Running speed detection method and device based on radar echo analysis and treadmill

Abstract

The invention relates to a running speed detection method and device based on radar echo analysis and a running machine. The method comprises the steps that original echo signals are collected, fixed frequency interference is suppressed through time-frequency transformation processing and an adaptive filtering algorithm, and denoised time-frequency domain signals are obtained; performing short-time Fourier transform processing on the denoised time-frequency domain signal, and extracting a motion signal spectrum sequence; based on the motion signal spectrum sequence, a Kalman filtering algorithm is adopted to estimate the motion state at the current moment, and a motion state estimation vector containing position, speed and acceleration components is obtained; and generating real-time speed data based on the vector, wherein the real-time speed data is used for representing the movement speed of the runner. According to the method, real-time detection of the non-contact running speed is realized through a radar signal processing and state estimation technology.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of radar moving target detection, and particularly relates to a running speed detection method and device based on radar echo analysis and a treadmill. BACKGROUND

[0002] With the rapid development of motion state monitoring technology, running speed as a core parameter for evaluating sports performance and formulating training plans, its accurate and efficient detection has become a research hotspot. Under this background, non-contact motion detection technology based on radar echo analysis has gradually attracted attention. This technology can realize non-physical contact motion parameter measurement through electromagnetic wave reflection, can avoid the interference of wearable devices on motion actions, and is not affected by environmental factors such as light and shielding, and thus has unique advantages in the field of motion monitoring.

[0003] At present, the running speed detection technology based on radar echo analysis usually collects radar echo signals of a moving target, extracts motion features through signal processing to calculate the speed. However, the existing technology still has many limitations in practical application: for example, the original echo signal often contains fixed frequency interference, if only simple filtering processing is adopted, the interference cannot be effectively suppressed, resulting in insufficient signal-to-noise ratio of the denoised signal, which affects the accuracy of subsequent feature extraction; in the time-frequency analysis link, the window parameters of the short-time Fourier transform are mostly fixed settings, which cannot be adaptively adjusted according to the dynamically changing motion state of the runner, and the motion signal spectrum sequence is easily extracted inaccurately; when the motion state is estimated, there is a lack of dynamic calibration mechanism for the phase shift of the radar echo signal, and the accumulation of phase error will cause speed measurement deviation, and there is a lack of real-time optimization of signal processing parameters based on the state estimation result, which is difficult to balance the detection accuracy and real-time requirement. SUMMARY

[0004] Therefore, it is necessary to provide a running speed detection method, device and treadmill based on radar echo analysis which can solve the above problems.

[0005] In a first aspect, the application provides a running speed detection method based on radar echo analysis, comprising:

[0006] The original echo signal is collected, after time-frequency transformation processing, an adaptive filtering algorithm is adopted to suppress the fixed frequency interference, and the denoised time-frequency domain signal is obtained;

[0007] The denoised time-frequency domain signal is subjected to short-time Fourier transform processing, and the motion signal spectrum sequence is extracted;

[0008] Based on the motion signal spectrum sequence, a Kalman filtering algorithm is adopted to estimate the motion state at the current time, and a motion state estimation vector is obtained;

[0009] Real-time speed data is generated based on the motion state estimation vector, and the real-time speed data is used to represent the motion speed of the runner.

[0010] In one embodiment, performing short-time Fourier transform processing on the denoised time-frequency domain signal to extract the motion signal spectrum sequence includes:

[0011] The Hanning window is used as the time domain analysis window. The window length of the Hanning window is 256ms, and the overlap rate between adjacent analysis windows is 75%.

[0012] Perform short-time Fourier transform on each signal segment intercepted by the Hanning window and calculate the amplitude spectrum;

[0013] The sequence formed by arranging the amplitude spectra of all Hanning windows in time order is used as the motion signal spectrum sequence.

[0014] In one embodiment, a Kalman filter algorithm is used to estimate the motion state at the current moment to obtain a motion state estimation vector, including:

[0015] Define the state vector x of the Kalman filter k =[p k , v k , a k ] T , where p k is the position component of the runner, v k is the velocity component, a k is the acceleration component;

[0016] Based on the state vector x k , using Newton's laws of motion, the equation of state is established using the following formula:

[0017]

[0018] Where F is the state transfer matrix, Δt is the signal sampling time interval, w k is the process noise;

[0019] The phase change is calculated by taking the phase difference of the motion signal spectrum sequence at adjacent moments. And based on the principle of radar speed measurement, the observation equation is established using the following formula:

[0020] z k =Hx k +n k ,

[0021] where z k is the observed value, λ is the wavelength of the radar signal, n k is the observation noise;

[0022] predicting a prior state estimation and a prior estimation covariance of a current time based on a state equation;

[0023] calculating a Kalman gain based on an observation value of the current time and a motion signal spectrum sequence of an observation equation;

[0024] updating the prior state estimation by using the Kalman gain to obtain a state estimation vector of the current time;

[0025] taking the state estimation vector as a motion state estimation vector.

[0026] In one of the embodiments, after obtaining the motion state estimation vector, the method further comprises:

[0027] dynamically calibrating a phase offset of the radar echo signal based on a real-time mapping relationship between the motion state estimation vector and the radar echo signal;

[0028] updating an analysis window length of a subsequent short-time Fourier transform processing according to the calibrated phase offset.

