in particular to a sensor position sensing method, device, garment, and product for smart clothing. Background Technology

[0002] In recent years, smart wearable systems incorporating wearable sensors such as pressure sensors, tension sensors, fiber optic sensors, and inertial sensors have been extensively researched to achieve functions such as motion capture, health monitoring, and human-computer interaction. The advantages of these smart garments are that they are not obstructed and have no limitations on range of motion, allowing users to use them freely and comfortably for extended periods. However, to ensure the accuracy of sensor measurements in smart garments, the sensors must be placed in the correct positions. However, due to the inherent characteristics of smart garments, issues such as sensor misalignment and motion artifacts in the fabric itself remain challenges for smart garment technology.

[0003] Traditional IMU localization methods determine position by transforming the IMU's acceleration from a local coordinate system to a global coordinate system and performing double integration after removing gravitational acceleration. However, this method leads to an accumulation of errors over time, quickly rendering the results invalid. Therefore, additional constraints on the IMU's motion are usually required to correct these errors, such as through pedestrian gait analysis. Data-driven IMU localization methods, on the other hand, utilize deep learning techniques to track the IMU's position. These are typically used for large-scale pedestrian localization, with errors measured in meters, making them unsuitable for the small-scale movements of sensors in smart clothing.

[0004] In addition, existing IMU position calculation methods often require combining multiple IMUs or IMUs with other sensors to reduce data drift. In smart clothing, the position detection of a single IMU is usually fixed on the sole of the shoe to reduce position tracking errors by taking advantage of some characteristics of human walking motion. However, this solution is not suitable for the small-scale movement of sensors in smart clothing. Summary of the Invention

[0005] Embodiments of the present invention provide a sensor position sensing method, device, garment, and product for smart clothing, which enables accurate prediction of cuff operations and height changes, improves the measurement accuracy of sensors in small-scale movements, and further enhances the measurement accuracy and natural interactive experience of smart clothing.

[0006] To achieve the above objectives, on the one hand, a sensor position sensing method for smart clothing is provided, wherein an IMU sensor is integrated into the cuff of the smart clothing, and the method includes:

[0007] The measurement data from the IMU sensor is acquired in real time, including acceleration data, angular velocity data, and rotation orientation data.

[0008] The zero-velocity detection method is used to determine whether the acceleration data in each frame of the measurement data is acceleration data when the IMU sensor is in motion or acceleration data when the IMU sensor is stationary.

[0009] When there are consecutive predetermined frames of acceleration data in the measurement data that represent the acceleration data of the IMU sensor in motion, the consecutive predetermined frames of measurement data are taken as motion data.

[0010] The current number of cuff folds is calculated based on the motion data, and the current number of cuff folds is compared with the previously calculated number of cuff folds to see if there is a change. If there is a change, the motion type of the motion data is cuff folding, and the first height change value of the IMU sensor relative to the reference point is estimated based on the current number of cuff folds and the preset value of each cuff fold, wherein the reference point is the initial position of the IMU sensor. If there is no change, a first predetermined number of frames of data are taken forward from the end of the motion data as first data, the first data is preprocessed according to a predetermined first processing rule, and the preprocessed first data is input into a pre-trained first long short-term memory model to obtain the motion type of the motion data. The motion type of the motion data includes: cuff pulling and cuff falling.

[0011] When the motion type of the motion data is cuff pulling, the motion data is divided into multiple second data, wherein each second data includes motion data of a second predetermined number of frames. The multiple second data are preprocessed according to a predetermined second processing rule, and the preprocessed multiple second data are respectively input into a pre-trained second long short-term memory model to obtain multiple second height change values ​​of the IMU sensor relative to the reference point.

[0012] The first height change value is added to the plurality of second height change values ​​to obtain the current height change value of the IMU sensor relative to the reference point.

[0013] Preferably, according to the sensor position sensing method, determining whether the acceleration data in each frame of the measurement data is acceleration data when the IMU sensor is in motion or acceleration data when the IMU sensor is stationary, using the zero-velocity detection method, includes:

[0014] Calculate the acceleration variance of a predetermined number of windows. When the acceleration variance is greater than or equal to a predetermined threshold, the acceleration data of the current frame is in motion. When the acceleration variance is less than the predetermined threshold, the acceleration data of the current frame is in a stationary state. The window length and the unit of the acceleration data are preset.

