Foot kicking action recognition method, vehicle tailgate opening method, device and readable medium

Abstract

This application discloses a kicking motion recognition method, a vehicle tailgate opening method, a device, and a readable medium. The main technical solution includes: acquiring radar channel impulse response data corresponding to a target motion; extracting features from the radar channel impulse response data to obtain multi-dimensional features of the target motion, wherein the multi-dimensional features are determined based on at least two of the motion amplitude features, trajectory features, and speed features; and identifying whether the target motion is a kicking motion based on the multi-dimensional features. This kicking motion recognition method can improve the detection rate of kicking motions, reduce the false alarm rate, and reduce the number of times the vehicle tailgate is mistakenly opened, thus improving the user experience.

Description

Technical Field

[0001] This application relates to the fields of radar signal processing and motion recognition technology, and in particular to methods for recognizing kicking motions, methods for opening vehicle tailgates, devices, and readable media. Background Technology

[0002] In non-contact human motion recognition applications (such as kicking to open a vehicle tailgate), existing solutions typically rely on capacitive or radar sensors for kicking motion recognition. Specifically, capacitive sensor-based solutions detect changes in capacitance caused by foot proximity to determine if a kick is occurring. However, this method is overly simplistic and suffers from deficiencies in accuracy, sensitivity, and anti-interference capabilities, resulting in a high false alarm rate for kicking motions. Radar-based solutions can identify kicks by analyzing the signal characteristics generated by the kicking motion in the radar echo, and compared to capacitive sensors, kicking motion recognition is more accurate. However, existing radar-based solutions rely on a single dimension or a few radar signal features, making it insufficient to distinguish between similar interfering actions (such as pedestrians passing by or limb swings), resulting in a low detection rate and a high false alarm rate for kicking motions. This further leads to poor performance in practical applications. Taking kicking to open a vehicle tailgate as an example, the high false alarm rate for kicking motions easily increases the number of times the tailgate is mistakenly opened. Therefore, how to comprehensively utilize the multi-dimensional information reflected by radar signals to achieve high detection rate and low false alarm rate recognition of kicking motions has become a pressing technical problem to be solved in this field. Summary of the Invention

[0003] This application provides a kicking action recognition method, a vehicle tailgate opening method, a device, and a readable medium to accurately recognize kicking actions, reduce false alarm rates, and further reduce the number of times the vehicle tailgate is accidentally opened, thereby improving user experience.

[0004] This application provides the following solution: According to the first aspect, a kicking action recognition method is provided, the method comprising: acquiring radar channel impulse response data corresponding to the target action; performing feature extraction on the radar channel impulse response data to obtain multi-dimensional features of the target action, wherein the multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features; and identifying whether the target action is a kicking action based on the multi-dimensional features of the target action.

[0005] According to a second aspect, a kicking action recognition device is provided, the device comprising: a data acquisition module configured to acquire radar channel impulse response data corresponding to a target action; an action feature extraction module configured to extract features from the radar channel impulse response data to obtain multi-dimensional features of the target action, the multi-dimensional features being determined based on at least two of action amplitude features, trajectory features, and velocity features; and an action classification and recognition module configured to identify whether the target action is a kicking action based on the multi-dimensional features of the target action.

[0006] According to a third aspect, a method for opening a vehicle tailgate is provided. The method includes: acquiring radar channel impulse response data corresponding to a target action triggered on a target vehicle; extracting features from the radar channel impulse response data to obtain multi-dimensional features of the target action, wherein the multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and speed features; and determining whether to control the tailgate of the target vehicle to perform an opening operation based on the multi-dimensional features of the target action.

[0007] According to a fourth aspect, a vehicle tailgate opening device is provided, the device comprising: a data acquisition module configured to acquire radar channel impulse response data corresponding to a target action triggered against a target vehicle; an action feature extraction module configured to extract features from the radar channel impulse response data to obtain multi-dimensional features of the target action, the multi-dimensional features being determined based on at least two of action amplitude features, trajectory features, and speed features; and a control module configured to determine whether to control the vehicle tailgate of the target vehicle to perform an opening operation based on the multi-dimensional features of the target action.