[0029] In one of the embodiments, the dynamically calibrating the phase offset of the radar echo signal comprises calculating a phase offset estimation value of the radar echo signal by using the following formula:

[0030]

[0031] wherein, is a phase offset estimation value of the kth time, is a measured phase value extracted from the motion signal spectrum sequence of the kth time, v k is a velocity component of the motion state estimation vector of the kth time, Δt is a signal sampling time interval, and λ is a wavelength of the radar transmitting signal;

[0032] updating a time domain analysis window length of the short-time Fourier transform by using the following formula:

[0033]

[0034] wherein, W new is an updated window length, W0 is an initial window length, α ∈ [0.05, 0.2] is a preset calibration coefficient, and the calibration coefficient is determined through a radar system calibration experiment, is an absolute value of the phase offset estimation value.

[0035] In one of the embodiments, after calculating the phase offset estimation value of the radar echo signal the method further comprises:

[0036] calculating a correction amount Δv of the velocity component based on the phase offset estimation value and the wavelength λ of the radar transmitting signal by using the following formula:k :

[0037]

[0038] wherein, is the phase offset estimate, λ is the wavelength of the radar transmitted signal, Δt is the signal sampling time interval;

[0039] correction amount Δv based on the velocity component k The state vector correction amount Δx is generated using the following formula k :

[0040]

[0041] The motion state estimation vector is compensated in real time using the correction amount Δx k The calibrated motion state estimation vector is generated using the following formula

[0042]

[0043] The calibrated motion state estimation vector is used as the basis for generating real-time speed data.

[0044] In one embodiment, the real-time speed data is generated based on the motion state estimation vector, further comprising:

[0045] extracting the velocity component at the current time from the motion state estimation vector to generate a real-time speed value;

[0046] extracting the acceleration component at the current time from the motion state estimation vector to generate a real-time acceleration value;

[0047] obtaining real-time speed values at a plurality of consecutive times to calculate a speed change curve;

[0048] integrating the real-time speed value, the real-time acceleration value and the speed change curve to form real-time speed data.

[0049] In a second aspect, the application also provides a running speed detection device based on radar echo analysis, comprising:

[0050] a signal acquisition and processing module for acquiring original echo signals, after time-frequency transformation processing, using an adaptive filtering algorithm to suppress fixed frequency interference, to obtain denoised time-frequency domain signals;

[0051] a motion feature extraction module for performing short-time Fourier transform processing on the denoised time-frequency domain signals to extract motion signal spectrum sequences;

[0052] ​a motion state estimation module, configured to estimate a motion state at a current time based on the motion signal spectrum sequence by using a Kalman filtering algorithm, to obtain a motion state estimation vector;

[0053] a speed data generation module, configured to generate real-time speed data based on the motion state estimation vector, the real-time speed data being used to represent a motion speed of the runner.

[0054] In a third aspect, the present application further provides a treadmill device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned radar echo analysis-based running speed detection method when executing the computer program.

[0055] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the above-mentioned radar echo analysis-based running speed detection method when executed by a processor.

[0056] The above-mentioned radar echo analysis-based running speed detection method, device and treadmill can suppress fixed frequency interference by performing time-frequency transformation processing on the original echo signal and using an adaptive filtering algorithm, eliminate background noise generated by stationary objects (such as instruments and walls) in the environment, and improve the signal-to-noise ratio; the continuous motion signal spectrum sequence is extracted by performing short-time Fourier transformation processing on the denoised time-frequency domain signal, and the periodic dynamic characteristics of the runner's limbs are captured; the motion state is estimated based on the spectrum sequence by using the Kalman filtering algorithm, the state prediction and observation correction mechanism is used, the motion continuity is modeled by using the state equation during the motion acceleration / deceleration stage, and the measurement jitter caused by the sudden change of speed is smoothed by fusing the real-time phase change amount by using the observation equation; the real-time speed data is generated according to the motion state estimation vector, and the high-precision speed detection with strong anti-interference ability and fast dynamic response is realized through the closed-loop processing procedure. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiments or the related art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0058] Figure 1 A radar echo analysis-based running speed detection method flow chart of the present application;

[0059] Figure 2 A radar echo analysis-based running speed detection device structure diagram of the present application. DETAILED DESCRIPTION

