Ultrasound sensor device for velocity estimation, and operating method thereof

The ultrasonic sensor device uses a composite chirp signal to measure Doppler velocity and positional relationships, addressing the challenge of weather-induced performance degradation in autonomous driving systems by accurately estimating object speed and position.

KR1020260161978APending Publication Date: 2026-11-10KOREA ADVANCED INST OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
KR1020250058009
Authority / Receiving Office
KR · KR
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-02
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing ultrasonic sensors struggle to accurately measure the speed of objects in adverse weather conditions due to their insensitivity to the Doppler effect, limiting their effectiveness in autonomous driving systems.

Method used

An ultrasonic sensor device that emits a composite chirp signal combining multiple chirp signals with varying frequencies, allowing for the estimation of both position and velocity of objects by analyzing echo signals for Doppler velocity and positional relationships using a processor or neural network.

Benefits of technology

The device can simultaneously estimate the position and velocity of moving objects regardless of weather conditions, enhancing accuracy and reliability in autonomous driving environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure PAT00015_ABST
    Figure PAT00015_ABST
Patent Text Reader

Abstract

An ultrasonic sensor device operated by at least one processor includes an ultrasonic transmitter that emits a composite chirp signal combined with a plurality of chirp signals whose frequency changes over a certain period of time, at least one ultrasonic receiver that receives echo signals reflected from an object, and a processor implemented to estimate velocity information of the object using chirp signal features included in the echo signals.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present disclosure relates to ultrasonic sensor technology. Background Technology

[0002] Autonomous driving technology is a technology that recognizes the driving environment, assesses risks, plans a driving route, and drives safely on its own. It is classified into stages based on the level of autonomy, and it is currently approaching Level 3, which enables low-speed urban driving during traffic congestion. In the future, for autonomous driving technology to reach a level of fully autonomous driving that requires no driver intervention, it is crucial for sensors to accurately collect information about surrounding objects even in various environments, including adverse weather conditions.

[0003] Cameras, LIDAR (Light Detection and Rranging) sensors, RADAR (Radio Detection and Rranging) sensors, and ultrasonic sensors can be used as sensors for autonomous driving. Cameras and LIDAR sensors have the disadvantage of significantly reduced performance in adverse weather conditions. RADAR and ultrasonic sensors are robust against environmental changes, as they experience almost no performance degradation due to weather. Although RADAR sensors have the disadvantage of relatively low resolution, ultrasonic sensors are utilized for object position measurement due to their advantages of being inexpensive and offering excellent spatial resolution unaffected by environmental conditions.

[0004] Meanwhile, during autonomous driving, the speed of surrounding objects must be accurately measured to precisely predict and track their location. Typically, speed is measured using images captured by a camera; however, since speed cannot be measured from a single frame, two frames are required for measurement, and speed measurement becomes difficult in adverse weather conditions. Therefore, a sensor device capable of measuring the speed of objects regardless of driving conditions, such as weather, is required. Prior art literature

[0005] (Patent Document 0001) KR 10-2191007 B(Patent Document 0002) KR 10-2261754 B(Patent Document 0003) KR 10-2023-0120506 A The problem to be solved

[0006] The present disclosure provides an ultrasonic sensor device that obtains an echo signal by radiating an ultrasonic chirp signal capable of measuring the Doppler effect, and estimates the velocity information of an object using the echo signal, and a method of operating the same.

[0007] The present disclosure provides an ultrasonic sensor device for simultaneously estimating the position and velocity of an object and a method of operating the same. means of solving the problem

[0008] An ultrasonic sensor device according to some embodiments comprises an ultrasonic transmitter that emits a composite chirp signal combined with a plurality of chirp signals whose frequency changes over a certain period of time, at least one ultrasonic receiver that receives echo signals reflected from an object, and a processor implemented to estimate velocity information of the object using chirp signal features included in the echo signals.

[0009] The above processor may be implemented to estimate the speed information using the Doppler velocity estimated from the echo signals.

[0010] The processor may be implemented to estimate the velocity information based on the positional relationship between the Doppler velocity estimated from the echo signals and the object.

[0011] The processor may be implemented to estimate the location information of the object from the chirp signal features included in the echo signals.

[0012] The processor may be implemented to estimate the location information of the object from the chirp signal features included in the echo signals.

[0013] The processor may be implemented to extract chirp signal features by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals.

