A method for filtering water mist clutter in bathroom fall detection radar

Abstract

The present invention discloses a method for filtering water mist clutter of a bathroom fall detection radar, comprising the steps of obtaining energy spectrum, three-dimensional coordinate information, speed, signal-to-noise ratio in three-dimensional target point cloud data and range Doppler data information, and forming multivariate features, using a sliding time window method to calculate the total deviation coefficient α between each frame and the current N-frame average value, and using the total deviation coefficient as an influencing factor to iteratively invert the original echo signal to obtain converged echo data. The present invention is not affected by light and water mist, protects personal privacy, and effectively suppresses static object clutter and false target interference caused by water mist sprayed from various shower heads.

Description

Technical Field

[0001] The invention belongs to the technical field of fall detection, and in particular relates to a method for filtering water mist clutter of a bathroom fall detection radar. Background Art

[0002] As China's population gradually ages, there may be negligence in the care of the elderly or unforeseen dangers, such as an elderly person falling abnormally in the bathroom. Therefore, many types of sensors related to indoor personnel detection and tracking have appeared on the market, including ultrasound, infrared, lidar, and optical cameras, etc. These sensors are affected to varying degrees by the light temperature of the external environment, resulting in false alarms. Millimeter-wave radar has all-weather characteristics and is more adaptable to the environment than other sensors. Especially when it comes to protecting personal privacy and life, millimeter-wave radar has an irreplaceable natural advantage. However, the static object clutter in the small space of the bathroom and the false target interference caused by the water mist sprayed from different models of shower heads seriously affect the normal operation of the radar in the bathroom environment. Summary of the invention

[0003] In view of the above problems, the present invention proposes a water mist clutter filtering method for a bathroom fall detection radar. Based on the iterative inversion of multivariate feature deviations, the water mist clutter of the bathroom fall detection radar is filtered out, and water mist and other stationary clutter are effectively suppressed during the operation of the millimeter-wave radar, so that the functions of human target tracking and fall detection in indoor environments can be accurately realized.

[0004] The water mist clutter filtering method of the bathroom fall detection radar disclosed in the present invention comprises the following steps:

[0005] The radar configuration initializes parameters, and the multi-transmitter and multi-receiver millimeter-wave radar transmits electromagnetic waves to collect raw echo data;

[0006] The original echo data is Fourier transformed to obtain the range Doppler data of each virtual array element;

[0007] Performing high-precision super-resolution angle measurement on the range Doppler data to obtain three-dimensional spatial data of dynamic targets in the monitoring scene;

[0008] Use CFAR constant false alarm detection to process the input signal and determine a threshold, then output 3D target point cloud data;

[0009] The energy spectrum, three-dimensional coordinate information, velocity, and signal-to-noise ratio in the three-dimensional target point cloud data and range Doppler data information are obtained to form multivariate features. The total deviation coefficient α between each frame and the current N-frame average value is calculated using a sliding time window method. The original echo signal is iteratively inverted using the total deviation coefficient as an influencing factor to obtain converged echo data.

[0010] Furthermore, the processed echo data is subjected to target tracking filtering using a Kalman filter algorithm to obtain a stable moving target track, and then the motion posture of the human target is identified based on the track information to determine whether the human target has fallen in the bathroom.

[0011] Furthermore, the weighted coefficient after the multivariate eigenvalues ​​are compounded is:

[0012]

[0013] Wherein, α is the total deviation coefficient after the multivariate eigenvalues ​​are compounded; ε0 is the weight factor of the range Doppler energy spectrum preset according to the statistical results of the measured data; α0 is the deviation coefficient of the range Doppler energy; ε1, ε2, ε3 are the preset weight factors of the three-dimensional coordinate information, speed, and signal-to-noise ratio; α1, α2, α3 are the deviation coefficients of the three-dimensional coordinate information, speed, and signal-to-noise ratio; n is 3.