[0008] According to a fifth aspect, an electronic device is provided, comprising: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any one of the first or third aspects.

[0009] According to a sixth aspect, a computer-readable medium is provided having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the method described in any implementation of the first or third aspect above.

[0010] Based on the specific solution provided in this application, the following technical effects are disclosed: This application extracts multi-dimensional features from radar channel impulse response data that can more comprehensively and accurately characterize kicking actions, and uses these features from different physical dimensions to jointly identify kicking actions. This significantly improves the ability to distinguish kicking actions from various interference actions, increases the detection rate of kicking actions, reduces the false alarm rate, and further reduces the number of times the vehicle tailgate is accidentally opened, thus improving the user experience. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart of a kicking action recognition method provided in an embodiment of this application.

[0013] Figure 2 This is a time-domain schematic diagram of the radar channel impulse response data corresponding to the kicking action provided in the embodiments of this application.

[0014] Figure 3 This is a schematic diagram illustrating the differences between omnidirectional and directional antennas in an ultra-wideband radar antenna provided in an embodiment of this application.

[0015] Figure 4 A three-dimensional visualization diagram of the speed characteristics of a kicking motion provided in an embodiment of this application.

[0016] Figure 5 A two-dimensional visualization diagram of the trajectory features of a kicking action provided in an embodiment of this application.

[0017] Figure 6 This is a schematic block diagram of a convolutional neural network used for training a kicking action recognition model, as provided in an embodiment of this application.

[0018] Figure 7 A flowchart illustrating a method for opening a vehicle tailgate as provided in an embodiment of this application.

[0019] Figure 8 This is a schematic diagram of the entire vehicle tailgate opening process provided in the embodiments of this application.

[0020] Figure 9 A schematic block diagram of a kicking action recognition device provided in an embodiment of this application.

[0021] Figure 10 This is a schematic block diagram of a vehicle tailgate opening device provided in an embodiment of this application.

[0022] Figure 11 A schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0024] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0026] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0027] When identifying kicking actions based on radar sensors, existing radar-based solutions typically use only a single or a few dimensions of signal features to determine the kicking action, resulting in insufficient kicking action recognition capabilities. In particular, when the detection range of kicking actions is expanded and the number of interference sources increases, the accuracy of kicking action recognition will decrease and the false alarm rate will increase.

[0028] In view of this, this application provides a new approach, which will be described in detail below with reference to the accompanying drawings and embodiments.

[0029] Figure 1 A flowchart illustrating the kicking motion recognition method provided in this application embodiment. Figure 1 As shown, the kicking motion recognition method may include the following steps: Step 101: Obtain the radar channel impulse response data corresponding to the target's action.

[0030] Step 102: Extract features from the radar channel impulse response data to obtain multi-dimensional features of the target action. The multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features.

[0031] Step 103: Identify whether the target action is a kicking action based on the multi-dimensional features of the target action.

[0032] As can be seen from the above process, this application can significantly improve the ability to distinguish kicking actions from various interference actions by extracting multi-dimensional features that can more comprehensively and accurately characterize kicking actions from radar channel impulse response data, and by using these features from different physical dimensions for joint identification of kicking actions, thereby increasing the detection rate of kicking actions and reducing the false alarm rate.

[0033] The following describes in detail each step of the above process and the effects that can be further produced, with reference to the embodiments.

[0034] First, the above step 101, namely "acquiring radar channel impulse response data corresponding to the target action", will be described in detail with reference to the embodiments.

[0035] This step aims to acquire CIR (Channel Impulse Response) data corresponding to the electromagnetic wave reflection signal modulated by the target's action within the radar detection area, in order to identify whether the target's action is a kicking action.