[0014] The above processor can be implemented to extract the chirp signal features by analyzing the strength and phase of the echo chirp signal, whose frequency changes along the time axis.

[0015] The above processor can be implemented to estimate the location information and speed information of the object from the echo signals using a network.

[0016] A method of operating an ultrasonic sensor device according to some embodiments includes the step of radiating a composite chirp signal in which a plurality of chirp signals with varying frequencies over a certain period of time are combined, and the step of estimating velocity information of an object using chirp signal features included in echo signals reflected from an object.

[0017] The step of estimating the speed information above can estimate the speed information using the Doppler speed estimated from the echo signals.

[0018] The step of estimating the above velocity information can estimate the velocity information based on the positional relationship between the Doppler velocity estimated from the echo signals and the object.

[0019] The above operation method may further include the step of estimating the location information of the object from the chirp signal features included in the echo signals.

[0020] The above chirp signal characteristics can be extracted by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals.

[0021] The above chirp signal characteristics can be extracted by analyzing the intensity and phase of the echo chirp signal, in which the frequency changes on the time axis.

[0022] The above operation method may further include a step of analyzing the location information of the object using chirp signal features included in the echo signals.

[0023] The above speed information may include at least one of the Doppler velocity of the object and the relative velocity of the object.

[0024] According to some embodiments, an ultrasonic sensor device operated by at least one processor includes a network that estimates at least one of position information and velocity information of an object from chirp signal features included in echo signals, wherein the echo signals are signals obtained by radiating a composite chirp signal in which a plurality of chirp signals with a frequency changing for a certain period of time are combined.

[0025] The above network can extract chirp signal features by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals.

[0026] The above network can extract the chirp signal features by analyzing the strength and phase of the echo chirp signal, whose frequency changes along the time axis.

[0027] The above network can be trained to estimate the velocity information based on the positional relationship between the Doppler velocity estimated from the echo signals and the object. Effects of the invention

[0028] The ultrasonic sensor device according to the present disclosure can simultaneously estimate the position and velocity of a moving object without being significantly affected by surrounding environments such as bad weather, thereby increasing the accuracy of object estimation. Brief explanation of the drawing

[0029] Figure 1 is a diagram illustrating chirp signals and echo signals. FIG. 2 is a diagram illustrating a composite chirp signal according to one embodiment. FIG. 3 is a diagram illustrating the Doppler effect of a composite chirp signal according to one embodiment. FIG. 4 is a diagram illustrating a speed profile according to one embodiment. FIG. 5 is a block diagram of an ultrasonic sensor device according to one embodiment. FIG. 6 is a diagram conceptually illustrating a network structure according to one embodiment. FIG. 7 is a diagram illustrating an exemplary network structure according to one embodiment. FIG. 8 is a flowchart illustrating a network training method according to one embodiment. FIG. 9 is a flowchart illustrating the operation method of an ultrasonic sensor device according to one embodiment. Specific details for implementing the invention

[0030] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0031] Throughout the specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "…part," "…unit," and "module" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0032] FIG. 1 is a diagram illustrating chirp signals and echo signals, FIG. 2 is a diagram illustrating a composite chirp signal according to one embodiment, FIG. 3 is a diagram illustrating the Doppler effect of a composite chirp signal according to one embodiment, and FIG. 4 is a diagram illustrating a velocity profile according to one embodiment.

[0033] Referring to Fig. 1, a bat emits ultrasound and receives echo signals reflected back from objects. Since unique patterns corresponding to the shapes of the objects are superimposed on the echo signals, the bat can analyze the patterns of the echo signals to avoid obstacles and hunt prey. In this way, the process by which a bat distinguishes prey and identifies the location of obstacles from ultrasonic echoes reflected from objects is called echolocation. To mimic this echolocation of bats, ultrasonic chirp signals similar to the ultrasound emitted by bats can be used.

[0034] A conventional ultrasonic sensor device (10) can radiate a hyperbolic frequency-modulated chirp (HFM chirp) signal and acquire an echo signal. For example, an HFM chirp signal can be generated by reducing the frequency curvedly (e.g., from 60 kHz to 30 kHz) for a certain period of time (e.g., 8 ms). This HFM chirp signal has the advantage of excellent spatial resolution, as it can obtain a much clearer chirp due to the pulse compression effect through cross-correlation with a matched filter. Meanwhile, although the echo signal frequency of a moving object generally differs from the radiated signal due to the Doppler effect, the HFM chirp signal reflects the Doppler effect while maintaining a hyperbolic frequency sweep pattern, so the echo signal of the moving object can appear identical to the radiated signal. That is, the HFM chirp signal has the characteristic of not being sensitive to the Doppler effect.

