Radar-based target tracking using motion detection

Abstract

Embodiments of the present disclosure relate to radar-based target tracking using motion detection. In an embodiment, a method includes receiving reflected radar signals with a millimeter wave radar, performing a range discrete Fourier transform (DFT) based on the reflected radar signals to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of range bins, determining, for each of the plurality of range bins, a respective intensity value based on changes in the respective I and Q signals over time, performing a peak search over the plurality of range bins to identify a peak range bin based on the respective intensity value for each of the plurality of range bins, and associating a target to the identified peak range bin.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to electronic systems and methods, and in particular embodiments, to radar-based human tracking using motion detection. BACKGROUND

[0002] Due to the rapid development of low-cost semiconductor technologies, such as silicon germanium (SiGe) and fine geometry complementary metal oxide semiconductor (CMOS) processes, applications in the millimeter-wave frequency regime have been of interest in the last few years. The availability of high-speed bipolar and metal oxide semiconductor (MOS) transistors has led to an increased demand for integrated circuits for millimeter-wave applications at, for example, 24 GHz, 60 GHz, 77 GHz, and 80 GHz, and beyond 100 GHz. Such applications include, for example, automotive radar systems and multi-gigabit communication systems.

[0003] In some radar systems, a distance between a radar and a target is determined by transmitting a frequency modulated signal, receiving a reflection (also referred to as a return) of the frequency modulated signal, and determining a distance based on a time delay and / or a frequency difference between the transmission and reception of the frequency modulated signal. Accordingly, some radar systems include a transmit antenna that transmits a radio frequency (RF) signal and a receive antenna that receives the reflected RF signal, and related RF circuitry for generating the transmitted signal and receiving the RF signal. In some cases, multiple antennas can be used to implement directional beams using phased array techniques. Multiple input multiple output (MIMO) configurations with multiple chipsets can also be used to perform coherent and non-coherent signal processing. SUMMARY

[0004] According to an embodiment, a method includes receiving, with a millimeter wave radar, a reflected radar signal; performing, based on the reflected radar signal, a distance discrete Fourier transform (DFT) to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of distance bins; determining, for each of the plurality of distance bins, a respective intensity value based on a change in the respective I signal and Q signal over time; performing, based on the respective intensity value for each of the plurality of distance bins, a peak search across the plurality of distance bins to identify a peak distance bin; and associating a target to the identified peak distance bin.

[0005] According to an embodiment, an apparatus includes: a millimeter-wave radar configured to transmit a chirped signal and receive a reflected chirped signal; and a processor configured to: perform a range discrete Fourier transform (DFT) based on the reflected chirped signal to generate an in-phase I signal and a quadrature Q signal for each of a plurality of range bins; determine a corresponding intensity value for each of the plurality of range bins based on the time-varying nature of the corresponding I and Q signals; perform a peak search across the plurality of range bins based on the corresponding intensity value for each of the plurality of range bins to identify a peak range bin; and associate a target with the identified peak range bin.

[0006] According to an embodiment, a method includes: receiving reflected radar signals using millimeter-wave radar; performing a range fast Fourier transform (FFT) based on the reflected radar signals to generate in-phase I signals and quadrature Q signals for each of a plurality of range bins; determining a corresponding short-term motion value for each of the plurality of range bins based on the changes in the corresponding I and Q signals in a single frame; performing a peak search across the plurality of range bins based on the corresponding short-term motion value for each of the plurality of range bins to identify a short-term peak range bin; and associating a target with the identified short-term peak range bin. Attached Figure Description

[0007] To gain a more complete understanding of the invention and its advantages, reference is now made to the following description in conjunction with the accompanying drawings, wherein:

[0008] Figure 1 A radar system according to an embodiment of the present invention is shown;

[0009] Figure 2 The invention illustrates an embodiment of the invention, by Figure 1 The sequence of radiation pulses emitted by the transmitting circuit;

[0010] Figure 3 A flowchart illustrating an embodiment of a method for detecting and tracking a human target according to an embodiment of the present invention is shown;

[0011] Figure 4 A state diagram for tracking a human target is shown according to an embodiment of the present invention;

[0012] Figures 5A-5D Showing the use of in Figure 4 An embodiment of the method for transitioning between states in a state diagram;

[0013] Figure 6 This illustrates an embodiment of the invention for use Figure 4 The flowchart of an implementation method for using a state machine to track the human body;

[0014] Figure 7The illustration shows an embodiment of the invention, by using Figure 6 A block diagram of the parameters tracked for each trajectory by the method;

[0015] Figures 8A-8D The use of, according to an embodiment of the present invention Figure 7 The parameters in Figure 4 The transitions between states in a state diagram;

[0016] Figure 9 A flowchart illustrating an embodiment of a method for generating distance data according to an embodiment of the present invention is shown;

[0017] Figure 10 and Figure 11 The illustrations, according to embodiments of the present invention, show people facing and moving away from each other, respectively. Figure 1 The millimeter-wave radar provides short-term and long-term moving maps;

[0018] Figure 12 Explanation shown Figure 10 and Figure 11 A map showing the maximum range of movement of the same group of people;

[0019] Figures 13-28 The present invention illustrates an embodiment of the present invention. Figure 10 and Figure 11 IQ curves for different frames of the map;

[0020] Figures 29-34 The present invention illustrates an embodiment of the present invention. Figures 10-12 The distance amplitude curves of FFT, STM and LTM for different frames of the map;

[0021] Figures 35-37 This illustrates an embodiment of the invention, when tracking such... Figure 10 and Figure 11 Pedestrians captured in the video Figure 6 The graph shows the output of the distance and velocity generation steps of the method;

[0022] Figure 38 Explanation shown Figures 35-37 The graph shown is a typical tracking curve of a pedestrian, which is generated by identifying the target based on the peak value of the distance FFT amplitude curve, and in which Doppler FFT is used to determine the target's velocity.