[0036] In the embodiments of this application, the aforementioned target action refers to a suspected kicking human action occurring within the radar detection area. The aforementioned radar channel impulse response data generally refers to the time-domain impulse response sequence or discrete sampling sequence acquired by the radar system, characterizing the transmitted signal after it has passed through the propagation channel. The aforementioned radar channel impulse response data includes the signal's time delay, amplitude, and phase along all propagation paths. For example... Figure 2 As shown, Figure 2 This figure shows a time-domain plot of radar channel impulse response data corresponding to a kicking motion. The horizontal axis represents time, and the vertical axis represents energy amplitude. Figure 2 The part highlighted in red shows how the kick signal changes over time; the amplitude of the kick signal indicates the intensity of the kicking motion.

[0037] As an example, the radar system transmits pulse signals at a fixed period and receives electromagnetic wave reflection signals modulated by the target's action. The received electromagnetic wave reflection signals are then processed to obtain the radar channel impulse response data corresponding to the target's action. It should be noted that a complete suspected kicking action is typically recorded by multiple consecutive periods of CIR data, forming a CIR data matrix (range cell × time frame).

[0038] As one possible approach, the aforementioned target action could be triggered against a target vehicle.

[0039] As another possible approach, the aforementioned target action can be triggered by a target device, which could be an electrically controlled component on a vehicle, such as the tailgate. The aforementioned radar channel impulse response data can be acquired using an ultra-wideband radar deployed on a vehicle equipped with the target device. This ultra-wideband radar employs an omnidirectional antenna for signal transmission and reception. "Omnidirectional" in omnidirectional antenna refers to the sensor's effective range in all directions; the aforementioned omnidirectional antenna can be a single antenna or an array antenna. Figure 3 As shown, Figure 3 The difference between omnidirectional and directional antennas is illustrated. Omnidirectional antennas have a wider detection range, while directional antennas only function in a specific direction in space. Because omnidirectional antennas, while expanding the detection range of radar sensors, also allow the sensors to detect more targets, they are more prone to false alarms from kicking the tailgate. Therefore, most kick-opening radars on the market use directional antennas to detect an area in a fixed direction. It should be noted that the solution in this application has strong resistance to false alarms, thus allowing the use of an omnidirectional antenna to significantly increase the detection range and improve the user experience of kicking the tailgate. Furthermore, the installation position of the aforementioned omnidirectional antenna can be determined according to specific scenario requirements; for example, it can be installed in the middle of the rear bumper or on the left or right sides of the rear bumper.

[0040] The following describes in detail step 102, namely, "extracting features from radar channel impulse response data to obtain multi-dimensional features of target action, wherein the multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features," with reference to an embodiment.

[0041] This step aims to extract characteristic information of target actions in different physical dimensions from radar channel impulse response data in order to better distinguish them from jamming signals.

[0042] In the embodiments of this application, the aforementioned motion amplitude characteristics can characterize the energy distribution of the radar signal corresponding to the aforementioned target motion in a two-dimensional space of range and time. The aforementioned trajectory characteristics can characterize the positional change of the aforementioned target motion in a two-dimensional space of range and time. The aforementioned velocity characteristics can characterize the distribution of the velocity component of the aforementioned target motion moving along the radar radial direction in a two-dimensional space of range and time.

[0043] As a feasible approach, when extracting features from radar channel impulse response data to obtain multi-dimensional features of target actions, the following can be performed: When the aforementioned multi-dimensional features include motion amplitude features, an amplitude extraction operation is performed on the aforementioned radar channel impulse response data to obtain motion amplitude features; and / or when the aforementioned multi-dimensional features include velocity features, Doppler processing is performed on the aforementioned radar channel impulse response data to obtain velocity features; and / or when the aforementioned multi-dimensional features include trajectory features, a difference operation is performed on the aforementioned motion amplitude features along the time dimension, and the differenced result is used as the trajectory feature. Specifically: When the aforementioned multi-dimensional features include action amplitude features, the complex form of the radar channel impulse response data can be moduloed to obtain a signal amplitude value matrix as the action amplitude feature. The horizontal axis of the signal amplitude value matrix represents the continuous time points where the action occurs, and the vertical axis represents the radial distance unit between the target and the radar antenna. Each element value in the matrix represents the magnitude of the radar reflected signal received at the corresponding time point and the corresponding distance unit.