[0035] As such, HFM chirp signals have the advantage of providing good spatial resolution and being insensitive to the Doppler effect, allowing for accurate position information to be obtained, and are therefore utilized in ultrasonic sensor devices. However, because HFM chirp signals are insensitive to the Doppler effect, they have the disadvantage of being difficult to obtain velocity information.

[0036] Referring to FIG. 2, the ultrasonic sensor device of the present disclosure obtains an echo signal by emitting an ultrasonic chirp signal (22) capable of measuring the Doppler effect in order to simultaneously estimate the position and velocity of an object.

[0037] The ultrasonic chirp signal (22) capable of measuring the Doppler effect may be a composite chirp signal formed by combining multiple chirp signals. For example, the composite chirp signal may be a signal formed by combining a first chirp signal in which the frequency decreases in a curve over a certain period of time and a second chirp signal in which the frequency increases in a curve over a certain period of time, but the number of chirp signals constituting the composite chirp signal may vary, and the frequency sweep pattern of the chirp signal may also be determined in various ways. For example, a composite chirp signal may be generated by combining selected chirp signal patterns among a pattern in which the frequency decreases over time and a pattern in which the frequency increases. In the description, an HFM chirp signal (HFM) in which the frequency decreases in a hyperbolic manner - )(20) and an HFM chirp signal (HFM) with a hyperbolic increase in frequency + A composite hyperbolic frequency-modulated (cHFM) chirp signal composed of (21) is described as an example of an ultrasonic chirp signal (22), but is not limited thereto. The composite chirp signal (22) can be generated as a frequency sweep pattern in which the frequency decreases in a curve (e.g., from 60 kHz to 30 kHz) for a certain period of time (e.g., 8 ms) and then increases in a curve (e.g., from 30 kHz to 60 kHz) for a certain period of time (e.g., 8 ms).

[0038] Referring to Figure 3, since the frequency change due to the Doppler effect does not affect the frequency sweep pattern of the chirp signal, the echo signal appears as if the complex chirp signal has shifted parallel to the time axis.

[0039] At this time, HFM - Chirp signals and HFM + Chirp signals move in different directions along the time axis. Through the width between two chirp signals included in these chirp signal features, the Doppler velocity v of a moving object d It can be measured. That is, HFM - Location information can be obtained through chirp signals, and HFM - HFM added to the chirp signal + Speed ​​information can be obtained through chirp signals.

[0040] In Fig. 3, T is the pulse duration, T' is T / D, and D is the Doppler coefficient. R is the echo detection point, and t + HFM + It is the point in time of maximum cross-correlation of the chirp signal, and t - HFM - It is the point of maximum cross-correlation of the chirp signal. △t + HFM + It is the time delay due to the Doppler effect of the chirp signal, and △t - HFM - It is a time delay caused by the Doppler effect of the chirp signal.

[0041] The width between two chirp signals due to the Doppler effect in the echo signal (t + - t - ) can be defined as in Equation 1, and △t + can be defined as in Equation 2, and the Doppler velocity V d is the width between two chirp signals (t) as in Equation 3. + - t -It can be calculated using ) and radiated signal characteristics (start frequency and end frequency, pulse period). In Equation 2, f1 is HFM + It is the starting frequency of the chirp signal, and f2 is HFM. + It is the end frequency of the chirp signal, and d is the Doppler coefficient. The Doppler coefficient d can be defined as in Equation 4.

[0042]

[0043]

[0044]

[0045]

[0046] Referring to Fig. 4, the Doppler velocity V d represents the projected velocity of the position vector between each ultrasonic sensor (ultrasonic receiver) and the object, and the relative velocity can be calculated from the Doppler velocity. V r is the magnitude of the object's relative velocity, α is the angle of the object's relative velocity, and V d θ is the Doppler velocity at the ultrasonic sensor, and θ is the angle between the ultrasonic sensor and the object.