[0060] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0061] In one embodiment, as shown in Figure 1 A radar echo analysis-based running speed detection method is provided. In a typical implementation environment, a terminal device (such as a smart treadmill integrated with a radar module, a wearable sports bracelet, or a smart phone) is provided with a millimeter wave radar sensor (such as a 60 GHz frequency band) inside. The radio frequency front end is responsible for directional transmission of continuous wave signals to the runner and collection of the original echo signals formed by reflection. The terminal device performs time-frequency transformation, adaptive filtering, and short-time Fourier transform algorithm through the embedded processor to output the motion signal spectrum sequence in real time. The Kalman filter algorithm module can be further run to dynamically construct the state equation and observation equation combined with the radar wavelength parameter, generate real-time speed / acceleration data through the motion state estimation vector, and display it on the terminal screen. In the application scenario, when the runner needs to monitor the motion performance in real time during indoor treadmill training or outdoor running, the radar sensor continuously captures the human motion micro-Doppler signal, and the processor accurately extracts the limb swing feature spectrum by suppressing the interference frequency of the environmental fixed object (such as the metal frame of the treadmill). In the sudden acceleration or speed change stage of the runner, the Kalman filter uses the state prediction based on Newton's law of motion and the observation correction based on phase difference to form a closed-loop feedback, which dynamically smooths the measurement noise caused by speed mutation. The generated real-time speed data is displayed locally on the terminal or transmitted to the cloud server through the network for motion data analysis and training plan optimization. This embodiment takes the method applied to the terminal as an example. It can be understood that the method can also be applied to the server, and can also be applied to a system including the terminal and the server, and is realized through the interaction of the terminal and the server. In this embodiment, the method includes the following steps:

[0062] S01, the original echo signal is collected, and after time-frequency transformation processing, an adaptive filtering algorithm is used to suppress fixed frequency interference to obtain a denoised time-frequency domain signal.

[0063] Wherein, the original echo signal can be captured by a radar sensor (such as a millimeter wave radar sensor), which contains Doppler components generated by runner motion reflection and fixed frequency interference caused by environmental fixed objects (such as walls or equipment). After collection, the original signal is processed by time-frequency transformation to convert the time-domain signal into time-frequency domain representation to reveal the dynamic distribution characteristics of the signal in the frequency and time dimensions. An adaptive filtering algorithm (such as LMS or RLS algorithm) is used to suppress interference of the transformed signal, and the filter parameters are adaptively adjusted based on the real-time spectral characteristics of the signal. By identifying and filtering out the fixed frequency interference components, the signal-to-noise ratio of the signal is improved. After processing, the denoised time-frequency domain signal is generated, which retains the runner motion characteristics and provides a high-fidelity input for subsequent short-time Fourier transform. Through the closed-loop parameter optimization mechanism, the limitations of traditional fixed filtering are effectively overcome, and high anti-interference capability is realized.

[0064] S02, the denoised time-frequency domain signal is processed by short-time Fourier transform to extract the motion signal spectrum sequence.

[0065] Wherein, the Hann window can be used as the time domain analysis window for segmenting the denoised time-frequency domain signal. The window length of the Hann window can be 256 ms and the overlap rate between adjacent analysis windows is set to 75%. The parameterized window design balances the time-frequency resolution and computational efficiency; short-time Fourier transform operation is performed on each signal segment cut by the Hann window to calculate the amplitude spectrum representing the local signal energy distribution; all window amplitude spectra arranged in time sequence are integrated into a continuous motion signal spectrum sequence, which constitutes a time sequence spectrum vector representing the periodic motion characteristics of the runner's limbs. Through the fixed overlap rate sliding window mechanism, the motion signal transient characteristics are retained while overcoming the frequency spectrum leakage problem caused by traditional fixed window length, providing high-resolution time-frequency feature input for subsequent state estimation.

[0066] S03, based on the motion signal spectrum sequence, the Kalman filter algorithm is used to estimate the motion state at the current time to obtain the motion state estimation vector.

[0067] Wherein, based on the motion signal spectrum sequence, a three-dimensional state vector containing position component, velocity component and acceleration component can be constructed and a state equation based on Newton's law of motion can be established. At the same time, the phase change is calculated by the difference between adjacent phases of the motion signal spectrum sequence to establish an observation equation, and the Kalman filter recursion is performed: the prior state estimation and covariance are predicted based on the state equation, the Kalman gain is calculated by combining the current observation value, and the state estimation vector that integrates motion continuity and real-time phase change is obtained by updating the prior estimation using the gain, which is used as the optimal estimation solution of the motion state. Through the closed-loop mechanism of kinematic modeling and observation noise suppression, the measurement jitter caused by sudden changes in velocity is effectively smoothed.

[0068] S04, generating real-time speed data based on the motion state estimation vector, the real-time speed data being used to represent the motion speed of the runner.

[0069] The motion state estimation vector (composed of position component, speed component and acceleration component, estimated by Kalman filtering algorithm) is taken as the input source, the speed component at the current time is extracted from the vector to generate the real-time speed value, and the acceleration component is extracted to generate the real-time acceleration value; a sequence of real-time speed values at continuous time points is obtained, and a speed change curve is generated by time series analysis algorithm (such as difference calculation or curve fitting), which represents the time-varying trend of the motion speed of the runner; the real-time speed value, the real-time acceleration value and the speed change curve are integrated to form structured real-time speed data, which is used to represent the motion speed of the runner in real time. Through multi-dimensional data fusion and dynamic curve generation mechanism, the abstract state vector is converted into data output that can directly and quantitatively represent the motion performance, providing high-precision input for treadmill display or cloud analysis.

