A method for human body existence detection based on distance-doppler domain millimeter wave radar

Abstract

The application belongs to the technical field of millimeter wave radar indoor human body detection application, and particularly relates to a millimeter wave radar human body existence detection method based on a distance-Doppler domain. The method filters out echo signals from static clutter, excludes static object interference, distinguishes moving human bodies based on distance change information, combines distance information and Doppler information, and distinguishes interference sources and static human bodies. The method improves the accuracy and robustness of human body detection, reduces the false alarm probability, and improves the detection effect.

Description

Technical Field

[0001] This invention belongs to the field of millimeter-wave radar application technology, specifically relating to a millimeter-wave radar human presence detection method based on the range-Doppler domain. Background Technology

[0002] In recent years, indoor human detection has found a wide range of applications, such as automatic lighting in smart homes, intrusion detection in security and surveillance, and daily activity monitoring in healthcare. To achieve indoor human detection, various types of sensors are typically used, such as visual sensors, infrared sensors, and radar sensors. However, visual sensors are affected by lighting conditions and may infringe on user privacy; infrared sensors are affected by thermal radiation and have lower resolution; millimeter-wave radar, on the other hand, offers all-weather, all-day operation, protects user privacy, and has high resolution. By emitting electromagnetic wave signals through a radar antenna, which are reflected when obstructed by a target, and then receiving and processing the reflected signals, information such as target distance and Doppler readings can be determined.

[0003] Existing research on indoor human detection mainly includes: (1) point cloud clustering, which uses radar to acquire target point cloud data. Point cloud is a three-dimensional dataset composed of discrete points. Clustering algorithms (such as k-means clustering, DBSCAN, etc.) are used to group the points in the point cloud to identify and locate human bodies. (2) deep learning classification, which uses convolutional neural networks or recurrent neural networks to train models for detecting and classifying human bodies. However, the above research is highly complex and requires high radar range resolution and velocity resolution. It is not suitable for low-frequency and narrow-band radars. Furthermore, it does not consider the impact of interference, which leads to the misjudgment of interference as human bodies in practical applications, resulting in false alarms. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems and shortcomings, and to overcome the issues of poor interference elimination and numerous false alarms in actual indoor scenarios caused by current point cloud data and deep learning-based methods. This invention provides a distance-Doppler information fusion method, which is applied to the differentiation between stationary human targets and moving human targets and interference, and has achieved excellent results in indoor human presence detection.

[0005] The technical solution of this invention is: a millimeter-wave radar anti-interference human presence detection method based on the range-Doppler domain, such as... Figure 1 As shown, it includes:

[0006] Step 1: Acquire radar echoes and perform 1DFFT on them to obtain range-slow time dimension data;

[0007] Step 2: Perform static clutter filtering on the distance-slow time dimension data. Specifically, forward difference is used: subtract the distance-slow time dimension data from the previous time dimension from the distance-slow time dimension data acquired at the current time, and traverse all distance-slow time dimension data to eliminate static clutter with constant energy and highlight the signal with energy change.

[0008] Step 3: Perform 2DFFT on the data after static clutter filtering to obtain range-Doppler data;

[0009] Step 4: For the data after static clutter filtering, determine whether the signal energy fluctuation exceeds a threshold to identify potential targets, including:

[0010] Select all slow-time dimension data under each distance gate, calculate the ratio of the maximum value to the mean value of the current distance gate signal energy. If the ratio is greater than n, it is determined that there is a potential target at the current distance gate. n is set according to experience.

[0011] Step 5: For the potential targets selected in the previous step, determine the moving targets and candidate targets by using the distance gate span decision. Specifically, the span decision is adopted: for a potential target, if the potential target appears in a continuous time and spans more than one continuous distance gate, that is, there is a continuous trajectory in the distance-slow time dimension, it is determined to be a moving target. If the potential target only appears in one distance gate, it is determined to be a candidate target.

[0012] Step 6: For candidate targets, based on the range-Doppler data obtained in Step 3, extract the Doppler information of the candidate targets, and combine the range and Doppler information to determine stationary human bodies and interference, including:

[0013] For the distance gate of the candidate target, extract its corresponding Doppler information. If the Doppler velocity is 0, it is judged to be a stationary human body. If the Doppler velocity is greater than 0, it is judged to be interference. Interference categories include oscillating fans and green plants.

[0014] Step 7: Mark the corresponding range gate for the detected interference, subtract the interference signal of that range gate from the range-slow time dimension data, and repeat this process until no signal energy meets the threshold, at which point the loop ends.

[0015] Step 8: Output the detection results. Candidate targets with non-zero Doppler velocity are judged as interference, candidate targets with zero Doppler velocity are judged as stationary human bodies, moving targets are judged as moving human bodies, and targets whose signal energy does not meet the threshold are judged as human bodies not existing.