[0047] The velocity information of an object is estimated based on the relationship between the Doppler velocity and the object's position, and a velocity profile can be utilized for estimating velocity information. A velocity profile can be expressed as the relationship between the Doppler velocity and the relative velocity as shown in Equation 5, but a velocity profile can be defined in various ways.

[0048]

[0049] The Doppler velocity V at each ultrasonic sensor via the complex chirp signal d If we estimate and the angle θ between the ultrasonic sensor and the object is known, the magnitude of the object's relative velocity V is obtained through the velocity profiles from two or more ultrasonic sensors. rThe relative velocity angle α between the object and the ultrasonic sensor can be calculated. The angle θ between the ultrasonic sensor and the object can be estimated from the Time of Flight (TOF) used for the object's position estimation. The magnitude of the object's relative velocity V r The relative velocity of the object from angle α, i.e., the velocity V in the X-axis direction x , velocity V in the Y-axis direction Y can be calculated.

[0050] The following describes in detail an ultrasonic sensor device that estimates the position and velocity of an object using a composite chirp signal capable of measuring the Doppler effect.

[0051] FIG. 5 is a block diagram of an ultrasonic sensor device according to one embodiment, FIG. 6 is a diagram conceptually explaining a network structure according to one embodiment, and FIG. 7 is a diagram exemplarily explaining a network structure according to one embodiment.

[0052] Referring to FIG. 5, the ultrasonic sensor device (100) can be implemented to acquire echo signals by radiating a composite chirp signal capable of measuring the Doppler effect, and to estimate the position information and velocity information (e.g., relative velocity) of an object using chirp signal features included in the echo signals. Chirp signal features can be extracted by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals. Chirp signal features can be extracted by analyzing the intensity and phase of the echo chirp signal, in which the frequency changes along the time axis. At this time, the position information of the object can be estimated based on chirp signal features such as signal intensity and Time of Flight (TOF) included in the echo signals. The position information of the object can be estimated based on the relationship between the Doppler velocity estimated from the echo signals and the object's position.

[0053] The ultrasonic sensor device (100) may include a network implemented to simultaneously estimate position information and velocity information of an object from chirp signal features included in echo signals. Here, the network may be composed of layers implemented to analyze features included in input data and output results, such as a neural network. The ultrasonic sensor device (100) may utilize a neural network mimicking the bat auditory system, which can be called a Bat-V Net in that it is a bat-mimicking neural network capable of measuring velocity. Here, the composite chirp signal capable of measuring the Doppler effect is, for example, an HFM chirp signal in which the frequency decreases hyperbola (HFM--- _ ) and an HFM chirp signal with hyperbolic frequency increase (HFM + It may be a composite hyperbolic frequency modulation (cHFM) chirp signal composed of ).

[0054] The ultrasonic sensor device (100) may include an ultrasonic transmitter (110) that emits an ultrasonic chirp signal, at least one ultrasonic receiver (120, 121, 122, 123) that receives echo signals reflected from an object, a processor (130) that estimates the position and velocity of an object from the echo signals, and a memory (140) that stores a program executed by the processor (130). In the description, it is assumed that echo signals are acquired using ultrasonic receivers placed at multiple locations, but the ultrasonic receivers do not necessarily need to be placed at multiple locations. In addition, the ultrasonic sensor device (100) may further include hardware necessary for the operation of the present invention. Meanwhile, the ultrasonic transmitter (110) and the multiple ultrasonic receivers (120, 121, 122, 123) may be connected to the processor (130) and the memory (150) via a wired or wireless network.

[0055] The ultrasonic transmitter (110) is a device that emits an ultrasonic signal capable of measuring the Doppler effect, i.e., a composite chirp signal, and may be a speaker that emits ultrasound by mimicking the mouth of a bat. The ultrasonic transmitter (110) may emit a composite chirp signal in which a first chirp signal in which the frequency decreases in a curve over a certain period of time and a second chirp signal in which the frequency increases in a curve over a certain period of time are combined. The composite chirp signal is an ultrasonic chirp signal capable of measuring the Doppler effect, for example, an HFM chirp signal (HFM) in which the frequency decreases in a hyperbolic manner. - ) and an HFM chirp signal with hyperbolic frequency increase (HFM + It may be composed of ), but the number of chirp signals constituting the composite chirp signal may vary, and the frequency sweep pattern of the chirp signal may also be determined in various ways. The ultrasonic chirp signal radiated from the ultrasonic transmitter (100) can be controlled by the processor (130). The processor (130) can perform imaging for various distances by controlling the length of the chirp.