[0070] The above-mentioned running speed detection method based on radar echo analysis collects the original echo signal and processes it by time-frequency transformation, uses an adaptive filtering algorithm to suppress fixed frequency interference, eliminates background noise generated by stationary objects in the environment, and improves the signal-to-noise ratio; the denoised time-frequency domain signal is processed by short-time Fourier transform to extract the motion signal spectrum sequence, capture the periodic dynamic characteristics of the runner's limbs, and overcome the signal sequence extraction deviation caused by the difficulty of fixed window parameters to adapt to dynamic motion state; based on the motion signal spectrum sequence, the Kalman filtering algorithm is used to estimate the motion state at the current time to obtain the motion state estimation vector containing position, speed and acceleration components, and the state equation is used to model the motion continuity and the observation equation is used to integrate the real-time phase change, which significantly improves the robustness of state estimation by smoothing the measurement jitter caused by speed mutation in the acceleration or deceleration stage; real-time speed data is generated based on the motion state estimation vector, and the outputs of each link are integrated through a closed-loop processing procedure to realize high-precision speed detection with strong anti-interference ability and fast dynamic response.

[0071] In one embodiment, the denoised time-frequency domain signal is processed by short-time Fourier transform to extract the motion signal spectrum sequence, including:

[0072] S11, using a Hanning window as a time domain analysis window, the window length of the Hanning window being 256 ms, and the overlap rate between adjacent analysis windows being 75%;

[0073] S12, performing short-time Fourier transform on the signal segment intercepted by each Hanning window to calculate the amplitude spectrum;

[0074] S13, arranging the sequence composed of the amplitude spectra of all Hanning windows in time sequence as the motion signal spectrum sequence.

[0075] Specifically, a Hanning window is used as the time domain analysis window, which is a weighted window function used to reduce spectral leakage, the window length is set to 256 ms, and the overlap rate between adjacent analysis windows is fixed at 75%, which balances the time resolution and frequency resolution through this parameterized design; the short-time Fourier transform operation is performed on the signal segment intercepted by each Hanning window, and the amplitude spectrum is calculated to represent the energy distribution characteristics of the signal in the local time period; the amplitude spectra of all Hanning windows are arranged in time sequence to form a continuous motion signal spectrum sequence, which is used as the time-frequency feature representation of the periodic motion of the runner's limbs for subsequent motion state estimation. Through the sliding window mechanism with fixed overlap rate, the motion transient characteristics are preserved while the spectral distortion is suppressed, and efficient motion signal feature extraction is realized.

[0076] In one embodiment, the Kalman filter algorithm is used to estimate the motion state at the current time to obtain a motion state estimation vector, including:

[0077] S21, define the state vector x of the Kalman filter k = [p k , v k , a k ] T , wherein p k is the position component of the runner, v k is the velocity component, and a k is the acceleration component;

[0078] S22, based on the state vector x k , the state equation is established using the following formula based on Newton's law of motion:

[0079]

[0080] where F is the state transition matrix, Δt is the signal sampling time interval, w k is the process noise;

[0081] S23, calculate the phase change amount by differentiating the phase difference of the adjacent time motion signal spectrum sequence, and establish the observation equation based on the radar speed measurement principle using the following formula:

[0082]

[0083] where z k is the observation value, λ is the wavelength of the radar transmitted signal, and n k is the observation noise;

[0084] S24, based on the state equation, predict the prior state estimation and prior estimation covariance at the current time;

[0085] S25, based on the observation equation and the current time observation value of the motion signal spectrum sequence, calculate the Kalman gain;

[0086] S26, update the prior state estimation using the Kalman gain to obtain the state estimation vector at the current time;

[0087] S27, take the state estimation vector as the motion state estimation vector.

[0088] Exemplarily, define the state vector of the Kalman filter as x k k k k T , wherein p k is the position component of the runner, representing the real-time distance information between the radar and the target, v k is the velocity component, representing the instantaneous motion rate of the runner, and a k is the acceleration component, representing the rate of change of the velocity; based on the state vector, the state equation x k = Fx k-1 + w k is established using Newton's law of motion, wherein the state transition matrix F is defined as Δt is the signal sampling time interval, reflecting the time step of continuous state update, w k is the process noise, simulating the uncertainty of the motion model; by calculating the phase change amount from the phase difference of the adjacent time motion signal spectrum sequence, and based on the radar speed measurement principle, the following formula is used to establish the observation equation z k = Hx k + n k , wherein the observation matrix z k is the observation value, derived from the phase change amount λ is the wavelength of the radar transmitted signal, and n k ​​​​To observe noise, the measurement error is represented; based on the state equation, the prior state estimation and the prior estimation covariance at the current time are predicted by extrapolating the current state through the state transition matrix F and the last time state vector; based on the observation equation and the current time observation value of the motion signal spectrum sequence, the Kalman gain is calculated, which dynamically balances the weight of the predicted value and the observation value; the prior state estimation is updated using the Kalman gain, and the state estimation vector at the current time is obtained by weighted fusion of prediction and observation data; the state estimation vector is output as the motion state estimation vector, which integrates the optimal estimation values of position, velocity and acceleration components, and is used for subsequent real-time speed data generation. Through Newton's law of motion modeling state continuity, radar speed measurement principle provides real-time observation feedback, and combined with the prediction-update mechanism of Kalman filter, effectively suppresses noise accumulation and smooths motion mutation, realizes high-precision motion state estimation.

