A human body recognition algorithm and sensor based on Doppler microwave sensing

Abstract

The present invention solves the problem that microwave sensing cannot effectively detect stationary or slightly moving objects, and proposes a human body recognition algorithm and sensor based on Doppler microwave sensing. The human sensing detector assembled through microwave radar devices continuously obtains dynamic data of human targets detected by the human sensing detector, and then uses the human body judgment algorithm to judge whether there is a person or not, so as to obtain accurate human body detection information, providing effective conditions for subsequent human sensing applications.

Description

Technical Field

[0001] The present invention relates to the technical field of human body sensing, in particular to a human body recognition algorithm and sensor based on Doppler microwave sensing. Background Art

[0002] The human body sensing technologies currently on the market mainly include infrared sensing and Doppler microwave sensing:

[0003] Infrared sensing: Utilizing the pyroelectric infrared sensing principle, the sensor triggers its action by collecting infrared energy changes. This is affected by ambient temperature. In the summer, when the ambient temperature is close to human body temperature, there is no infrared energy change when a person enters the sensing range, resulting in very low sensitivity to human body signals. Disadvantages: The sensor is significantly affected by ambient temperature, and the infrared probe must be exposed, which reduces the overall aesthetics of the product.

[0004] Doppler microwave sensing: The wavelength of radiation emitted by an object changes due to the relative motion between the source and the observer. In front of the moving source, the wave is compressed, resulting in a shorter wavelength and a higher frequency. When moving behind the source, the opposite effect occurs: the wavelength becomes longer and the frequency becomes lower. The difference between the transmitted and received frequencies caused by the Doppler effect is called the Doppler shift. It reveals how wave properties change during motion. Microwave radar sensing simultaneously transmits a microwave signal and receives the reflected wave. The difference between the two signals is mixed to produce a new low-frequency signal, called an intermediate frequency (IF), which is the difference between the transmitted and reflected frequencies. The higher the operating frequency of microwave radar sensing, the narrower the gain and detection beam, and the longer the detection range. Using the Doppler principle, a moving object entering the sensing range changes the radar signal waveform, triggering the radar sensor. Features: It only senses moving people and can penetrate plastic for concealed installation, making it aesthetically pleasing. Disadvantages: It cannot sense people who are sitting still or sleeping.

[0005] Reference is made to a split-type microwave sensing LED lamp, a microwave sensing device, and an LED lamp with Chinese patent publication number CN204119580U, wherein the split-type microwave sensing LED lamp includes a microwave sensing device and an LED lamp, wherein the microwave sensing device and the LED lamp are wirelessly connected; the microwave sensing device is used to detect human movement and send a wireless control signal; the LED lamp has a built-in wireless communication module that can receive the wireless control signal and control the LED lamp according to the signal.

[0006] Microwaves can detect the distance, relative speed, and azimuth of objects, but they cannot effectively detect stationary or slightly moving objects, and therefore cannot accurately detect the presence of human bodies in indoor spaces. This results in a poor user experience in many applications that determine the presence of human beings. For example, in the indoor application scenario of "turning on the lights when people arrive and turning them off when they leave," the lights may be turned off when people enter a slightly moving state or a completely stationary state, or the number of people counted indoors may not be accurate. Summary of the Invention

[0007] The present invention solves the problem that microwave sensing cannot effectively detect stationary or slightly moving objects, and proposes a human body recognition algorithm and sensor based on Doppler microwave sensing. The human sensing detector assembled through microwave radar devices continuously obtains dynamic data of human targets detected by the human sensing detector, and then uses the human body judgment algorithm to judge whether there is a person or not, so as to obtain accurate human body detection information, providing effective conditions for subsequent human sensing applications.