[0056] Multiple ultrasonic receivers (120, 121, 122, 123) are input devices that receive echo signals reflected back from an object, and may be microphones that receive echo signals by mimicking a bat's ear.

[0057] A channel may be assigned to distinguish each ultrasonic receiver. The received ultrasonic echoes are transmitted to a processor (130). The number of receivers can be varied, but for the description, it is assumed that four channels of ultrasonic echoes are received by four receivers. The arrangement of the receivers can be designed in various ways.

[0058] The processor (130) is a device that controls the operation of the ultrasonic sensor device (100) and may be a processor of various types that processes instructions included in a program, for example, a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), GPU (Graphic Processing Unit), etc.

[0059] The memory (140) stores a program executed by the processor (130) and may be, for example, ROM (read only memory), RAM (random access memory), etc. The program includes instructions described to execute the operation of the present invention.

[0060] A processor (130) may be implemented to estimate the position and velocity information of an object from echo signals. A network implemented to estimate the position and velocity information of an object from echo signals may be utilized. For example, the processor (130) may be implemented to convert channel-specific echo signals into channel-specific echo images and to estimate the position and velocity information of an object from the echo images. The echo image may be a spectrogram in the time-frequency domain generated through a Fourier transform such as the Short-Time Fourier Transform (STFT). Here, the position information of an object may be expressed as the distance and angle (r, θ) of the object. Here, the velocity information of an object may include the relative velocity of the object. The relative velocity may be calculated from the Doppler velocity estimated from the width between two chirp signals. To this end, a velocity profile expressing the relationship between the Doppler velocity and the relative velocity may be utilized during the network training phase.

[0061] The network of the present disclosure may be trained to receive echo signals or echo images transformed from echo signals as input, and to estimate the position and velocity of an object from chirp signal features included in the echo signals. The position information of the object may be inferred from the analysis results obtained by comparing chirp signal features (e.g., signal strength and flight time) included in the echo signals. The velocity information of the object may be calculated using Doppler velocity. Doppler velocity may be inferred from the width between two chirp signals moving in different directions along the time axis.

[0062] The analysis results that a network can output by analyzing echo signals can be defined in various ways, for example, including the distance and angle (r, θ) of an object and the relative velocity of an object (velocity in the X-axis direction, velocity in the Y-axis direction), and may further include the Doppler velocity obtained during the estimation process and the confidence score for the existence of the object.

[0063] The network can be implemented with an artificial intelligence model that has learned at least one task and can be executed by a processor (130). Meanwhile, the artificial intelligence model may be completed in a separate device, and the ultrasonic sensor device (100) may be equipped with the learned artificial intelligence model. In this case, the ultrasonic sensor device (100) may update the artificial intelligence model by linking with a server (not shown) and transmit measurement data to the server. For convenience of explanation, the ultrasonic sensor device (100) is described as training an artificial intelligence model using learning data and estimating object information using the learned artificial intelligence model.

[0064] Referring to FIG. 6, the network (200) can be implemented to estimate output data (y) from input data (x).

[0065] The input data (x) is echo signals, or echo images converted from echo signals, and the echo signals are signals received by radiating a complex chirp signal.

[0066] The output data (y) may consist of an analysis result (y1) containing object location information and an analysis result (y2) containing object velocity information. The analysis result (y1) containing object location information includes the object's distance and angle (r i , θ i ), Doppler velocity (v) at each ultrasonic receiver i_1 , v i_2 ), confidence in object existence (c i It may include ). The analysis result (y2) containing the object's velocity information is the relative velocity (v) in the X-axis direction. x_i ), relative velocity in the Y-axis direction (v y_i It may include ). Here, i can be assigned to the identified object.

[0067] The network (200) may include an encoder implemented to estimate the position information and velocity information of an object by comparing and analyzing chirp signal features (signal strength, flight time, width between two chirp signals, etc.) included in the echo images. For example, the encoder may include a first encoder (E1) (210) implemented to estimate the position information and Doppler velocity of an object, and a second encoder (E2) (220) implemented to estimate the velocity information of an object based on the relationship between the Doppler velocity and the position of the object, but the encoders for estimating the position information and velocity information of an object do not necessarily need to be implemented separately. The first encoder (210) may be implemented to output an analysis result (y1) containing the object position information, and the second encoder (220) may be implemented to output an analysis result (y2) containing the object velocity information. The second encoder (E2) (220) can be implemented to calculate the relative velocity of an object from the Doppler velocity analyzed by the first encoder (E1) based on the velocity profile. Since the Doppler velocity is inferred from chirp signal features (e.g., width between two chirp signals) moving in different directions on the time axis, it can be inferred by the first encoder (210) along with object position information inferred on the time axis.