[0044] When the aforementioned multi-dimensional features include velocity features, the velocity feature of a kicking action is based on the human kicking posture. A normal human kick consists of one "go" and one "return" motion, which corresponds to approaching or moving away from the target device's location. For radar, approaching and moving away can be distinguished by "positive" and "negative" velocities. Therefore, the aforementioned velocity features include positive and / or negative velocity features. Positive velocity features represent the velocity when the target moves closer to the radar, while negative velocity features represent the velocity when the target moves away from the radar. Further, to obtain the velocity features, for each range cell of the signal amplitude value matrix corresponding to the radar channel impulse response data, the following steps can be performed: extract the multi-frame amplitude sequence corresponding to that range cell from the signal amplitude value matrix, perform a short-time Fourier transform on the multi-frame amplitude sequence to obtain the velocity spectra for that range cell, integrate the energy in the positive velocity interval of each velocity spectrum or take the maximum value as the positive velocity feature value for that range cell at the current moment, and then arrange the positive velocity feature values ​​of all range cells at all times in order of distance and time to obtain a positive velocity value matrix as the positive velocity feature. Similarly, the negative velocity eigenvalues ​​of all distance units at all times can be arranged in order of distance and time to obtain a negative velocity value matrix as the negative velocity feature. For example... Figure 4 As shown, Figure 4 A three-dimensional visualization of the velocity characteristics of the target's motion is shown. The three-dimensional visualization includes velocity dimension (0-140), distance dimension (0-10), and energy amplitude dimension (0-300). Figure 4 Specifically, it includes subgraph (a) and subgraph (b). Figure 4 Image (a) shows the positive velocity characteristics of the target's movement and the change in energy amplitude as the target approaches the radar. Figure 4Figure (b) shows the negative velocity characteristics of the target's movement and the change in energy amplitude as the target moves away from the radar.

[0045] When the aforementioned multi-dimensional features include trajectory features, the difference between the next frame and the previous frame can be calculated along the time dimension from the signal amplitude value matrix of the aforementioned action amplitude features to obtain the trajectory features. For example, Figure 5 As shown, Figure 5 This is a two-dimensional visualization of the trajectory characteristics of a real kicking action. The visualization uses time as the horizontal axis and distance as the vertical axis, showing the entire process of a kicking action, from the toes approaching the radar to the toes moving away from the radar.

[0046] Therefore, by extracting key features such as the amplitude, speed and trajectory of the target action from the radar channel impulse response data, more comprehensive and discriminative motion representation information can be provided for subsequent kick recognition, thereby reducing the false alarm rate of kick recognition.

[0047] The following describes step 103, namely "identifying whether the target action is a kicking action based on the multi-dimensional features of the target action", in detail with reference to the embodiments.

[0048] This step aims to use multi-dimensional features to make a comprehensive judgment on the target action in order to accurately determine whether the target action is a kicking action.

[0049] As a feasible approach, when identifying whether a target action is a kicking action based on its multi-dimensional features, a multi-channel feature image can first be constructed based on these features. This multi-channel feature image is then input into a pre-defined kicking action recognition model to determine whether the target action is indeed a kicking action. The kicking action recognition model can be a convolutional neural network that takes the multi-channel feature image corresponding to the action as input and an action type label as output. The action type label characterizes whether the target action corresponding to the input multi-channel feature image is a kicking action. Specifically, each feature in the multi-dimensional features can generate a single-channel image. These single-channel images are stacked to form a multi-channel feature image, which is then used as input to the kicking action recognition model. The final output is an action type label characterizing whether the target action is a kicking action. This approach transforms the radar feature processing problem into an image classification problem, fully utilizing the powerful spatial feature extraction and cross-channel information fusion capabilities of the kicking action recognition model to improve the kicking action detection rate while significantly reducing the false alarm rate.

[0050] Furthermore, as a feasible approach, when constructing a multi-channel feature image based on the multi-dimensional features of the target action, one can first fill the numerical values ​​of each dimensional feature into a preset two-dimensional coordinate grid to obtain a single-channel image. This two-dimensional coordinate grid can be composed of a time dimension and a distance dimension. Then, the single-channel images corresponding to each dimensional feature are stacked according to their channel dimensions to obtain a multi-channel feature image. Thus, by strictly aligning different dimensional features in both the spatial (distance) and temporal dimensions, the convolutional kernel of the kicking action recognition model can effectively perform cross-channel learning.