[0008] In order to achieve the above objectives, the following technical solutions are proposed:

[0009] A human body recognition algorithm based on Doppler microwave sensing includes the following steps:

[0010] S1, acquiring dynamic data of a human target in real time based on Doppler microwave sensing, wherein the dynamic data includes a distance λ, a relative velocity v, and an azimuth angle φ from the human target to a test point;

[0011] S2, using the DBSCAN clustering algorithm to classify and label the dynamic data, and obtaining the data geometric center of the dynamic data by class to generate classification information, the classification information including:

[0012] Classification mark Oid;

[0013] The geometric center of the classified data Oloc[λ0, φ0];

[0014] Array D[[λ1, ν1, φ1], [λ2, ν2, φ2], …, [λn, νn, φn]] of dynamic data;

[0015] S3, set the existing target location data set Ldata (Oid, Oloc, D),

[0016] The distance change value Δλ and the azimuth change value Δφ between the real-time target position Oloc[λ(t), φ(t)] and the existing target position data are calculated in sequence to obtain a set of change values [Δλ, Δφ]; where: λ(t) is the distance value of the geometric center of the target object acquired at the current moment; φ(t) is the azimuth value of the geometric center of the target object acquired at the current moment;

[0017] S4, substitute a set of change values [Δλ, Δφ] into the correlation formula:

[0018] Y(t)=Δλ+2*λ(t)*sin(Δφ / 2),

[0019] Where Y(t) is the change space distance, a set of change space distance [Y] values is obtained, and it is determined whether the minimum Y value is less than Ymin. If so, the classification information corresponding to the minimum Y value is updated to the existing target position data set; if not, the classification information corresponding to the minimum Y value is added to the existing target position data set; Ymin is the minimum allowable Euclidean distance reference error value;

[0020] S5, extract the target position data set record that has not been updated, compare the relative speed of the last updated record to see if it is greater than the reference speed Vmin, if so, determine that the record is in an invalid state and remove it from the existing target position data set, if not, do nothing;

[0021] S6, counting the number of records in the target location data set, when the number is greater than 0, determining that the current state is a occupied state, when the number is 0, determining that the current state is an unoccupied state.

[0022] The present invention continuously obtains the distance λ, relative speed v and azimuth φ from the human target to the test point based on Doppler microwave sensing, then calculates the position distribution range of the human target through a human body judgment algorithm, and then matches the relative movement speed with the distribution range of the previous position. This can improve the matching hit rate, that is, improve the recognition rate of microwave sensing for stationary objects, so as to obtain accurate human body detection information and provide effective conditions for subsequent human sensing applications.

[0023] Preferably, the distance λ is calculated as follows:

[0024]

[0025] Where: c is the speed of light, T is the period of the modulated triangle wave, ΔF is the frequency modulation bandwidth, fb+ is the intermediate frequency output on the rising edge of the triangle wave, and fb- is the intermediate frequency output on the falling edge of the triangle wave;

[0026] The calculation formula of the relative speed v is as follows:

[0027]

[0028] Where: f is the frequency of the transmitted signal;

[0029] The calculation formula of the azimuth angle φ is as follows:

[0030]

[0031] Where: Δd is the differential distance.

[0032] Preferably, the S2 further includes setting profile coefficients of the distance λ, relative speed v and azimuth angle φ, and adjusting the coefficients according to actual working conditions.

[0033] The present invention sets default profile radius coefficients for distance, relative speed, and azimuth based on experience, and adjusts these coefficients based on actual operating conditions. Based on the currently set profile coefficients, the DBSCAN clustering algorithm is used to classify and label the collected reflection point signals [λn, vn, φn]. Using the clustering algorithm to classify the collected reflection point signals, the profile coefficient values can be adjusted through machine learning to obtain more accurate classification results.

[0034] Preferably, the calculation formula of Δλ is as follows:

[0035] Δλ=λ(i)-λ(t),

[0036] Where: λ(i) is the distance value of the data geometric center Oloc stored in the existing target location dataset Ldata;

[0037] The calculation formula of Δφ is as follows:

[0038] Δφ=φ(i)-φ(t), where: φ(i) is the azimuth value of the data geometric center Oloc stored in the existing target position data set Ldata.

[0039] A human body recognition sensor based on Doppler microwave sensing adopts the above-mentioned human body recognition algorithm based on Doppler microwave sensing, including a microwave radar, a main control module and a storage module. The microwave radar is used to use Doppler microwave sensing to obtain dynamic data of human targets in real time and transmit the dynamic data to the main control module. The main control module is equipped with a human body recognition algorithm to obtain human body recognition results based on the obtained dynamic data. The storage module is used to store and update dynamic data records.