[0068] Referring to FIG. 7, the first encoder (210) and the second encoder (220) of the network (200) can be implemented in various ways. For example, the first encoder (210) can be trained to analyze the echo image of each channel using a network structure that mimics the auditory system of a bat and output an analysis result containing object location information. The second encoder (220) can be trained to output an analysis result containing object velocity information from the Doppler velocity analyzed by the first encoder (E1) based on a velocity profile.

[0069] The auditory system of bats can be composed of four layers. The four layers may be the cochlear nucleus (CN) layer, the superior olivary complex (SOC) layer, the nucleus of the lateral lemniscus (NLL) layer, and the inferior colliculus (IC) layer.

[0070] When a signal enters the bat's ear, the CN layer extracts non-linear spectral information of the sound and temporal information such as time of flight (TOF). The CN layer consists of the dorsal cochlear (DC) and the ventral cochlear (VC), where the DC analyzes the non-linear spectral information of the sound and the VC extracts temporal information.

[0071] The SOC layer estimates the location of an object through the differences in arrival times and strengths of echo signals received per channel. The SOC layer consists of MSO (Medial Superior Olives) and LSO (Lateral Superior Olives).

[0072] Data processed at all layers converges in the NLL layer and is reconstructed through the IC layer.

[0073] A first encoder (210) mimicking the auditory system of such a bat may construct a CN layer that extracts spectral and time information through a residual block, construct a SOC layer that estimates the position of an object by concatenating channel data output from the CN layer, and construct an NLL layer and an IC layer that integrate data through a convolution layer. At this time, unlike the existing bat-mimicking neural network that estimates the position of an object, the first encoder (210) may further include a layer that extracts the Doppler velocity from the width between two chirp signals moving in different directions along the time axis.

[0074] The result analyzed by the first encoder (210) can be output as an analysis result (y1) through a pooling layer and a fully connected layer. The analysis result (y1) may include the distance and angle of an object, Doppler velocity, and confidence in the existence of the object.

[0075] The second encoder (220) can be constructed as a layer that estimates the velocity information of an object from the Doppler velocity estimated in the first encoder (210). The relative velocity (y2) of the object is the relative velocity magnitude V of the object. r The relative velocity of the object can be expressed as an angle α, or the relative velocity in the X-axis direction calculated therefrom (v x_i ), relative velocity in the Y-axis direction (v y_i It can be output as ). The second encoder (220) uses the Doppler velocity V through the velocity profile of FIG. 4 and Equation 4 during the training process. d It can be trained to calculate the relative velocity of an object from.

[0076] A training device (not shown) can train the network (200) using training data. The training data may consist of input data (echo signals or echo images) and output data (correct answer data) of the neural network (200). The network (200) may be trained through a loss function in which loss functions for the object's location information, the object's velocity information (Doppler velocity and relative velocity), and the object's probability of existence are summed. Data inferred from the input data ( The loss function between ) and the ground truth data (p) can be defined as Equation 6, the loss function for the object's position can be defined as Equation 7, the loss function for Doppler velocity and relative velocity can be defined as Equation 8, and the loss function for the object's existence probability can be defined as Equation 9. Here, I i objindicates whether an object appears in grid cell i. λ dis and λ ang is 100, and λ v_d1, λ v_d2, λ v_x, λ v_y It can be 80 days.

[0077]

[0078]

[0079]

[0080]

[0081] In this way, the ultrasonic sensor device (100) can simultaneously estimate the position information and speed information of a moving object without being significantly affected by surrounding environments such as bad weather, and thereby increase the accuracy of the object's position.

[0082] Meanwhile, object tracking is the task of continuously detecting and tracking specific objects within a series of video frames. In various fields such as computer vision, robotics, autonomous vehicles, and security, real-time object tracking plays a crucial role in many applications. A representative estimation algorithm used for object tracking is the Kalman filter.