[0089] In one of the embodiments, after obtaining the motion state estimation vector, it further comprises:

[0090] S31, based on the real-time mapping relationship between the motion state estimation vector and the radar echo signal, dynamically calibrating the phase offset of the radar echo signal;

[0091] S32, according to the calibrated phase offset, updating the analysis window length of the subsequent short-time Fourier transform processing.

[0092] Specifically, after obtaining the motion state estimation vector at the current time through the Kalman filter algorithm, since the phase of the radar echo signal is easily offset by environmental interference (such as multipath reflection, system noise), and this offset will accumulate and affect the accuracy of subsequent state estimation, therefore, dynamic calibration is needed based on the current motion state estimation vector; based on the real-time mapping relationship between the motion state estimation vector and the radar echo signal, the phase offset of the radar echo signal is dynamically calibrated; the phase offset is the phase deviation caused by environmental interference or system error in the signal acquisition process; at this time, the phase offset estimation value can be calculated and calibrated, according to the calibrated phase offset estimation value, the analysis window length of the subsequent short-time Fourier transform processing is updated, the time domain window size of the intercepted signal segment is adjusted, and the updated short-time Fourier transform window length will be applied to the next round of signal processing, so that the time-frequency analysis resolution is adaptive to the motion state of the runner (such as shortening the window to improve the time resolution when the speed changes rapidly, and lengthening the window to improve the frequency resolution when the speed is uniform). Through the formation of a closed-loop adaptive mechanism, the signal processing parameters are optimized in real time based on state estimation feedback, which can effectively suppress the accumulation of phase error and dynamically balance the time-frequency resolution to adapt to the change of the runner's motion state, and improve the accuracy and real-time performance of speed detection.

[0093] In one of the embodiments, S41, the phase offset of the radar echo signal collection is dynamically calibrated, including calculating the phase offset estimation value of the radar echo signal using the following formula:

[0094]

[0095] wherein, is the phase offset estimation value at the kth moment, is the measured phase value extracted from the motion signal spectrum sequence at the kth moment, v k is the speed component of the motion state estimation vector at the kth moment, Δt is the signal sampling time interval, and λ is the wavelength of the radar transmitted signal;

[0096] S42, the time domain analysis window length of the short-time Fourier transform is updated, which is realized by the following formula:

[0097]

[0098] wherein, W new is the updated window length, W0 is the initial window length, and α ∈ [0.05, 0.2] is a preset calibration coefficient, which is determined by radar system calibration experiments, is the absolute value of the phase offset estimation value.

[0099] Exemplarily, the phase offset estimation value of the radar echo signal can be calculated by the formula wherein is the measured phase value extracted from the motion signal spectrum sequence at the kth moment (representing the actual phase state of signal collection), v k is the speed component of the motion state estimation vector at the kth moment (representing the instantaneous motion rate of the runner), 2·v k is the double Doppler effect caused by the radar round trip path, Δt is the signal sampling time interval (defining the time difference between consecutive sampling points), and λ is the wavelength of the radar transmitted signal (a constant determined by the radar hardware parameters); based on the calibrated phase offset estimation value, the time domain analysis window length of the short-time Fourier transform is updated, wherein the analysis window length is used to divide the time domain window size of the signal segment, and the updating operation is realized by the formula: Exemplarily, W0 is the initial window length (256 ms), and α ∈ [0.05, 0.2] is a preset calibration coefficient (determined by radar system calibration experiments, used to control the response strength of window adjustment), ​The absolute value of the phase offset estimation value is used to quantify the deviation amplitude; by forming a closed-loop adaptive mechanism, the signal processing parameters are optimized in real time based on state estimation feedback, the time-frequency resolution is dynamically balanced to adapt to the changes of the runner's motion state, the phase error accumulation is suppressed, and the accuracy and real-time performance of the speed detection are improved.

[0100] In one embodiment, the phase offset estimation value of the radar echo signal is calculated Then, the method further comprises:

[0101] S51, based on the phase offset estimation value and the wavelength λ of the radar transmitted signal, the correction amount Δv of the speed component is calculated by the following formula k :

[0102]

[0103] wherein, is the phase offset estimation value, λ is the wavelength of the radar transmitted signal, and Δt is the signal sampling time interval;

[0104] S52, based on the correction amount Δv of the speed component k , the state vector correction amount Δx is generated using the following formula k :

[0105]

[0106] S53, the motion state estimation vector is compensated in real time using the correction amount Δx k , and the calibrated motion state estimation vector is generated using the following formula

[0107]

[0108] S54, the calibrated motion state estimation vector is used as the basis for generating real-time speed data.

[0109] Specifically, after calculating the phase offset estimation value of the radar echo signal , based on the phase offset estimation value and the wavelength λ of the radar transmitted signal, the correction amount Δv of the speed component is calculated by the formula k , characterizes the phase deviation introduced by environmental interference or system error during signal acquisition, λ is a constant determined by radar hardware parameters, reflecting the characteristics of electromagnetic waves, and Δt is the signal sampling time interval (representing the time difference between consecutive sampling points, used to quantify the time resolution); based on the correction amount Δv of the speed component k ​(quantifying the magnitude of the bias of the speed estimation), using the formula generating a state vector correction amount Δx k which is adjusted only for the speed component, maintaining the stability of the position and acceleration components; using the correction amount Δx k compensating the motion state estimation vector x k (including the position component p k , the speed component V k and the acceleration component a k ) in real time, through the formula generating a calibrated motion state estimation vector which integrates the phase calibration information, improving the accuracy of the state estimation; using the calibrated motion state estimation vector as the basis for generating real-time speed data, for subsequent extraction of the speed component and output of the real-time motion parameters of the runner. By feeding back the phase offset error to compensate the state vector in real time, the cumulative error is suppressed and the robustness of the Kalman filter estimation is enhanced, achieving higher precision and dynamic response capability in the detection of running speed.