[0040] The beneficial effects of the present invention are as follows: the present invention continuously obtains the distance λ, relative speed v and azimuth φ from the human target to the test point based on Doppler microwave sensing, and then calculates the position distribution range of the human target through a human judgment algorithm, and then matches the relative movement speed with the distribution range of the previous position, which can improve the matching hit rate, that is, improve the recognition rate of microwave sensing for stationary objects, so as to obtain accurate human detection information and provide effective conditions for subsequent human sensing applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is an algorithm flow chart of an embodiment;

[0042] Figure 2 7 is a diagram of a continuous frequency modulation microwave waveform of a microwave radar of an embodiment;

[0043] Figure 3 The present invention is a schematic diagram of the principle of microwave radar measuring azimuth angle in an embodiment. DETAILED DESCRIPTION

[0044] Example:

[0045] This embodiment proposes a human body recognition algorithm based on Doppler microwave sensing, referring to Figure 1 , including the following steps:

[0046] S1, based on Doppler microwave sensing, obtains the dynamic data of the human target in real time. The dynamic data includes the distance λ, relative velocity v and azimuth angle φ from the human target to the test point;

[0047] Currently, human body sensors using Doppler microwave sensing technology are primarily microwave radars. Microwave radars used for human body detection typically employ the simple, low-cost FMCW radar system, which is suitable for close-range detection. Microwave radars can measure the distance, speed, and direction of moving objects.

[0048] Microwave radar emits directional microwaves of the appropriate wavelength through a transmitting antenna. When these microwaves encounter obstacles or targets, they are reflected back and received by a receiving antenna. Based on the microwave's wavelength, the time it takes for the microwaves to travel is calculated using the formula: τ = 2 * distance / speed of light. Knowing the relative distance between the microwave radar and a moving target allows the target's position to be determined.

[0049] The radar antenna emits a series of continuous frequency-modulated microwaves. The frequency changes with time according to the modulation voltage. It is generally a continuous triangular wave. The transmission and reception signals are referenced Figure 2 shown. Figure 2 The solid line in the middle is the transmitted signal, and the dashed line is the reflected signal from a relatively stationary and moving object. The reflected and transmitted signal waveforms are identical, differing only by a delay time td.

[0050] The frequency difference between the transmitted and reflected signals at a given moment is the intermediate frequency (IF) fb of the mixing output. The Doppler effect causes a frequency shift in the reflected signal from a moving object. The IF frequencies output at the rising and falling edges of the triangle wave are fb+ and fb-, respectively.

[0051] The calculation formula of distance λ is as follows:

[0052]

[0053] Where: c is the speed of light, T is the period of the modulated triangle wave, ΔF is the frequency modulation bandwidth, fb+ is the intermediate frequency output on the rising edge of the triangle wave, and fb- is the intermediate frequency output on the falling edge of the triangle wave;

[0054] In this embodiment: the speed of light c = 300,000,000 m / s, the minimum frequency of the difference signal of fb- plus fb+ = fs / N = 500 Hz, at the sampling frequency fs = 512 kHz, the number of sampling points N = 1024, ΔF = 1 GHz, T = 2 ms, and the geometric distance Δd between the two receiving antennas of the human detector = 1 cm.

[0055] Due to the Doppler effect, the frequency variation of microwave radar is closely related to the relative speed between the radar and the object. The relative speed of the radar and the object can be determined by the frequency variation of the reflected microwaves. The closer the radar and the object, the higher the frequency of the reflected microwave echo, and vice versa.

[0056] The calculation formula of relative velocity v is as follows:

[0057]

[0058] Where: f is the frequency of the transmitted signal;

[0059] refer to Figure 3 The microwave radar's two receiving antennas can calculate the azimuth of the monitored target by measuring the phase difference between microwaves reflected from the same target. The azimuth angle φ is calculated using trigonometric calculations based on the geometric distance d between microwave radar receiving antennas RX1 and RX2 and the phase difference b between the reflected echoes received by the two microwave radar antennas.

[0060] The calculation formula of the azimuth angle φ is as follows:

[0061]

[0062] Where: Δd is the differential distance from the target to each antenna.