[0051] It's important to note that constructing multi-channel feature images is not limited to simple channel stacking. For example, each single-channel image can be normalized first to balance the influence of different physical dimensions; alternatively, an attention mechanism can be used to assign weights to features from different channels or spatial locations before combining them. The channel order of the multi-channel feature image can be arbitrary, as long as it matches the order used during model training. The image size can be adjusted according to network input requirements, such as scaling it to 64. 64. The input model can also be formed by inputting multiple branch networks separately, and then fusing them at the feature layer or decision layer.

[0052] Furthermore, the kicking motion recognition model can be obtained through the following steps: The first step is to obtain training data containing multiple training samples. These training samples include radar channel impulse response data and their corresponding action type labels, with action types including kicking actions or non-kicking actions.

[0053] The second step involves training a convolutional neural network using the aforementioned training data to obtain the kicking action recognition model. This training includes: The first sub-step involves extracting features from the radar channel impulse response data in the training samples to obtain multi-dimensional features. These multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features.

[0054] The second sub-step involves inputting the multi-dimensional features corresponding to the training samples into a convolutional neural network to obtain the action type labels predicted by the convolutional neural network. For example... Figure 6 As shown, Figure 6 A schematic block diagram of the above-described convolutional neural network is shown, which consists of an input layer, multiple convolutional-pooling layers, flattening layers, fully connected layers, and activation functions. The convolutional-pooling layers may include convolutional layers, rectified linear units, and pooling layers. As an example, the activation function can be the softmax function.

[0055] The third sub-step involves updating the parameters of the convolutional neural network by utilizing the difference between the action type labels predicted by the convolutional neural network and the action type labels in the corresponding training samples.

[0056] In this way, by using training data containing real kicking actions and various interference action samples, and by training a convolutional neural network based on multi-dimensional features, the kicking action recognition model can learn the complex and non-linear mapping relationship between these multi-dimensional features and kicking action categories from the data, thereby obtaining a classification model with high generalization ability and recognition accuracy specifically for this kicking action recognition scenario.

[0057] Furthermore, after identifying whether the target action is a kicking action based on its multi-dimensional characteristics, the system can also control the tailgate of the target vehicle to open in response to the target action being a kicking action. Specifically, when the target action is a kicking action, a tailgate opening control message can be sent directly to the tailgate drive motor and locking solenoid valve via the CAN (Controller Area Network) bus. Alternatively, a flag bit can be changed, and the underlying control loop program can detect the change in the flag bit and execute a preset tailgate opening program.

[0058] Continue to refer to Figure 7 , Figure 7 A flowchart illustrating a vehicle tailgate opening method provided in an embodiment of this application. Figure 7 As shown, the method for opening the tailgate of this vehicle may include the following steps: Step 701: Obtain radar channel impulse response data corresponding to the target action triggered by the target vehicle.

[0059] Step 702: Extract features from the radar channel impulse response data to obtain multi-dimensional features of the target action. The multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features.

[0060] Step 703: Based on the multi-dimensional characteristics of the target action, determine whether to control the tailgate of the target vehicle to perform an opening operation.

[0061] As can be seen from the above process, this application obtains radar channel impulse response data corresponding to the target action triggered by the target vehicle, and performs feature extraction on the radar channel impulse response data to obtain multi-dimensional features of the target action. The multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and speed features. Based on the multi-dimensional features of the target action, it is determined whether to control the tailgate of the target vehicle to perform an opening operation. Thus, it is possible to comprehensively judge whether to open the tailgate of the target vehicle based on the multi-dimensional features of the target action, thereby reducing the number of times the tailgate of the target vehicle is accidentally opened and improving the user experience.

[0062] The following describes in detail each step of the above process and the effects that can be further produced, with reference to the embodiments.