[0110] In one of the embodiments, generating real-time speed data based on the motion state estimation vector further includes:

[0111] S61, extracting the speed component at the current time from the motion state estimation vector to generate a real-time speed value;

[0112] S62, extracting the acceleration component at the current time from the motion state estimation vector to generate a real-time acceleration value;

[0113] S63, obtaining real-time speed values at consecutive time points to calculate and generate a speed change curve;

[0114] S64, integrating the real-time speed value, the real-time acceleration value and the speed change curve to form real-time speed data.

[0115] Exemplarily, during implementation, the velocity component at the current moment is extracted from the motion state estimation vector (directly obtained by indexing the vector elements) to generate a real-time speed value (outputting the runner's current speed in meters per second); the acceleration component at the current moment is extracted from the same vector (by indexing the vector elements) to generate a real-time acceleration value (reflecting the rate of change of speed in meters per second2); a sequence of real-time speed values ​​for multiple consecutive moments (such as within 10 sampling periods) is obtained (by caching historical data), and time series analysis (such as differential operation or polynomial fitting algorithm) is performed based on the sequence to calculate and generate a speed change curve (the curve is a time-speed function that visualizes the dynamic trend of running speed); the real-time speed value, real-time acceleration value and speed change curve are integrated (encapsulated through data structures, such as JSON format or array) to form real-time speed data (the data is output as a structured output for a treadmill display or a cloud-based motion analysis system). Through multi-dimensional feature extraction and dynamic curve generation, the state vector is efficiently converted into quantifiable and traceable motion parameters, thereby realizing real-time monitoring and training optimization of running speed.

[0116] The above-mentioned running speed detection method based on radar echo analysis collects the original echo signal and performs time-frequency transformation processing, and adopts an adaptive filtering algorithm (such as LMS or RLS algorithm) to suppress the fixed frequency interference generated by fixed objects in the environment, effectively eliminating background noise to improve the signal-to-noise ratio; then, the denoised time-frequency domain signal is subjected to short-time Fourier transform processing, in which a Hanning window is used as the time domain analysis window (the window length is 256ms, and the overlap rate of adjacent analysis windows is 75%) to extract the motion signal spectrum sequence, capture the periodic dynamic characteristics of the runner's limbs, and overcome the spectrum leakage problem caused by fixed window parameters; based on the motion signal spectrum sequence, the Kalman filter algorithm is used to estimate the motion state at the current moment, and the state vector x is defined by k =[p k , v k , a k ] T And use Newton's laws of motion to establish the state equation x k =Fx k-1 +w k And calculate the phase change based on phase difference Establish the observation equation z k =Hx k +n k , predict the prior state estimate and update the Kalman gain to generate a motion state estimation vector containing position, velocity and acceleration components, thereby smoothing the measurement jitter caused by the sudden change in velocity during the acceleration / deceleration phase; in addition, the phase offset of the radar echo signal is dynamically calibrated based on the motion state estimation vector (using the formula calculating the offset estimation value), and updating the window length of the short-time Fourier transform according to the calibration result (by implementing the formula ) to calculate the velocity component correction amount Δv k compensating the state vector in real time (using the formula ), forming a closed-loop feedback mechanism; extracting the velocity component from the motion state estimation vector to generate a real-time velocity value, extracting the acceleration component to generate a real-time acceleration value, and integrating the velocity value sequence at consecutive time points to calculate the velocity change curve, forming structured real-time velocity data. The adaptive filtering algorithm effectively suppresses fixed frequency interference and improves the signal-to-noise ratio of the original signal; the parameterized short-time Fourier transform window combined with the dynamic calibration mechanism optimizes the analysis accuracy and avoids the deviation of the motion signal sequence caused by fixed parameters; the state equation modeling of the Kalman filter effectively suppresses the accumulation of phase errors and the jitter of velocity measurement; the closed-loop process integrates real-time state estimation and parameter adaptive adjustment, achieving high-precision (strong anti-interference ability), fast dynamic response, and real-time benefits of non-contact running speed detection, which is suitable for treadmill equipment and cloud-based motion analysis systems.

[0117] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, there is no strict order limitation for the execution of these steps, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiments of the present application also provide a radar echo analysis-based running speed detection device for implementing the radar echo analysis-based running speed detection method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more radar echo analysis-based running speed detection device embodiments provided below can refer to the limitations of the radar echo analysis-based running speed detection method described above, and will not be repeated here.