[0063] S2, set the silhouette coefficients of distance λ, relative speed ν, and azimuth angle φ, and adjust the coefficients according to the actual working conditions. Based on the currently set silhouette coefficients, use the DBSCAN clustering algorithm to classify and label the dynamic data, and obtain the data geometric center of the dynamic data by class to generate classification information. The classification information includes:

[0064] Classification mark Oid;

[0065] The geometric center of the classified data Oloc[λ0, φ0];

[0066] Array D[[λ1, ν2, φ1], [λ2, ν2, φ2], …, [λn, νn, φn]] of dynamic data;

[0067] According to the calculation formula:

[0068]

[0069] Find the minimum distance:

[0070] Δλ=300000000*(2 / 1000)*500 / 4*10000000000=0.075; set the silhouette coefficient as 8 times the minimum calculated data; considering that the thickness of the human body is generally 300mm, the distance difference must meet the minimum condition of greater than 0.3m;

[0071] Calculation results:

[0072] The silhouette coefficient of the distance λ is 8*0.075=0.6.

[0073] According to the calculation formula:

[0074]

[0075] Minimum speed v = 300000000 / (4*512000)*((1 / 2*1000) / 2) = 0.0366

[0076] The silhouette coefficient of velocity v is 8*0.0366=0.29.

[0077] According to the calculation formula:

[0078]

[0079] Calculation results:

[0080] The silhouette coefficient of the azimuth angle φ is 8*(2*3.14*0.01 / 0.075)=0.037.

[0081] The present invention sets default profile radius coefficients for distance, relative speed, and azimuth based on experience, and adjusts these coefficients based on actual operating conditions. Based on the currently set profile coefficients, the DBSCAN clustering algorithm is used to classify and label the collected reflection point signals [λn, vn, φn]. Using the clustering algorithm to classify the collected reflection point signals, the profile coefficient values can be adjusted through machine learning to obtain more accurate classification results.

[0082] S3, set the existing target location data set Ldata (Oid, Oloc, D), and sequentially calculate the distance change value Δλ and azimuth change value Δφ between the real-time target location Oloc [λ(t), φ(t)] and the existing target location data to obtain a set of change values [Δλ, Δφ]; where: λ(t) is the distance value of the geometric center of the target object acquired at the current moment; φ(t) is the azimuth value of the geometric center of the target object acquired at the current moment;

[0083] The calculation formula of Δλ is as follows:

[0084] Δλ=λ(i)-λ(t),

[0085] Where: λ(i) is the distance value of the data geometric center Oloc stored in the existing target location dataset Ldata;

[0086] The calculation formula of Δφ is as follows:

[0087] Δφ=φ(i)-φ(t),

[0088] Where: φ(i) is the azimuth value of the data geometric center Oloc stored in the existing target location dataset Ldata. S4, substitute a set of change values [Δλ, Δφ] into the correlation formula:

[0089] Y(t)=Δλ+2*λ(t)*sin(Δφ / 2),

[0090] Where Y(t) is the change spatial distance, a set of change spatial distance [Y] values is obtained, and it is determined whether the minimum Y value is less than Ymin. If so, the classification information corresponding to the minimum Y value is updated to the existing target position data set; if not, the classification information corresponding to the minimum Y value is added to the existing target position data set; Ymin is the minimum allowable Euclidean distance reference error value; in this embodiment, Ymin = 40 cm.

[0091] S5. Extract the target position data set record that has not been updated yet and compare whether the relative speed of the last updated record is greater than the reference speed Vmin. If so, the record is deemed invalid and removed from the existing target position data set. If not, no operation is performed. In this embodiment, Vmin = 40 cm / s.

[0092] S6, counting the number of records in the target location data set, when the number is greater than 0, determining that the current state is a occupied state, when the number is 0, determining that the current state is an unoccupied state.

[0093] The present invention continuously obtains the distance λ, relative speed ν and azimuth φ of the human target from the test point based on Doppler microwave sensing, then calculates the position distribution range of the human target through a human body judgment algorithm, and then matches the relative movement speed with the distribution range of the previous position. This can improve the matching hit rate, that is, improve the recognition rate of microwave sensing for stationary objects, so as to obtain accurate human body detection information and provide effective conditions for subsequent human sensing applications.