[0014] Furthermore, the calculation formula of the deviation coefficient of multivariate features is expressed as follows:

[0015]

[0016]

[0017] Among them, σ0 is the range Doppler energy deviation; is the current N-frame sliding window average value of the range Doppler energy; σ1, σ2, σ3 are the deviations of the three-dimensional coordinate information, speed, and signal-to-noise ratio; A i is the current N-frame sliding window average of 3D coordinate information, speed, and signal-to-noise ratio;

[0018] The current N-frame sliding window average of multivariate features is expressed as follows:

[0019] N=[power max *δ]

[0020]

[0021]

[0022] Where power max is the maximum value of the range Doppler energy spectrum in the current frame, δ is the sliding window adjustment coefficient, N is the sliding window size, where A0(j) is the range Doppler energy spectrum of the jth frame; when i=1, A i (j) is the 3D coordinate value of the jth frame; when i=2, A i (j) is the velocity value of the jth frame; when i=3, A i (j) is the signal-to-noise ratio value of the jth frame.

[0023] Furthermore, the calculation formula for the deviation of the energy spectrum and other multivariate features is as follows:

[0024]

[0025]

[0026] Further, the iterative inversion of the original echo signal by using the total deviation coefficient as an influencing factor to obtain converged echo data includes:

[0027] The complex signal formula model e(t,z) of the transmission echo of the kth virtual array element in the millimeter-wave original radar multi-transmit and multi-receive array is expressed as follows:

[0028]

[0029] Where A is the amplitude of the transmitted signal; f is the frequency of the radar transmitted signal; t represents the fast time with each frequency modulation as the starting point; T c is the frequency modulation period; K is the frequency modulation slope; is the initial phase of the radar transmitting signal; z is the frequency modulation index of the transmitting antenna;

[0030] Among them, the sum of the original echo signals of all the transmitting and receiving virtual array elements is E(t,z), where j represents the reception of the jth virtual array element:

[0031]

[0032] The new echo complex signal model E(t,z) is constructed by the total deviation coefficient α after the multivariate eigenvalue compounding:

[0033]

[0034] Repeatedly construct a new echo complex signal model until α < α is satisfied T Or M>M T ,

[0035] Among them, α T is the preset training error threshold, M is the number of iterations, and M T The maximum number of iterations allowed.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] It uses millimeter-wave radar processing algorithms to detect human body falling postures, and has the performance of not being affected by light and water mist. Compared with traditional cameras, it has the advantage of protecting personal privacy.

[0038] The diversified feature extraction and echo iterative inversion method is used to effectively suppress static object clutter and false target interference caused by water mist sprayed from various shower heads. Experiments have verified that the fall radar has a good detection effect.

[0039] It uses an iterative inversion algorithm based on multivariate feature deviations, which has adaptive capabilities and can cover a variety of complex bathroom scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flow chart of the water mist clutter filtering method of the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.

[0042] refer to Figure 1 The steps of the water mist clutter filtering method of the present invention are as follows:

[0043] Radar original echo signal: radar configuration initialization parameters, multi-transmitter multi-receiver millimeter-wave radar transmits electromagnetic waves to collect original echo data;

[0044] Range Doppler data: The original echo data is Fourier transformed to obtain the range Doppler data of each virtual array element;

[0045] Millimeter-wave radar angle measurement: Perform high-precision super-resolution angle measurement on the original Doppler data to obtain three-dimensional spatial data of dynamic targets in the monitoring scene;

[0046] CFAR constant false alarm detection: CFAR constant false alarm detection first processes the input signal and determines a threshold, then outputs 3D target point cloud data;

[0047] Multivariate feature deviation iterative inversion algorithm: Based on the multivariate features such as energy spectrum, three-dimensional coordinate information, velocity, signal-to-noise ratio, etc. in the target point cloud and Doppler data information, the sliding time window calculates the total deviation coefficient α of each frame and the average value of the current N frames. The present invention uses this total deviation coefficient as an influencing factor to iteratively invert the original echo signal to obtain new echo data, and repeats the cycle many times until α < α is satisfied. T Or M>M T , where α T is the preset training error threshold, M is the number of iterations, and M T The maximum number of iterations allowed.