[0119] In one exemplary embodiment, as shown in Figure 2 a radar echo analysis-based running speed detection device is provided, comprising:

[0120] The signal acquisition and processing module 101 is configured to acquire a raw echo signal, perform time-frequency conversion processing on the raw echo signal, and use an adaptive filtering algorithm to suppress fixed frequency interference, so as to obtain a time-frequency domain signal after noise reduction.

[0121] The motion feature extraction module 102 is configured to perform short-time Fourier transform processing on the time-frequency domain signal after noise reduction, and extract a motion signal spectrum sequence.

[0122] The motion state estimation module 103 is configured to use a Kalman filtering algorithm to estimate a motion state at a current time based on the motion signal spectrum sequence, so as to obtain a motion state estimation vector.

[0123] The speed data generation module 104 is configured to generate real-time speed data based on the motion state estimation vector, and the real-time speed data is used to represent a motion speed of a runner.

[0124] In one embodiment, the motion feature extraction module 102 is further configured to:

[0125] A Hanning window is used as a time domain analysis window, the length of the Hanning window is 256 ms, and the overlap rate between adjacent analysis windows is 75%;

[0126] Short-time Fourier transform is performed on a signal segment cut from each Hanning window, and an amplitude spectrum is calculated.

[0127] A sequence formed by arranging the amplitude spectra of all Hanning windows in time sequence is used as the motion signal spectrum sequence.

[0128] In one embodiment, the motion state estimation module 103 is further configured to:

[0129] A state vector x of the Kalman filter is defined as k = [p k , v k , a k ] T , wherein p k is a position component of the runner, v k is a speed component, and a k is an acceleration component.

[0130] Based on the state vector x k , a state equation is established using the following formula based on Newton's law of motion:

[0131]

[0132] , wherein F is a state transition matrix, Δt is a signal sampling time interval, w k is process noise.

[0133] The phase change amount is calculated by differentiating the phase difference of adjacent motion signal spectrum sequences And based on the radar speed measurement principle, the following formula is used to establish the observation equation:

[0134]

[0135] Where z k is the observation value, λ is the wavelength of the radar transmitted signal, n k is the observation noise;

[0136] Based on the state equation, the prior state estimation and the prior estimation covariance at the current time are predicted;

[0137] Based on the observation equation and the observation value of the motion signal spectrum sequence at the current time, the Kalman gain is calculated;

[0138] The prior state estimation is updated using the Kalman gain to obtain the state estimation vector at the current time;

[0139] The state estimation vector is used as the motion state estimation vector.

[0140] In one of the embodiments, the motion feature extraction module 102 is further configured to:

[0141] Based on the real-time mapping relationship between the motion state estimation vector and the radar echo signal, the phase offset of the radar echo signal is dynamically calibrated;

[0142] According to the calibrated phase offset, the analysis window length of the subsequent short-time Fourier transform processing is updated.

[0143] In one of the embodiments, the motion feature extraction module 102 is further configured to calculate the phase offset estimation value of the radar echo signal using the following formula:

[0144]

[0145] Where, is the phase offset estimation value at the kth time, is the measured phase value extracted from the motion signal spectrum sequence at the kth time, v k is the speed component of the motion state estimation vector at the kth time, Δt is the signal sampling time interval, and λ is the wavelength of the radar transmitted signal;

[0146] The time domain analysis window length of the short-time Fourier transform is updated, which is realized by the following formula:

[0147]

[0148] Where, W new is the updated window length, W0 is the initial window length, and α ∈ [0.05, 0.2] is a preset calibration coefficient, which is determined by radar system calibration experiment, |Δφ| is the absolute value of the phase offset estimation value.

[0149] In one embodiment, the motion state estimation module 103 is further configured to:

[0150] Based on the phase offset estimation value and the wavelength λ of the radar transmitted signal, the correction amount Δv of the velocity component is calculated by the following formula k :

[0151]

[0152] wherein, is the phase offset estimation value, λ is the wavelength of the radar transmitted signal, and Δt is the signal sampling time interval;

[0153] Based on the correction amount Δv of the velocity component k , the state vector correction amount Δx is generated by using the following formula k :

[0154]

[0155] The motion state estimation vector is compensated in real time by using the correction amount Δx k , and the calibrated motion state estimation vector is generated by using the following formula

[0156]

[0157] The calibrated motion state estimation vector is used as the basis for generating real-time velocity data.

[0158] In one embodiment, the velocity data generation module 104 is further configured to:

[0159] extract the velocity component at the current time from the motion state estimation vector to generate a real-time velocity value;

[0160] extract the acceleration component at the current time from the motion state estimation vector to generate a real-time acceleration value;

[0161] obtain real-time velocity values at a plurality of continuous time points to calculate and generate a velocity change curve;

[0162] integrate the real-time velocity value, the real-time acceleration value, and the velocity change curve to form real-time velocity data.

[0163] In one embodiment, a treadmill device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a radar echo analysis-based running speed detection method as described above.

[0164] In one embodiment, a computer readable storage medium is provided, having stored thereon a computer program, which when executed by a processor implements the steps of any of the above method embodiments.

[0165] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts are described in the part of the method embodiments. The above described device embodiments are only schematic and the components illustrated as separate components can or can not be physically separate and the components illustrated as a unit can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement without creative labor.