[0063] First, the above step 701, namely "acquiring radar channel impulse response data corresponding to the target action triggered by the target vehicle", will be described in detail with reference to the embodiments.

[0064] This step corresponds to step 101 above, and the radar channel impulse response data corresponding to the target action can be obtained from the vehicle-mounted ultra-wideband radar of the target vehicle.

[0065] The following describes in detail step 702, namely, "to extract features from the above radar channel impulse response data to obtain multi-dimensional features of the above target action, wherein the above multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features," with reference to the embodiments.

[0066] This step corresponds to step 102 mentioned above. For the specific implementation process, please refer to step 102 mentioned above. It will not be repeated here.

[0067] The following describes in detail step 703, namely, "determining whether to control the tailgate of the target vehicle to perform an opening operation based on the multi-dimensional characteristics of the target action", with reference to the embodiments.

[0068] This step is used to identify kicking motions and make a decision to open the tailgate.

[0069] One possible approach is to control the tailgate of the target vehicle to open in response to the identification of the target action as a kicking action based on the aforementioned multi-dimensional features. Specifically, referring to step 103 above, it is first determined whether the target action is a kicking action. If it is a kicking action, a tailgate opening command is generated and issued to control the tailgate of the target vehicle to open. It should be noted that if the identification result is not a kicking action, no control operation is performed, or only logging is performed.

[0070] like Figure 8As shown, Figure 8 The diagram illustrates the entire vehicle tailgate opening process described above. First, radar channel impulse response data corresponding to the target action is acquired. Then, preferably, multi-dimensional features such as the target action amplitude, positive velocity, negative velocity, and trajectory are extracted. These multi-dimensional features are then input into a kicking action recognition model, which outputs a kicking action judgment result. "Yes" indicates that the target action is a kicking action, and "No" indicates that the target action is not a kicking action. If the kicking action judgment result is "Yes", the vehicle tailgate is controlled to perform the opening operation.

[0071] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0072] Further reference Figure 9 , Figure 9 A schematic block diagram of the kicking motion recognition device is shown. This kicking motion recognition device is related to... Figure 1 Corresponding to the method embodiments shown, the kicking action recognition device 900 includes a first data acquisition module 901, a first action feature extraction module 902, and an action classification and recognition module 903. The main functions of each component module are as follows: The first data acquisition module 901 is configured to acquire radar channel impulse response data corresponding to the target action.

[0073] The first action feature extraction module 902 is configured to extract features from the aforementioned radar channel impulse response data to obtain multi-dimensional features of the aforementioned target action. The aforementioned multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features.

[0074] The action classification and recognition module 903 is configured to identify whether the target action is a kicking action based on the multi-dimensional features of the target action.

[0075] As one possible approach, the target action is triggered on the target vehicle. After identifying whether the target action is a kicking action based on the multi-dimensional characteristics of the target action, the kicking action recognition device 900 is further configured to control the tailgate of the target vehicle to perform an opening operation in response to the target action being a kicking action.

[0076] As one possible implementation, when the first action feature extraction module 902 extracts features from the radar channel impulse response data to obtain multi-dimensional features of the target action, it is specifically configured to perform amplitude extraction on the radar channel impulse response data to obtain action amplitude features when the multi-dimensional features include action amplitude features; and / or perform Doppler processing on the radar channel impulse response data to obtain velocity features when the multi-dimensional features include velocity features; and / or perform differential operation on the action amplitude features along the time dimension when the multi-dimensional features include trajectory features, and use the differential result as the trajectory feature.

[0077] As one possible implementation, the aforementioned velocity characteristics include positive velocity characteristics and / or negative velocity characteristics. Positive velocity characteristics are the velocity characteristics when the target moves closer to the radar, and negative velocity characteristics are the velocity characteristics when the target moves away from the radar.

[0078] As one possible approach, when the action classification and recognition module 903 identifies whether the target action is a kicking action based on the multi-dimensional features of the target action, it is specifically configured to construct a multi-channel feature image based on the multi-dimensional features of the target action; input the multi-channel feature image into a preset kicking action recognition model to obtain whether the target action is the kicking action.