[0015] Preferably, according to the sensor position sensing method, calculating the current number of cuff folds based on the motion data includes:

[0016] A three-dimensional coordinate system is constructed based on the initial frame data in the motion data, wherein the three-dimensional coordinate system includes x, y, and z axes, the x and y axes are two axes in the horizontal direction, and the z axis is an axis in the vertical direction; and the orientation of the IMU sensor is projected onto the xz plane of the three-dimensional coordinate system;

[0017] Calculate the rotation angle of each frame of motion data relative to the y-axis, and then sum the rotation angles to obtain the total rotation angle of the motion data, where:

[0018]

[0019] Δθ k θ represents the rotation angle of the k-th frame of the motion data. k V represents the total rotation angle. k V′ represents the rotation vector of the k-th frame of the motion data. k V represents the rotation vector projected onto the xz plane from the k-th frame of motion data; k-1 V′ represents the rotation vector of the (k-1)th frame of the motion data. k-1 This represents the rotation vector projected onto the xz plane from the (k-1)th frame of the motion data; where the (k-1)th frame is the frame preceding the kth frame in the motion data.

[0020] The number of cuff folds is calculated using the total rotation angle, where:

[0021]

[0022] T k This indicates the number of times the cuff is folded.

[0023] Preferably, according to the sensor position sensing method, preprocessing the first data or the second data includes:

[0024] When the coordinate system of the first data or the second data is the IMU sensor coordinate system:

[0025] When the unit of the acceleration data is m / s 2 At that time, the acceleration data is scaled according to a first predetermined ratio;

[0026] When the unit of the angular velocity data is rad / s, the angular velocity data is scaled according to a second predetermined ratio;

[0027] When the number of cuff folds is odd, the x-axis and z-axis data of the acceleration data and the angular velocity data are inverted.

[0028] When the IMU sensor is integrated into the cuff of the left hand, the y-axis data of the acceleration data is inverted, and the x-axis and z-axis data of the angular velocity data are inverted.

[0029] Preferably, according to the sensor position sensing method, the preprocessing of the first data further includes: when the number of frames of the motion data is less than the first predetermined number of frames, using data that is stationary at the end of the motion data to append to the motion data to complete the data.

[0030] Preferably, according to the sensor position sensing method, the preprocessing of the second data further includes: when the number of frames of the motion data is less than the second predetermined number of frames, supplementing the motion data with data in a stationary state evenly before and after the motion data, until the number of frames in the motion state is equal to the second predetermined number of frames.

[0031] Preferably, according to the sensor position sensing method, the loss function of the first long short-term memory model is the cross-entropy function, and the activation function is the ReLU function;

[0032] The loss function of the second long short-term memory model is the mean squared error loss function, and the activation function is the ReLU function.

[0033] On the other hand, the present invention provides a sensor position sensing device for smart clothing, comprising a memory and a processor, wherein the memory stores at least one program, the at least one program being executed by the processor to implement the sensor position sensing method as described above.

[0034] In another aspect, the present invention provides a smart garment, comprising: an IMU sensor integrated in the cuff; and, as described above, a sensor position sensing device for the smart garment.

[0035] In another aspect, the present invention provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, it implements the sensor position sensing method as described above.

[0036] The above technical solution has the following technical effects:

[0037] The technical solution of this invention integrates an IMU sensor into the cuff of smart clothing to acquire multi-axis measurement data in real time, including acceleration, angular velocity, and rotational orientation, and utilizes the zero-velocity detection method (AMVD: Acceleration-Moving Variance). The detector determines the motion state of each frame of acceleration data. When a predetermined number of consecutive frames of acceleration data are detected to be in motion, the data segment is extracted as motion data. The current number of cuff folds is calculated, and the type of motion data is determined by comparing the current number of cuff folds with the previously calculated number of cuff folds. If the number of folds changes, the type is determined to be cuff fold, and the first height change value of the IMU sensor relative to the reference point is estimated based on the pre-stored cuff height data, where the reference point is the initial position of the IMU sensor. Otherwise, the first predetermined number of frames of data are extracted from the end of the motion data as first data. The first data is pre-processed according to the predetermined first processing rule and then input into the trained first long short-term memory (LSTM) model to determine the cuff type. When the type is identified as cuff pulling, the motion data is divided into multiple second data, where each second data includes a second predetermined number of frames of motion data. The multiple second data are pre-processed according to the predetermined second processing rule and then input into the second long short-term memory (LSTM) model to obtain multiple second height change values ​​of the IMU sensor relative to the reference point. Finally, the first height change value and multiple second height change values ​​are added together to obtain the current height change value of the IMU sensor relative to the reference point. This enables accurate prediction of cuff height changes, improving the user experience and accuracy of smart clothing in natural interactions.