[0166] The above described embodiments only express several implementation manners of the present application, the description is more specific and detailed, but it should not be understood as a limitation to the patent scope of the application. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. A running speed detection method based on radar echo analysis, characterized in that: The method comprises: The original echo signal is collected, processed by time-frequency transformation, and an adaptive filtering algorithm is used to suppress fixed frequency interference to obtain the denoised time-frequency domain signal; Performing short-time Fourier transform processing on the denoised time-frequency domain signal to extract a motion signal spectrum sequence; Based on the motion signal spectrum sequence, a Kalman filter algorithm is used to estimate the motion state at the current moment to obtain a motion state estimation vector; Real-time speed data is generated based on the motion state estimation vector, and the real-time speed data is used to represent the motion speed of the runner.

2. The method according to claim 1, characterized in that The step of performing short-time Fourier transform processing on the denoised time-frequency domain signal to extract a motion signal spectrum sequence includes: A Hanning window is used as the time domain analysis window, wherein the window length of the Hanning window is 256 ms, and the overlap rate between adjacent analysis windows is 75%; Perform short-time Fourier transform on each signal segment intercepted by the Hanning window and calculate the amplitude spectrum; A sequence formed by arranging the amplitude spectra of all Hanning windows in time order is used as the motion signal spectrum sequence.

3. The method according to claim 1, characterized in that The Kalman filter algorithm is used to estimate the motion state at the current moment to obtain the motion state estimation vector, including: Define the state vector x of the Kalman filter k =[p k , v k , a k ] T , where p k is the position component of the runner, v k is the velocity component, a k is the acceleration component; Based on the state vector x k , using Newton's laws of motion, the equation of state is established using the following formula: Where F is the state transfer matrix, Δt is the signal sampling time interval, w k is the process noise; The phase change is calculated by taking the phase difference of the motion signal spectrum sequence at adjacent moments. And based on the principle of radar speed measurement, the observation equation is established using the following formula: where z k is the observed value, λ is the wavelength of the radar signal, n k is the observation noise; Based on the state equation, predict the prior state estimate and the prior estimate covariance at the current moment; Calculating a Kalman gain based on the observation equation and the current observation value of the motion signal spectrum sequence; Using the Kalman gain to update the prior state estimate to obtain a state estimate vector at a current moment; The state estimation vector is used as the motion state estimation vector.

4. The method according to claim 1, wherein After obtaining the motion state estimation vector, the method further includes: Dynamically calibrating the phase offset of the radar echo signal based on the real-time mapping relationship between the motion state estimation vector and the radar echo signal; The analysis window length of the subsequent short-time Fourier transform processing is updated according to the calibrated phase offset.

5. The method according to claim 4, characterized in that: The dynamic calibration of the phase offset of the radar echo signal acquisition includes calculating the phase offset estimate of the radar echo signal using the following formula: in, is the estimated phase offset at the kth moment, is the measured phase value extracted from the motion signal spectrum sequence at the kth moment, v k is the velocity component of the motion state estimation vector at time k, Δt is the signal sampling time interval, and λ is the wavelength of the radar transmission signal; The updating of the time domain analysis window length of the short-time Fourier transform is achieved by the following formula: Among them, W new is the updated window length, W0 is the initial window length, α∈[0.05,0.2] is the preset calibration coefficient, which is determined by the radar system calibration experiment. is the absolute value of the phase offset estimate.

6. The method according to claim 1, characterized in that In calculating the phase offset estimate of the radar echo signal After that, it also includes: Based on the phase offset estimate With the wavelength λ of the radar signal, the correction value Δv of the velocity component is calculated by the following formula k : in, is the phase offset estimate, λ is the wavelength of the radar signal, and Δt is the signal sampling time interval; The correction amount Δv based on the velocity component k , use the following formula to generate the state vector correction Δx k : Using the correction amount Δx k The motion state estimation vector is compensated in real time, and the calibrated motion state estimation vector is generated using the following formula: The calibrated motion state estimation vector As the basis for generating the real-time speed data.

7. The method according to claim 1, characterized in that The generating of real-time speed data based on the motion state estimation vector further includes: Extracting the velocity component at the current moment from the motion state estimation vector to generate a real-time velocity value; Extracting the acceleration component at the current moment from the motion state estimation vector to generate a real-time acceleration value; Obtaining the real-time speed values ​​at a plurality of consecutive moments and calculating and generating a speed change curve; The real-time speed value, the real-time acceleration value and the speed change curve are integrated to form the real-time speed data.

8. A running speed detection device based on radar echo analysis, characterized in that: The device comprises: The signal acquisition and processing module is used to collect the original echo signal, and after time-frequency transformation, an adaptive filtering algorithm is used to suppress fixed frequency interference to obtain the denoised time-frequency domain signal; A motion feature extraction module is used to perform short-time Fourier transform processing on the denoised time-frequency domain signal to extract a motion signal spectrum sequence; A motion state estimation module is used to estimate the motion state at the current moment using a Kalman filter algorithm based on the motion signal spectrum sequence to obtain a motion state estimation vector; The speed data generating module is used to generate real-time speed data based on the motion state estimation vector, and the real-time speed data is used to represent the motion speed of the runner.

9. A treadmill device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images