[0016] The beneficial effects of this invention are as follows: Based on the traditional range-slow time dimension of millimeter-wave radar, it proposes a human detection method that fuses range-Doppler data. A key innovation is the ability to detect moving human bodies, stationary human bodies, and interference while maintaining low computational complexity. Compared to traditional detection algorithms, it first eliminates static object interference through static clutter filtering; secondly, it introduces a method for setting thresholds based on signal energy fluctuations to determine the presence of potential targets at the current range gate. This dynamic threshold setting allows the system to adaptively adapt to different environments and target characteristics, improving detection robustness; simultaneously, it introduces a range gate span decision method to distinguish between continuously moving targets with changing positions and targets with unchanged positions; finally, by extracting Doppler information from candidate targets, it can identify stationary human bodies and interference sources, such as oscillating fans and green plants. In summary, this invention improves the accuracy and robustness of indoor human detection, and this method is suitable for applications requiring target detection and classification in complex indoor environments. Attached Figure Description

[0017] Figure 1 This is the overall flowchart of the present invention. Detailed Implementation

[0018] Figure 1 This paper demonstrates a human presence detection system based on joint decision-making in the range-Doppler domain. A top-down design approach is used to detail the algorithm architecture. The algorithm module primarily consists of two modules: a range-slow time dimension module and a range-Doppler dimension module, which perform joint decision-making. The data processing relationships between these modules in the overall process are as follows: Figure 1 As shown.

[0019] After the millimeter-wave radar receives the echo signal, it performs 1DFFT processing to obtain preliminary range-slow time dimension data. Forward difference is then applied to this data to filter out static clutter. On one hand, the signal fluctuation amplitude of the clutter-filtered range-slow time dimension data is assessed. An adaptive threshold is set; if the signal fluctuation exceeds this threshold, a suspected target is identified at that range gate. If, after traversing all range gates, all signal fluctuations are less than the threshold, it is determined that there are no targets in the scene, and no human body is present. On the other hand, 2DFFT is performed on the clutter-filtered data to obtain range-Doppler domain data. Further judgment is made for range gates with suspected targets. If the signal crosses multiple consecutive range gates, a moving target is identified at that range gate, confirming the presence of a moving human body. If the signal continuously appears at the same range gate, the velocity of the corresponding range gate in the range-Doppler domain is checked for zero. If the velocity is zero, a stationary human body is identified; if the velocity is non-zero, interference is identified. For interference signals, subtract the interference signal from the distance-slow time dimension, and then check again whether the signal fluctuation is greater than the threshold. Repeat this process until all interference is subtracted.

[0020] In summary, the anti-interference human detection method based on the distance-Doppler domain proposed in this invention reduces computational complexity while eliminating interference, thereby improving the subsequent detection rate and reducing false alarms.

Claims

1. A method for detecting human presence using millimeter-wave radar based on the range-Doppler domain, characterized in that, Includes the following steps: Step 1: Acquire radar echoes and perform 1DFFT on them to obtain range-slow time dimension data; Step 2: Perform static clutter filtering on the distance-slow time dimension data. Specifically, subtract the distance-slow time dimension data from the previous time dimension from the distance-slow time dimension data acquired at the current time, and iterate through all distance-slow time dimension data to eliminate static clutter with constant energy. Step 3: Perform 2DFFT on the data after static clutter filtering to obtain range-Doppler data; Step 4: For the data after static clutter filtering, determine whether the signal energy fluctuation is greater than the threshold to identify potential targets. Specifically, select all slow time dimension data under each range gate, calculate the ratio of the maximum value to the mean value of the current range gate signal energy. If the ratio is greater than n, it is determined that there is a potential target at the current range gate, where n is an empirical value. Step 5: For the selected potential targets, determine the moving targets and candidate targets by the distance gate span judgment. Specifically: for a potential target, if the potential target appears in a continuous time and spans more than one continuous distance gate, that is, there is a continuous trajectory in the distance-slow time dimension, it is determined to be a moving target. If the potential target only appears in one distance gate, it is determined to be a candidate target. Step 6: For candidate targets, based on the distance-Doppler data obtained in Step 3, extract the Doppler information of the candidate targets, and combine the distance and Doppler information to determine stationary human bodies and interference. Specifically, for the distance gate of the candidate target, extract its corresponding Doppler information. If the Doppler velocity is 0, it is judged as a stationary human body. If the Doppler velocity is greater than 0, it is judged as interference. Interference categories include oscillating fans and green plants. Step 7: Mark the corresponding range gate for the detected interference, subtract the interference signal of that range gate from the range-slow time dimension data, and repeat this process until no signal energy meets the threshold, at which point the loop ends. Step 8: Output the detection results. Candidate targets with non-zero Doppler velocity are judged as interference, candidate targets with zero Doppler velocity are judged as stationary human bodies, moving targets are judged as moving human bodies, and targets whose signal energy does not meet the threshold are judged as human bodies not existing.

Images