[0083] The Kalman filter optimally estimates state variables using currently observed and predicted information; in the prediction phase, it derives the predicted value from the previous correction value, while in the correction phase, it derives the current correction value using the Kalman gain. Although velocity values ​​are used in the process of predicting the previously corrected position, there are limitations in accurately predicting the position of an object whose velocity is constantly changing because these values ​​are also obtained from past position information. Furthermore, the Kalman gain is calculated through the error covariance of the predicted value; however, since the error covariance of the predicted value relies solely on position information, there are limitations when considering relative reliability.

[0084] To overcome these limitations, if the velocity information estimated in the present disclosure is input into a Kalman filter along with the position information, a more accurate position can be predicted in the prediction stage. Additionally, if the error covariance of the predicted value is calculated by considering both position and velocity when calculating the Kalman gain, a more accurate value can be obtained in the correction stage, thereby making the optimal position correction value more accurate. Therefore, by inputting the position and velocity information output from the neural network (200) of the present disclosure into a Kalman filter, the object estimation accuracy of the existing system can be further stabilized.

[0085] FIG. 8 is a flowchart illustrating a network training method according to one embodiment.

[0086] Referring to FIG. 8, the training device acquires training data consisting of input data regarding the echo signals of a complex chirp signal and correct data (output data) including position information and velocity information of an object (S110). The input data may be echo signals or echo images converted from the echo signals, which can be determined according to network characteristics. The velocity information may include Doppler velocity and relative velocity.

[0087] The training device trains the network (200) to infer the position information and velocity information of an object from chirp signal features included in the echo signals using training data (S120). The training method may vary. The training device may train the network (200) through a loss function such as Equations 6 through 9 so that the loss between the data inferred by the network (200) from the input data and the correct data is minimized. The network (200) may include an encoder that analyzes the chirp signal features included in the echo signals to estimate the position information and velocity information of an object. The encoder may include, for example, a first encoder (210) that estimates the position information of an object from chirp signal features such as flight time and estimates the Doppler velocity from chirp signal features moving in different directions on the time axis (e.g., width between two chirp signals), and a second encoder (220) that estimates velocity information (e.g., relative velocity) from the Doppler velocity based on the position relationship between the Doppler velocity and the object. The position relationship between the Doppler velocity and the object may be obtained based on a velocity profile.

[0088] FIG. 9 is a flowchart illustrating the operation method of an ultrasonic sensor device according to one embodiment.

[0089] Referring to FIG. 9, the ultrasonic sensor device (100) emits a composite chirp signal in which a plurality of chirp signals with varying frequencies over a certain period of time are combined (S210). The composite chirp signal is an ultrasonic chirp signal capable of measuring the Doppler effect, for example, an HFM chirp signal (HFM) in which the frequency decreases hyperbolically. - ) and an HFM chirp signal with hyperbolic frequency increase (HFM + It may be a composite hyperbolic frequency modulation (cHFM) chirp signal composed of ), but the number of chirp signals constituting the composite chirp signal may vary, and the chirp signal pattern may also be determined in various ways.

[0090] The ultrasonic sensor device (100) estimates the location information of an object using chirp signal features included in the echo signals (S220). The ultrasonic sensor device (100) can estimate the location information of an object by analyzing the echo images converted from the echo signals.

[0091] The ultrasonic sensor device (100) estimates the velocity information of an object using the Doppler velocity estimated from echo signals (S230). The ultrasonic sensor device (100) can estimate the Doppler velocity by analyzing chirp signal features (e.g., width between two chirp signals) moving in different directions along the time axis. The ultrasonic sensor device (100) can estimate the velocity information based on the positional relationship between the Doppler velocity and the object. The velocity information may include a relative velocity corresponding to the Doppler velocity.

[0092] The ultrasonic sensor device (100) outputs an analysis result including the location information of an object and the velocity information of an object estimated from echo signals (S240). The velocity information may include at least one of the Doppler velocity and the relative velocity of the object. The ultrasonic sensor device (100) may output an analysis result that further includes the reliability of the estimated object location.

[0093] Chirp signal features can be extracted by analyzing at least two of the changes in intensity, phase, and frequency of the echo chirp signals included in the echo signals. Chirp signal features can be extracted by analyzing the intensity and phase of the echo chirp signal, where the frequency changes over the time axis.