[0038] In a further embodiment, by preprocessing the acceleration and angular velocity data in the motion data, including scaling the data proportionally and adjusting the data direction according to the sensor placement, it is ensured that the operation of the cuff can be accurately reflected regardless of how the cuff is folded or which arm it is worn on. This allows the smart clothing to maintain good measurement accuracy without strictly limiting the sensor placement, making the smart clothing closer to ordinary clothing worn every day.

[0039] In a further embodiment, by supplementing static state data when data is insufficient, all data segments can meet the requirements of model input, thereby improving the model's prediction accuracy for cuff operation types (such as folding or pulling) and their height changes. Ultimately, this achieves highly accurate tracking of smart garment sleeve operations, enhancing the practicality and user experience of smart garments. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating a sensor position sensing method for smart clothing according to an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram illustrating the results of motion state detection using AMVD on each frame of acceleration data in the measurement data in a sensor position sensing method for smart clothing according to an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram illustrating the classification of cuff operations and the prediction of cuff pulling distance in a sensor position sensing method for smart clothing according to an embodiment of the present invention.

[0043] Figure 4 The following is a flowchart illustrating the change in cuff height as a motion type of cuff pulling in a sensor position sensing method for smart clothing according to an embodiment of the present invention.

[0044] Figure 5 This is a diagram of a sensor position sensing device for smart clothing according to an embodiment of the present invention. Detailed Implementation

[0045] To further illustrate the various embodiments, the present invention provides accompanying drawings. These drawings are part of the disclosure of the present invention, primarily used to illustrate the embodiments and to explain the operating principles of the embodiments in conjunction with the relevant descriptions in the specification. With reference to these drawings, those skilled in the art should be able to understand other possible implementations and the advantages of the present invention. Components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0046] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0047] Example 1:

[0048] In order to achieve accurate prediction of cuff operation and height changes, improve the measurement accuracy of sensors in small-scale movements, and thus enhance the measurement accuracy and natural interaction experience of smart clothing, this invention provides a sensor position sensing method for smart clothing. Figure 1 This is a flowchart illustrating a sensor position sensing method for smart clothing according to an embodiment of the present invention. Figure 1 As shown, an IMU sensor is integrated into the cuff of the smart garment, including:

[0049] Real-time acquisition of measurement data from the IMU sensor, including acceleration data, angular velocity data, and rotation orientation data;

[0050] The zero-velocity detection method is used to determine whether the acceleration data in each frame of the measurement data is acceleration data when the IMU sensor is in motion or acceleration data when the IMU sensor is stationary.

[0051] When there are consecutive predetermined frames of acceleration data in the measurement data that represent the acceleration data of the IMU sensor in motion, the consecutive predetermined frames of measurement data will be used as motion data.

[0052] The current number of cuff folds is calculated based on the motion data. The current number of cuff folds is compared to the previously calculated number of cuff folds to determine if there is a change. If there is a change, the motion type of the motion data is cuff folding. The first height change of the IMU sensor relative to a reference point is estimated based on the current number of cuff folds and a pre-set value for each cuff fold. The reference point is the initial position of the IMU sensor. If there is no change, a first predetermined number of frames of data are extracted from the end of the motion data as the first data. The first data is preprocessed according to a predetermined first processing rule, and the preprocessed first data is input into a pre-trained first long short-term memory model to obtain the motion type of the motion data. The motion types of the motion data include: cuff pulling and cuff falling.

[0053] When the motion type of the motion data is cuff pulling, the motion data is divided into multiple second data, wherein each second data includes a second predetermined number of motion data frames. The multiple second data are preprocessed according to a predetermined second processing rule, and the preprocessed multiple second data are respectively input into a pre-trained second long short-term memory model to obtain multiple second height change values ​​of the IMU sensor relative to the reference point.

[0054] The first altitude change value is added to multiple second altitude change values ​​to obtain the current altitude change value of the IMU sensor relative to the reference point.