[0094] This embodiment also proposes a human body recognition sensor based on Doppler microwave sensing, which adopts the above-mentioned human body recognition algorithm based on Doppler microwave sensing, including a microwave radar, a main control module and a storage module. The microwave radar is used to use Doppler microwave sensing to obtain dynamic data of human targets in real time and transmit the dynamic data to the main control module. The main control module is equipped with a human body recognition algorithm to obtain human body recognition results based on the obtained dynamic data. The storage module is used to store and update dynamic data records.

Claims

1. A human body recognition algorithm based on Doppler microwave sensing, characterized in that: The following steps are involved: S1, acquiring dynamic data of a human target in real time based on Doppler microwave sensing, wherein the dynamic data includes a distance λ, a relative velocity v, and an azimuth angle φ from the human target to a test point; S2, using the DBSCAN clustering algorithm to classify and label the dynamic data, and obtaining the data geometric center of the dynamic data by class to generate classification information, the classification information including: Classification mark Oid; The geometric center of the classified data Oloc[λ0, φ0]; Array D[[λ1, v1, φ1], [λ2, v2, φ2], …, [λn, νn, φn]] of dynamic data; S3, set the existing target location data set Ldata (Oid, Oloc, D), The distance change value Δλ and the azimuth change value Δφ between the real-time target position Oloc[λ(t), φ(t)] and the existing target position data are calculated in sequence to obtain a set of change values [Δλ, Δφ]; where: λ(t) is the distance value of the geometric center of the target object acquired at the current moment; φ(t) is the azimuth value of the geometric center of the target object acquired at the current moment; S4, substitute a set of change values [Δλ, Δφ] into the associated formula: Y(t)=Δλ+2*λ(t)*sin(Δφ / 2), Where Y(t) is the change space distance, a set of change space distance [Y] values is obtained, and it is determined whether the minimum Y value is less than Ymin. If so, the classification information corresponding to the minimum Y value is updated to the existing target position data set; if not, the classification information corresponding to the minimum Y value is added to the existing target position data set; Ymin is the minimum allowable Euclidean distance reference error value; S5, extract the target position data set record that has not been updated, compare the relative speed of the last updated record to see if it is greater than the reference speed Vmin, if so, determine that the record is in an invalid state and remove it from the existing target position data set, if not, do nothing; S6, counting the number of records in the target location data set, when the number is greater than 0, determining that the current state is a occupied state, when the number is 0, determining that the current state is an unoccupied state.

2. The human body recognition algorithm based on Doppler microwave sensing according to claim 1 is characterized in that: The calculation formula of the distance λ is as follows: Where: c is the speed of light, T is the period of the modulated triangle wave, ΔF is the frequency modulation bandwidth, fb+ is the intermediate frequency output on the rising edge of the triangle wave, and fb- is the intermediate frequency output on the falling edge of the triangle wave; The calculation formula of the relative speed v is as follows: Where: f is the frequency of the transmitted signal; The calculation formula of the azimuth angle φ is as follows: Where: Δd is the differential distance.

3. The human body recognition algorithm based on Doppler microwave sensing according to claim 1 is characterized in that: Said S2 also includes setting the profile coefficients of the distance λ, the relative speed v and the azimuth angle φ, and adjusting the coefficients according to the actual working conditions.

4. The human body recognition algorithm based on Doppler microwave sensing according to claim 1 is characterized in that: The calculation formula of Δλ is as follows: Δλ=λ(i)-λ(t), Where: λ(i) is the distance value of the data geometric center Oloc stored in the existing target location dataset Ldata; The calculation formula of Δφ is as follows: Δφ=φ(i)-φ(t), Where: φ(i) is the azimuth value of the data geometric center Oloc stored in the existing target location dataset Ldata.

5. The human body recognition algorithm based on Doppler microwave sensing according to claim 1 is characterized in that: The reference speed Vmin is 40 cm / s.

6. A human body recognition sensor based on Doppler microwave induction, using the human body recognition algorithm based on Doppler microwave induction according to claim 1, characterized in that: It includes a microwave radar, a main control module and a storage module. The microwave radar is used to use Doppler microwave sensing to obtain dynamic data of human targets in real time and transmit the dynamic data to the main control module. The main control module is equipped with a human recognition algorithm to obtain human recognition results based on the obtained dynamic data. The storage module is used to store and update dynamic data records.

Images