[0048] Human target tracking and fall detection: After the multivariate feature deviation iterative inversion algorithm, the Kalman filter algorithm is used to track and filter the target, and finally a stable moving target track is obtained. The posture of the human target is identified according to the target height and width in the track information, and it is accurately determined whether the human target has fallen in the bathroom.

[0049] The above step 5 is based on the iterative inversion algorithm of multivariate feature deviation. The range Doppler data and target point cloud data formed in step 2, step 3 and step 4 have multivariate feature quantities such as target energy spectrum, three-dimensional coordinate information, speed, signal-to-noise ratio, etc. After collecting multiple sets of measured data, the statistical results of target feature changes are as follows:

[0050] The characteristic distribution of radar data of human targets is as follows:

[0051] 1) The azimuth and distance directions of the three-dimensional coordinate information are active and changeable within the detection range, and the height direction is uniformly between 0 and 1.8m;

[0052] 2) The velocity distribution varies from 0 to 5 m / s, depending on the person's motion state, and approaches 0 when stationary;

[0053] 3) When a person is at rest, the energy is small and close to 0, and the energy fluctuates when moving;

[0054] 4) The signal-to-noise ratio of human targets is high.

[0055] The characteristic distribution characteristics of radar data of water mist clutter from various shower heads are as follows:

[0056] 1) The azimuth, distance and height of the three-dimensional coordinate information are fixed and unchanged;

[0057] 2) The speed is relatively constant and uniform, which is related to the water jet speed;

[0058] 3) The energy is evenly distributed without obvious fluctuations for a long time;

[0059] 4) The signal-to-noise ratio is low.

[0060] The multivariate feature deviation iterative inversion algorithm of the present invention has adaptive capability, and assigns weight factor ε to the diversified features through feature quantities analyzed by a large number of samples. Then, a sliding time window is used to calculate the deviation of each feature between each frame and the average value of the current N frames, and the final total deviation coefficient α is obtained by combining the weight factors of each feature.

[0061] The specific calculation formula is as follows, and the weighted coefficient after the multivariate eigenvalue is compounded is:

[0062]

[0063] Where α is the total deviation coefficient after the multivariate eigenvalues ​​are compounded; ε0 is the weight factor of the range Doppler energy spectrum preset according to the statistical results of the measured data; α0 is the deviation coefficient of the range Doppler energy; ε i Preset weight factors for other multivariate features (3D coordinate information, speed, signal-to-noise ratio); α i is the deviation coefficient of other multivariate features (3D coordinate information, speed, signal-to-noise ratio); N is the number of other multivariate features.

[0064] The derivation formulas of various multivariate feature deviations are expressed as follows:

[0065]

[0066]

[0067] Among them, σ0 is the range Doppler energy deviation; is the average value of the range Doppler energy sliding window; σ i is the deviation of other multivariate features (3D coordinate information, speed, signal-to-noise ratio); is the current N-frame sliding window average of other multivariate features (3D coordinate information, speed, signal-to-noise ratio);

[0068] The state value of the current N-frame sliding window average is expressed as follows. The sliding window size N is affected by the maximum energy of the target point cloud. If the maximum energy of the target point cloud is too large, it means that the target detection is stable and the clutter is relatively weak, so the number of sliding windows can be appropriately reduced. On the contrary, it means that the target energy is weak and the clutter is relatively strong, so the number of sliding windows can be appropriately increased:

[0069] N=[power max *δ]

[0070]

[0071]

[0072] Where power max is the maximum value of the range Doppler energy spectrum in the current frame, δ is the sliding window adjustment coefficient, when the clutter is relatively strong, the δ value is set larger, when the clutter is relatively weak, the δ value is set smaller; N is the sliding window size, where A0(j) is the range Doppler energy spectrum of the jth frame; when i=1, A i (j) is the 3D coordinate value of the jth frame; when i=2, A i (j) is the velocity value of the jth frame; when i=3, A i (j) is the signal-to-noise ratio value of the jth frame.