[0055] Example 2:

[0056] Existing IMU position calculations typically require combining multiple IMUs or fusing with other sensors to reduce data drift. In applications using a single IMU, a common practice is to mount it to the sole of a shoe, leveraging the characteristics of human walking motion to reduce errors in tracking the body's position. However, this gait-based approach is not suitable for the position detection needs of sensors in smart clothing during small-scale movements.

[0057] Therefore, embodiments of the present invention provide a sensor position sensing method for smart clothing. This method uses supervised learning and trains an LSTM model to predict the height change of an IMU sensor relative to its initial position during motion. The method includes:

[0058] 1. Acquire measurement data from the IMU sensor in real time.

[0059] Preferably, the measurement data includes acceleration data, angular velocity data, and rotational orientation data, totaling 9 axes of data.

[0060] Preferably, the IMU sensor is attached to the cuff of the smart garment by sewing a small pocket into the cuff.

[0061] In one specific implementation, the IMU sensor acquires and transmits data at a sampling rate of 60Hz.

[0062] 2. Use the zero velocity detection (AMVD) method to determine whether the acceleration data in each frame of the measurement data is acceleration data when the IMU sensor is in motion or acceleration data when the IMU sensor is stationary.

[0063] Preferably, Figure 2 This is a schematic diagram illustrating the motion state detection results of each frame of acceleration data in the measurement data using AMVD in a sensor position sensing method for smart clothing according to an embodiment of the present invention. Figure 2 As shown, the acceleration variance of a predetermined number of windows is calculated. When the acceleration variance is greater than or equal to a predetermined threshold, the acceleration data of the current frame is in motion; when the acceleration variance is less than the predetermined threshold, the acceleration data of the current frame is in a stationary state. The window length and the unit of acceleration data are preset.

[0064] In one specific implementation, the window length of AMVD is set to 30 frames, and the acceleration unit is set to m / s². 2 The threshold is set to 0.08. Each frame of acceleration data in the measurement data is contained in 30 different windows. If any one of these windows is considered stationary by AMVD, then the acceleration data of that frame is considered stationary.

[0065] In one specific implementation, the measurement data containing m frames of data is used as X. m Where the data in frame t is x t AMVD is used to determine whether the IMU sensor is currently in motion.

[0066] 3. When there are consecutive predetermined frames of acceleration data in the measurement data that represent the acceleration data of the IMU sensor in motion, the consecutive predetermined frames of measurement data will be used as motion data.

[0067] Preferably, the number of consecutive predetermined frames is 30 frames.

[0068] 4. Calculate the current number of cuff folds based on the motion data, and compare the current number of cuff folds with the number of cuff folds obtained in the previous calculation to see if there is any change;

[0069] If a change occurs, the motion type of the motion data is cuff folding, and the first height change value of the IMU sensor relative to the reference point is estimated based on the current number of cuff folds and the preset value of each cuff fold. The reference point is the initial position of the IMU sensor.

[0070] If there is no change, the data of the first predetermined number of frames is taken from the end of the motion data as the first data, the first data is preprocessed according to the predetermined first processing rule, and the preprocessed first data is input into the pre-trained first long short-term memory model to obtain the motion type of the motion data; preferably, the first predetermined number of frames is 60 frames; preferably, the motion type of the motion data includes: cuff pulling and cuff falling.

[0071] Figure 3 This is a schematic diagram illustrating the classification of cuff operations and the prediction of cuff pulling distance in a sensor position sensing method for smart clothing according to an embodiment of the present invention. The classification of cuff operations corresponds to the motion type of the motion data; the cuff pulling distance corresponds to the height change value of the IMU sensor installed in the cuff relative to a selected reference point. Preferably, the reference point is the initial position of the IMU sensor. Specifically, as shown... Figure 3 As shown, the sleeve operations are categorized into: pulling the cuff to fold it over and allowing the cuff to fall naturally under gravity; the pulling distance of the cuff is the distance between the position of the cuff equipped with the IMU sensor after the corresponding operation and the initial position of the cuff after the garment is worn. In one specific implementation, this cuff pulling distance is used to represent the height change value of the IMU sensor relative to a selected reference point.