[0094] The ultrasonic sensor device (100) can estimate the position information and velocity information of an object by using a network trained to estimate the position information and velocity information of an object from chirp signal features included in the echo signals. The network (200) can be implemented to estimate the position information of an object based on the signal strength and flight time difference included in the echo signals, and to estimate the velocity information of an object using the Doppler velocity estimated from the chirp signal features and the position information of the object.

[0095] In this way, the ultrasonic sensor device (100) according to the present disclosure can simultaneously estimate the position and speed of a moving object without being significantly affected by surrounding environments such as bad weather, and thereby increase the accuracy of the moving object estimation.

[0096] The embodiments of the present invention described above are not implemented only through devices and methods, but may also be implemented through a program that realizes a function corresponding to the configuration of the embodiments of the present invention or a recording medium on which such program is recorded.

[0097] Although embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention as defined in the following claims also fall within the scope of the present invention.

Claims

Claim 1 An ultrasonic sensor device comprising: an ultrasonic transmitter that emits a composite chirp signal combining a plurality of chirp signals whose frequency changes over a certain period of time; at least one ultrasonic receiver that receives echo signals reflected from an object; and a processor implemented to estimate velocity information of the object using chirp signal features included in the echo signals. Claim 2 In claim 1, the ultrasonic sensor device is implemented such that the processor estimates the velocity information using the Doppler velocity estimated from the echo signals. Claim 3 An ultrasonic sensor device according to claim 1, wherein the processor is implemented to estimate the velocity information based on the positional relationship between the Doppler velocity estimated from the echo signals and the object. Claim 4 An ultrasonic sensor device according to claim 1, wherein the processor is implemented to estimate position information of the object from chirp signal features included in the echo signals. Claim 5 An ultrasonic sensor device according to claim 1, wherein the processor is implemented to estimate position information of the object from chirp signal features included in the echo signals. Claim 6 An ultrasonic sensor device according to claim 1, wherein the processor is implemented to extract chirp signal features by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals. Claim 7 An ultrasonic sensor device according to claim 1, wherein the processor is implemented to extract chirp signal features by analyzing the intensity and phase of the echo chirp signal whose frequency changes along the time axis. Claim 8 In claim 1, the ultrasonic sensor device is implemented such that the processor uses a network to estimate the position information and velocity information of the object from the echo signals. Claim 9 A method of operation for an ultrasonic sensor device comprising the steps of: radiating a composite chirp signal in which a plurality of chirp signals with varying frequencies over a certain period of time are combined; and estimating velocity information of an object using chirp signal features included in echo signals reflected from an object. Claim 10 In claim 9, the step of estimating the speed information is a method of operation in which the speed information is estimated using the Doppler speed estimated from the echo signals. Claim 11 In claim 9, the step of estimating the velocity information is a method of operation in which the velocity information is estimated based on the positional relationship between the Doppler velocity estimated from the echo signals and the object. Claim 12 A method of operation according to claim 9, further comprising the step of estimating location information of the object from chirp signal features included in the echo signals. Claim 13 In claim 9, the above chirp signal features are extracted by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals, in a method of operation. Claim 14 In claim 9, the above chirp signal feature is extracted by analyzing the intensity and phase of the echo chirp signal, in which the frequency changes on the time axis. Claim 15 A method of operation according to claim 9, further comprising the step of analyzing the location information of the object using chirp signal features included in the echo signals. Claim 16 In claim 9, the method of operation, wherein the velocity information includes at least one of the Doppler velocity of the object and the relative velocity of the object. Claim 17 An ultrasonic sensor device operated by at least one processor, comprising a network that estimates at least one of position information and velocity information of an object from chirp signal features included in echo signals, wherein the echo signals are signals obtained by radiating a composite chirp signal in which a plurality of chirp signals with a frequency changing for a certain period of time are combined. Claim 18 In claim 17, the ultrasonic sensor device, wherein the network extracts chirp signal features by analyzing at least two of the intensity change, phase change, and frequency change of the echo chirp signal included in the echo signals. Claim 19 In claim 17, the network is an ultrasonic sensor device that extracts chirp signal features by analyzing the intensity and phase of the echo chirp signal, whose frequency changes along the time axis. Claim 20 In paragraph 17, the ultrasonic sensor device, wherein the network is trained to estimate the velocity information based on the positional relationship between the Doppler velocity estimated from the echo signals and the object.