[0073] The energy spectrum and other multivariate eigenvalue deviations are derived as follows:

[0074]

[0075]

[0076] After the total deviation coefficient is obtained, iterative inversion is required to reconstruct new echo data from the original echo data and the total deviation coefficient.

[0077] The specific steps are as follows:

[0078] The complex signal formula model e(t,z) of the transmission echo of the kth virtual array element in the millimeter-wave original radar multi-transmit and multi-receive array is expressed as follows:

[0079]

[0080] Where A is the amplitude of the transmitted signal; f is the frequency of the radar transmitted signal; t represents the fast time with each frequency modulation as the starting point; T c is the frequency modulation period; K is the frequency modulation slope; is the initial phase of the radar transmitting signal; z is the frequency modulation index of the transmitting antenna.

[0081] Among them, the sum of the original echo signals of all the transmitting and receiving virtual array elements is the superimposed signal E(t,z), where j represents the reception of the jth virtual array element:

[0082]

[0083] The new echo complex signal model E'(t,z) is constructed by the weighted coefficient α after the multivariate eigenvalue compounding:

[0084]

[0085] Repeatedly construct a new echo complex signal model until α < α is satisfied T Or M>M T (where α T is the preset training error threshold, M is the number of iterations, and M T To allow the maximum number of iterations to be executed), iterative inversion is performed to obtain converged echo data, and Fourier transform, angle measurement and CFAR detection radar signal processing are continued. Finally, the target is tracked and filtered through the Kalman filter algorithm, and a stable moving target track is obtained. The motion posture of the human target is identified according to the trajectory information, and it is accurately determined whether the human target falls in the bathroom.

[0086] Compared with the prior art, the present invention has the following beneficial effects:

[0087] It uses millimeter-wave radar processing algorithms to detect human body falling postures, and has the performance of not being affected by light and water mist. Compared with traditional cameras, it has the advantage of protecting personal privacy.

[0088] The diversified feature extraction and echo iterative inversion method is used to effectively suppress static object clutter and false target interference caused by water mist sprayed from various shower heads. Experiments have verified that the fall radar has a good detection effect.

[0089] It uses an iterative inversion algorithm based on multivariate feature deviations, which has adaptive capabilities and can cover a variety of complex bathroom scenes.

[0090] As used herein, the word "preferred" is intended to be used as an example, instance, or illustration. Any aspect or design described herein as "preferred" is not necessarily to be construed as being more advantageous than other aspects or designs. On the contrary, the use of the word "preferred" is intended to present concepts in a specific way. The term "or" as used in this application is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" means any one of the naturally included permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.

[0091] Moreover, although the present disclosure has been shown and described with respect to one or implementations, those skilled in the art will think of equivalent variations and modifications based on the reading and understanding of this specification and the accompanying drawings. The present disclosure includes all such modifications and variations, and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the above-mentioned components (such as elements, etc.), the terms used to describe such components are intended to correspond to any component (unless otherwise indicated) that performs the specified function of the component (such as it is functionally equivalent), even if the structure is not equivalent to the disclosed structure of the function in the exemplary implementation of the present disclosure shown herein. In addition, although the specific features of the present disclosure have been disclosed with respect to only one of several implementations, such features can be combined with one or other features of other implementations that may be desired and advantageous for a given or specific application. Moreover, insofar as the terms "including", "having", "containing" or their variations are used in specific embodiments or claims, such terms are intended to be included in a manner similar to the term "comprising".

[0092] The functional units in the embodiments of the present invention may be integrated into a processing module, or each unit may exist physically separately, or multiple or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc. The above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.