[0072] In one specific implementation, calculating the current number of cuff folds based on motion data includes: constructing a three-dimensional coordinate system based on the initial frame data in the motion data, and projecting the orientation of the IMU sensor onto the xz plane of the three-dimensional coordinate system; preferably, the three-dimensional coordinate system includes x, y, and z axes, where the x-axis and y-axis are two axes in the horizontal direction, and the z-axis is an axis in the vertical direction;

[0073] Calculate the rotation angle of each frame of motion data relative to the y-axis, then sum the calculated rotation angles to obtain the total rotation angle of the motion data. The number of cuff folds is then calculated using the total rotation angle. Preferably, the total rotation angle and the number of cuff folds are calculated using the following formulas:

[0074]

[0075] Δθ kθ represents the rotation angle of the k-th frame of motion data. k V represents the total rotation angle. k V′ represents the rotation vector of the k-th frame of motion data. k V represents the rotation vector projected onto the xz plane from the k-th frame of motion data; k-1 V′ represents the rotation vector of the (k-1)th frame of motion data. k-1 This represents the rotation vector projected onto the xz plane from the (k-1)th frame of motion data; where the (k-1)th frame is the frame preceding the kth frame in the motion data.

[0076] The number of cuff folds is calculated using the total rotation angle, where:

[0077]

[0078] T k Indicates the number of times the cuffs are folded.

[0079] Preferably, the number of cuff folds at the final k-th frame is expressed as the result of the rotation angle divided by π and rounded off.

[0080] In one specific implementation, the first processing rule includes: when the coordinate system of the first data is the IMU sensor coordinate system; and / or,

[0081] When the unit of acceleration data is m / s 2 When the acceleration data is scaled according to a first predetermined ratio, and when the unit of the angular velocity data is rad / s, the angular velocity data is scaled according to a second predetermined ratio, thereby performing preliminary standardization of the data; preferably, the first predetermined ratio is 1 / 30; preferably, the second predetermined ratio is 1 / 10.

[0082] When the IMU sensor is placed 180° reversed around the y-axis, that is, when the number of cuff folds is odd, the x-axis and z-axis data of acceleration and angular velocity data are inverted to compensate for the data difference before and after folding.

[0083] When the IMU sensor is integrated into the cuff of the left hand, the y-axis data of acceleration data is inverted, and the x-axis and z-axis data of angular velocity data are inverted to compensate for the data differences between the left and right hands.

[0084] When the number of frames in the motion data is less than the first predetermined number of frames, data that is in a stationary state after the end of the motion data is appended to the end of the motion data to make the length of all data remain at the first predetermined number of frames.

[0085] Preferably, the first long short-term memory model, used as a model for classifying motion types, employs a cross-entropy function as the loss function and a ReLU function as the activation function. The loss function measures the difference between the model's predicted values ​​and the true values, and optimizes the model parameters by minimizing this difference. The activation function introduces non-linear characteristics, enabling the neural network to learn and represent complex function mappings. The cross-entropy function, which measures the difference between two probability distributions, is widely used in classification problems. The ReLU function is a common activation function; it is computationally fast, alleviates the gradient vanishing problem, and helps improve the model's sparsity.

[0086] In one specific implementation, the input data of the first long short-term memory model includes acceleration data and angular velocity data of the motion data after being processed by a predetermined first processing rule in the last 60 frames. Only the last 60 frames of data are used because the main difference between the cuff pulling type and the cuff falling type of motion data lies in the data pattern at the end of the motion.

[0087] 5. When the motion type of the motion data is cuff pulling, the motion data is divided into multiple second data sets. Each second data set includes motion data for a second predetermined number of frames. The multiple second data sets are preprocessed according to a predetermined second processing rule. The preprocessed second data sets are then input into a pre-trained second long short-term memory model to obtain multiple second height change values ​​of the IMU sensor relative to the reference point. Preferably, the second predetermined number of frames is 60 frames.

[0088] In one specific implementation, the second processing rule includes: when the coordinate system of the second data is the IMU sensor coordinate system; and / or,

[0089] When the unit of acceleration data is m / s 2 When the acceleration data is scaled according to a first predetermined ratio, and when the unit of the angular velocity data is rad / s, the angular velocity data is scaled according to a second predetermined ratio, thereby performing preliminary standardization of the data; preferably, the first predetermined ratio is 1 / 30; preferably, the second predetermined ratio is 1 / 10.

[0090] When the IMU sensor is placed 180° reversed around the y-axis, that is, when the number of cuff folds is odd, the x-axis and z-axis data of acceleration and angular velocity data are inverted to compensate for the data difference before and after folding.