[0079] As one possible approach, when the action classification and recognition module 903 constructs a multi-channel feature image based on the multi-dimensional features of the target action, it is specifically configured to fill the value of the dimension feature into a preset two-dimensional coordinate grid for each dimension feature to obtain a single-channel image. The two-dimensional coordinate grid is composed of time and distance dimensions. The single-channel images corresponding to each dimension feature are stacked according to the channel dimensions to obtain a multi-channel feature image.

[0080] As one possible approach, the aforementioned target action is triggered against the target device, and the aforementioned radar channel impulse response data is acquired by an ultra-wideband radar deployed on a vehicle on which the aforementioned target device is installed. The aforementioned ultra-wideband radar uses an omnidirectional antenna for signal transmission and reception.

[0081] As one possible implementation, the aforementioned kicking action recognition model is obtained through the following steps: acquiring training data containing multiple training samples, the training samples including radar channel impulse response data and their corresponding action type labels, the action type including kicking action or non-kicking action; training a convolutional neural network using the training data to obtain the aforementioned kicking action recognition model, wherein the training includes: extracting features from the radar channel impulse response data in the training samples to obtain multi-dimensional features, the multi-dimensional features being determined based on at least two of the action amplitude feature, trajectory feature, and velocity feature; inputting the multi-dimensional features corresponding to the training samples into the convolutional neural network to obtain the action type label predicted by the convolutional neural network; and updating the parameters of the convolutional neural network using the difference between the action type label predicted by the convolutional neural network and the action type label in the corresponding training samples.

[0082] Further reference Figure 10 , Figure 10 A schematic block diagram of a vehicle tailgate opening device is shown. This vehicle tailgate opening device is related to... Figure 7 Corresponding to the method embodiments shown, the vehicle tailgate opening device 1000 includes a second data acquisition module 1001, a second motion feature extraction module 1002, and a control module 1003. The main functions of each component module are as follows: The second data acquisition module 1001 is configured to acquire radar channel impulse response data corresponding to the target action.

[0083] The second action feature extraction module 1002 is configured to extract features from the aforementioned radar channel impulse response data to obtain multi-dimensional features of the aforementioned target action. The aforementioned multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features.

[0084] The control module 1003 is configured to determine whether to control the tailgate of the target vehicle to open based on the multi-dimensional characteristics of the target action.

[0085] As one possible implementation, when the control module 1003 determines whether to control the tailgate of the target vehicle to perform an opening operation based on the multi-dimensional characteristics of the target action, it is specifically configured to control the tailgate of the target vehicle to perform an opening operation in response to the recognition that the target action is a kicking action based on the multi-dimensional characteristics.

[0086] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0087] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0088] In addition, embodiments of this application also provide an electronic device, including: One or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any of the foregoing method embodiments.

[0089] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0090] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the methods described in the foregoing method embodiments.

[0091] in, Figure 11The architecture of an electronic device is illustrated, which may include a processor 1110, a video display adapter 1111, a disk drive 1112, an input / output interface 1113, a network interface 1114, and a memory 1120. The processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, and memory 1120 can communicate with each other via a communication bus 1130.

[0092] The processor 1110 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.

[0093] The memory 1120 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1120 can store the operating system 1121 for controlling the operation of the electronic device 1100, and the basic input / output system (BIOS) 1122 for controlling the low-level operations of the electronic device 1100. Additionally, it can store a web browser 1123, a data storage management system 1124, a kick gesture recognition device 900, and a vehicle tailgate opening device 1000, etc. The aforementioned kick gesture recognition device 900 and vehicle tailgate opening device 1000 can be application programs that specifically implement the aforementioned steps in this embodiment. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 1120 and executed by the processor 1110.

[0094] Input / output interface 1113 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0095] Network interface 1114 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0096] Bus 1130 includes a pathway for transmitting information between various components of the device, such as processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, and memory 1120.

[0097] It should be noted that although the above-described device only shows the processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, memory 1120, bus 1130, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the solution of this application, and does not necessarily include all the components shown in the figures.