[0093] To sum up, the above embodiment is an implementation mode of the present invention, but the implementation mode of the present invention is not limited by the embodiment. Any other changes, modifications, substitutions, combinations, and simplifications that deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for filtering water mist clutter of bathroom fall detection radar, characterized in that: The following steps are involved: The radar configuration initializes parameters, and the multi-transmitter and multi-receiver millimeter-wave radar transmits electromagnetic waves to collect raw echo data; The original echo data is Fourier transformed to obtain the range Doppler data of each virtual array element; Performing high-precision super-resolution angle measurement on the range Doppler data to obtain three-dimensional spatial data of dynamic targets in the monitoring scene; Use CFAR constant false alarm detection to process the input signal and determine a threshold, then output 3D target point cloud data; The energy spectrum, three-dimensional coordinate information, speed, and signal-to-noise ratio in the three-dimensional target point cloud data and range Doppler data information are obtained, and multivariate features are formed. The total deviation coefficient of each frame and the average value of the current N frames is calculated using the sliding time window method. , using the total deviation coefficient as an influencing factor to iteratively invert the original echo signal to obtain converged echo data; The total coefficient of variation for: ； in is the total deviation coefficient after the multivariate eigenvalues ​​are combined; is a weight factor of the range Doppler energy spectrum preset according to the statistical results of the measured data; is the deviation coefficient of range Doppler energy; , , The weight factors of the preset three-dimensional coordinate information, speed, and signal-to-noise ratio; , , is the deviation coefficient of three-dimensional coordinate information, speed, and signal-to-noise ratio; n is 3.

2. The method for filtering water mist clutter of bathroom fall detection radar according to claim 1, characterized in that: The converged echo data is filtered using the Kalman filter algorithm to track the target and obtain a stable moving target track. The motion posture of the human target is then identified based on the trajectory information to determine whether the human target has fallen in the bathroom.

3. The method for filtering water mist clutter of bathroom fall detection radar according to claim 1, characterized in that: The calculation formula of the deviation coefficient of multivariate features is expressed as follows: ； in, is the range Doppler energy deviation; is the current N-frame sliding window average value of the range Doppler energy; , , is the deviation of three-dimensional coordinate information, speed, and signal-to-noise ratio; is the current N-frame sliding window average of 3D coordinate information, speed, and signal-to-noise ratio; The current N-frame sliding window average of multivariate features is expressed as follows: ； in power max is the maximum value of the range Doppler energy spectrum in the current frame, is the sliding window adjustment coefficient, N is the sliding window size, where A 0( j ) is the j Range Doppler power spectrum of the frame; i =1, A i ( j ) is the j The three-dimensional coordinate value of the frame; i =2, A i ( j ) is the j The speed value of the frame; i =3, A i ( j ) is the j The signal-to-noise ratio value of the frame.

4. The method for filtering water mist clutter of bathroom fall detection radar according to claim 1, characterized in that: The formula for calculating the deviation of energy spectrum and other multivariate features is as follows: 。 5. The method for filtering water mist clutter of bathroom fall detection radar according to claim 1, characterized in that: The iterative inversion of the original echo signal by using the total deviation coefficient as an influencing factor to obtain converged echo data includes: Millimeter wave original radar multiple transmit and multiple receive array k The formula model of the complex signal of the transmitted echo of virtual array elements The expression is as follows: ； in A is the transmitted signal amplitude; f The frequency of the radar transmitting signal; t Indicates the fast time with each frequency modulation as the starting point of time; T c is the frequency modulation period; K is the frequency modulation slope; is the initial phase of the radar transmission signal; z is the frequency modulation index of the transmitting antenna; Among them, the sum of the original echo signals of all the transmitting and receiving virtual array elements is E ( t , z ),in j Expressed as j Reception of virtual array elements: ； The total deviation coefficient is Constructing a new echo complex signal model E ( t , z ): ； Repeatedly construct a new echo complex signal model until it satisfies or , in, is the preset training error threshold, M is the number of iterations, M T The maximum number of iterations allowed.

Images