[0091] When the IMU sensor is integrated into the cuff of the left hand, the y-axis data of acceleration data is inverted, and the x-axis and z-axis data of angular velocity data are inverted to compensate for the data differences between the left and right hands.

[0092] When the number of frames in the motion data is less than the second predetermined number of frames, static data is evenly distributed before and after the motion data until the number of frames in the motion data equals the second predetermined number of frames. Preferably, when the motion data cannot be divided into multiple complete 60-frame windows, i.e., the window length is not a multiple of 60, static data is added before and after the motion data. For example, if a segment of motion data has 201 frames, and it is 39 frames short when divided into 4 windows, then 20 frames of static data are added before the start of the motion data, and 19 frames of static data are added at the end of the motion data, thereby ensuring that a segment of data can be divided into several complete 60-frame windows.

[0093] Preferably, the loss function of the second long short-term memory model is the mean squared error loss function, and the activation function is the ReLU function; the mean squared error loss function used is used to calculate the mean of the squared error, which is more suitable for regression problems, especially when the output is a continuous value.

[0094] In one specific implementation, Figure 4 This is a flowchart illustrating the change in cuff height as a motion type of cuff pulling in a sensor position sensing method for smart clothing, according to an embodiment of the present invention. Figure 4 As shown, the motion data segment is divided into multiple windows of 60 frames each. The input data includes the acceleration and angular velocity data of the motion data within the 60 frames after processing by a predetermined second processing rule. The output prediction result is the height change value of the IMU sensor relative to the reference point in 60 frames, in centimeters.

[0095] 6. Add the first height change value to multiple second height change values ​​to obtain the current height change value of the IMU sensor relative to the reference point.

[0096] In one specific implementation, for a continuous IMU data stream, the IMU data stream is divided into a static state and a moving state. The IMU data continuously in the moving state is regarded as motion data, and a predetermined algorithm is used to distinguish the motion data into three types: the user folds (e.g., rolls up) the cuff, the cuff falls naturally under the action of gravity, and the user pulls the cuff. If the motion data is detected as the motion type of the user pulling the cuff, then a deep learning method is used to predict the distance change of the IMU sensor relative to the initial position of the IMU sensor.

[0097] The method described in this embodiment of the invention uses deep learning to predict end-to-end IMU sensor height changes based on the patterns of sleeve manipulation by users when wearing smart clothing. This solves the problem that existing IMU positioning schemes cannot be used for small-scale movements of sensors on smart clothing, and achieves high-precision adaptive sensor position perception for natural interaction with smart clothing.

[0098] Example 3:

[0099] The present invention also provides a sensor position sensing device for smart clothing, such as... Figure 5 As shown, the device includes a processor 501, a memory 502, a bus 503, and a computer program stored in the memory 502 and executable on the processor 501. The processor 501 includes one or more processing cores. The memory 502 is connected to the processor 501 via the bus 503. The memory 502 is used to store program instructions. When the processor executes the computer program, it implements the steps in the above-described method embodiment of Embodiment 1 of the present invention.

[0100] Furthermore, as an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0101] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0102] Example 4:

[0103] The present invention also provides a smart garment, comprising: an IMU sensor integrated in the cuff; and a sensor position sensing device for smart garment as described above.

[0104] Example 5:

[0105] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0106] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. A sensor position sensing method for smart clothing, characterized in that, An IMU sensor is integrated into the cuff of the smart garment, including: The measurement data from the IMU sensor is acquired in real time, including acceleration data, angular velocity data, and rotation orientation data. The zero-velocity detection method is used to determine whether the acceleration data in each frame of the measurement data is acceleration data when the IMU sensor is in motion or acceleration data when the IMU sensor is stationary. When there are consecutive predetermined frames of acceleration data in the measurement data that represent the acceleration data of the IMU sensor in motion, the consecutive predetermined frames of measurement data are taken as motion data. The current number of cuff folds is calculated based on the motion data, and the current number of cuff folds is compared with the previously calculated number of cuff folds to see if there is a change. If there is a change, the motion type of the motion data is cuff folding, and the first height change value of the IMU sensor relative to the reference point is estimated based on the current number of cuff folds and the preset value of each cuff fold, wherein the reference point is the initial position of the IMU sensor. If there is no change, a first predetermined number of frames of data are taken forward from the end of the motion data as first data, the first data is preprocessed according to a predetermined first processing rule, and the preprocessed first data is input into a pre-trained first long short-term memory model to obtain the motion type of the motion data. The motion type of the motion data includes: cuff pulling and cuff falling. When the motion type of the motion data is cuff pulling, the motion data is divided into multiple second data, wherein each second data includes motion data of a second predetermined number of frames. The multiple second data are preprocessed according to a predetermined second processing rule, and the preprocessed multiple second data are respectively input into a pre-trained second long short-term memory model to obtain multiple second height change values ​​of the IMU sensor relative to the reference point. The first height change value is added to the plurality of second height change values ​​to obtain the current height change value of the IMU sensor relative to the reference point.