[0098] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer program product. This computer program product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described above in various embodiments or some parts of embodiments of this application.

[0099] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for recognizing kicking motions, characterized in that, The method includes: Acquire radar channel impulse response data corresponding to the target's actions; Feature extraction is performed on the radar channel impulse response data to obtain multi-dimensional features of the target's action. The multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features. The target action is identified as a kicking action based on its multi-dimensional features.

2. The method according to claim 1, characterized in that, The target action is triggered against a target vehicle. After identifying whether the target action is a kicking action based on the multi-dimensional features of the target action, the method further includes: In response to the target action being a kicking action, the tailgate of the target vehicle is controlled to open.

3. The method according to claim 1, characterized in that, The step of extracting features from the radar channel impulse response data to obtain multi-dimensional features of the target action includes: When the multi-dimensional features include motion amplitude features, an amplitude extraction operation is performed on the radar channel impulse response data to obtain the motion amplitude features; and / or When the multi-dimensional features include velocity features, Doppler processing is performed on the radar channel impulse response data to obtain the velocity features; and / or When the multi-dimensional features include trajectory features, a difference operation is performed on the motion amplitude features along the time dimension, and the differenced result is used as the trajectory features.

4. The method according to claim 1, characterized in that, The velocity characteristics include positive velocity characteristics and / or negative velocity characteristics. Positive velocity characteristics are the velocity characteristics when the target moves closer to the radar, and negative velocity characteristics are the velocity characteristics when the target moves away from the radar.

5. The method according to claim 1, characterized in that, The step of identifying whether the target action is a kicking action based on the multi-dimensional features of the target action includes: Construct a multi-channel feature image based on the multi-dimensional features of the target action; The multi-channel feature image is input into a preset kicking action recognition model to determine whether the target action is the kicking action.

6. The method according to claim 5, characterized in that, The step of constructing a multi-channel feature image based on the multi-dimensional features of the target action includes: For each dimensional feature, the value of the dimensional feature is filled into a preset two-dimensional coordinate grid to obtain a single-channel image. The two-dimensional coordinate grid is composed of a time dimension and a distance dimension. The single-channel images corresponding to each dimension of features are stacked to obtain a multi-channel feature image.

7. The method according to any one of claims 1-6, characterized in that, The target action is triggered against the target device, and the radar channel impulse response data is acquired by an ultra-wideband radar deployed on a vehicle on which the target device is installed. The ultra-wideband radar uses an omnidirectional antenna for signal transmission and reception.

8. The method according to claim 5, characterized in that, The kicking motion recognition model is obtained through the following steps: Acquire training data containing multiple training samples, wherein the training samples include radar channel impulse response data and their corresponding action type labels, and the action type includes kicking action or non-kicking action; The training data is used to train a convolutional neural network to obtain the kicking action recognition model, wherein the training includes: Feature extraction is performed on the radar channel impulse response data in the training samples to obtain multi-dimensional features, which are determined based on at least two of the action amplitude features, trajectory features, and velocity features. The multi-dimensional features corresponding to the training samples are input into a convolutional neural network to obtain the action type label predicted by the convolutional neural network. The parameters of the convolutional neural network are updated by utilizing the difference between the action type labels predicted by the convolutional neural network and the action type labels in the corresponding training samples.

9. A method for opening a vehicle tailgate, characterized in that, The method includes: Acquire radar channel impulse response data corresponding to target actions triggered by the target vehicle; Feature extraction is performed on the radar channel impulse response data to obtain multi-dimensional features of the target's action. The multi-dimensional features are determined based on at least two of the action amplitude features, trajectory features, and velocity features. Based on the multi-dimensional characteristics of the target action, determine whether to control the tailgate of the target vehicle to open.

10. The method according to claim 9, characterized in that, The step of determining whether to control the tailgate of the target vehicle to open based on the multi-dimensional characteristics of the target action includes: In response to the recognition that the target action is a kicking action based on the multi-dimensional features, the tailgate of the target vehicle is controlled to open.

11. An electronic device, characterized in that, include: One or more processors; A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1 to 10.

12. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 10.

Images