2. The sensor position sensing method according to claim 1, characterized in that, Determining whether each frame of acceleration data in the measurement data represents acceleration data from when the IMU sensor is in motion or when the IMU sensor is stationary using the zero-velocity detection method includes: Calculate the acceleration variance of a predetermined number of windows. When the acceleration variance is greater than or equal to a predetermined threshold, the acceleration data of the current frame is in motion. When the acceleration variance is less than the predetermined threshold, the acceleration data of the current frame is in a stationary state. The window length and the unit of the acceleration data are preset.

3. The sensor position sensing method according to claim 1, characterized in that, The current number of cuff folds is calculated based on the aforementioned motion data, including: A three-dimensional coordinate system is constructed based on the initial frame data in the motion data, wherein the three-dimensional coordinate system includes x, y, and z axes, the x and y axes are two axes in the horizontal direction, and the z axis is an axis in the vertical direction; and the orientation of the IMU sensor is projected onto the xz plane of the three-dimensional coordinate system; Calculate the rotation angle of each frame of motion data relative to the y-axis, and then sum the rotation angles to obtain the total rotation angle of the motion data, where: Δθ k θ represents the rotation angle of the k-th frame of the motion data. k V represents the total rotation angle. k V′ represents the rotation vector of the k-th frame of the motion data. k V represents the rotation vector projected onto the xz plane from the k-th frame of motion data; k-1 V′ represents the rotation vector of the (k-1)th frame of the motion data. k-1 This represents the rotation vector projected onto the xz plane from the (k-1)th frame of the motion data; where the (k-1)th frame is the frame preceding the kth frame in the motion data. The number of cuff folds is calculated using the total rotation angle, where: T k This indicates the number of times the cuff is folded.

4. The sensor position sensing method according to claim 1, characterized in that, Preprocessing the first data or the second data includes: When the coordinate system of the first data or the second data is the IMU sensor coordinate system: When the unit of the acceleration data is m / s 2 At that time, the acceleration data is scaled according to a first predetermined ratio; When the unit of the angular velocity data is rad / s, the angular velocity data is scaled according to a second predetermined ratio; When the number of cuff folds is odd, the x-axis and z-axis data of the acceleration data and the angular velocity data are inverted. When the IMU sensor is integrated into the cuff of the left hand, the y-axis data of the acceleration data is inverted, and the x-axis and z-axis data of the angular velocity data are inverted.

5. The sensor position sensing method according to claim 1, characterized in that, Preprocessing the first data further includes: when the number of frames of the motion data is less than the first predetermined number of frames, using data that is at the end of the motion data and is in a stationary state to append to the motion data to complete it.

6. The sensor position sensing method according to claim 1, characterized in that, The preprocessing of the second data further includes: when the number of frames of the motion data is less than the second predetermined number of frames, supplementing the motion data with data in a stationary state evenly before and after the motion data, until the number of frames in the motion state is equal to the second predetermined number of frames.

7. The sensor position sensing method according to claim 1, characterized in that, The loss function of the first long short-term memory model is the cross-entropy function, and the activation function is the ReLU function; The loss function of the second long short-term memory model is the mean squared error loss function, and the activation function is the ReLU function.

8. A sensor position sensing device for smart clothing, characterized in that, It includes a memory and a processor, the memory storing at least one program, the at least one program being executed by the processor to implement the sensor position sensing method as described in any one of claims 1 to 7.

9. A smart garment, characterized in that, include: An IMU sensor integrated into the cuff; And, as described in claim 8, a sensor position sensing device for smart clothing.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the sensor position sensing method as described in any one of claims 1 to 7.

Images