[0023] Figures 39-41 This illustrates an embodiment of the invention, showing how to track departure and orientation in a step-by-step manner. Figure 1 Millimeter-wave radar when pedestrians are walking Figure 6 The graph shows the output of the distance and velocity generation steps of the method;

[0024] Figure 42 This illustrates an embodiment of the invention, showing how to track pedestrians using a frame skipping mode. Figure 6 The graph shows the output of the distance and velocity generation steps of the method;

[0025] Figure 43 This illustrates an embodiment of the invention, showing how to track a walking person using a low-power mode. Figure 6 The graph shows the output of the distance and velocity generation steps of the method;

[0026] Figure 44 and Figure 45 This illustrates an embodiment of the invention, showing how to track a walking person using a low-power mode and a frame skipping mode. Figure 6 The method outputs curves for distance and velocity generation steps; and

[0027] Figures 46-48 This illustrates an embodiment of the invention, showing how to track objects moving away in a stepwise manner. Figure 1 Millimeter-wave radar of walking people Figure 6 The method outputs a graph of distance and velocity generation steps.

[0028] Unless otherwise stated, corresponding numbers and symbols in the various figures generally denote corresponding parts. The figures are drawn to clearly illustrate relevant aspects of the preferred embodiments and are not necessarily drawn to scale. Detailed Implementation

[0029] The manufacture and use of the disclosed embodiments are discussed in detail below. However, it should be understood that the present invention provides many applicable inventive concepts that can be implemented in various specific contexts. The specific embodiments discussed are merely illustrative of specific ways of constructing and using the invention and do not limit the scope of the invention.

[0030] The following description illustrates various specific details to provide a thorough understanding of several exemplary embodiments according to the specification. Embodiments may be obtained without one or more specific details, or by utilizing other methods, components, materials, etc. In other instances, known structures, materials, or operations have not been shown or described in detail so as not to obscure different aspects of the embodiments. References to "embodiment" in this specification indicate that a particular configuration, structure, or feature described with respect to that embodiment is included in at least one embodiment. Therefore, phrases such as "in one embodiment" that may appear at different points in this specification do not necessarily refer precisely to the same embodiment. Furthermore, in one or more embodiments, particular forms, structures, or features may be combined in any suitable manner.

[0031] Embodiments of the present application will be described in the context of a radar-based human tracking system and method using motion detection. Embodiments of the present application can be used to track other types of targets, such as animals (e.g., dogs) or autonomous objects, such as robots.

[0032] In embodiments of the present application, millimeter wave radar performs target detection by a moving investigation of each range bin, instead of performing a conventional range FFT peak search. The moving investigation is performed by a micro-Doppler evaluation in the in-phase (I) and quadrature (Q) plane, instead of using a conventional Doppler Fast Fourier Transform (FFT). In some embodiments, the millimeter wave radar uses multiple states to track one or more targets.

[0033] Radar, such as millimeter wave radar, can be used to detect and track humans. Conventional frequency-modulated continuous wave (FMCW) radar systems sequentially transmit a linearly increasing frequency waveform, known as a chirp, which is collected by a receiving antenna after being reflected by an object. The radar can operate as a monostatic radar, in which a single antenna works as both a transmitting and receiving antenna, or a bistatic radar, in which dedicated antennas are used for transmitting and receiving radar signals, respectively.

[0034] The transmitted and received signals are then mixed with each other in the RF section, generating an intermediate frequency (IF) signal that is digitized using an analog-to-digital converter (ADC).

[0035] The IF signal is known as the beat signal and contains the beat frequencies of all targets. After bandpass filtering the IF signal, a Fast Fourier Transform (FFT) is applied to the digitized and filtered IF signal to extract the range information of all targets from the radar data. This process is known as range FFT and generates range data.

[0036] The first dimension of the range data includes all samples per chirp (fast time) for range estimation. The second dimension of the range data includes data of the same range bin from different chirps in a frame (slow time) for velocity estimation.

[0037] Conventionally, targets are detected based on a peak search over the fast time dimension of the range data, where a target is detected when the amplitude of a range bin is above a threshold. The target velocity is estimated by using a so-called Doppler FFT along the slow time dimension of the corresponding range bin (where the target was detected).

[0038] In embodiments of the invention, the millimeter wave radar performs target detection through a moving investigation of each range bin. In some embodiments, the moving investigation includes determining a short term motion (STM) value and a long term motion (LTM) value for each range bin. The short term motion value is determined for each range bin based on the I and Q signals of a single frame. The long term motion value is determined for each range bin based on the I and Q signals over multiple frames. A peak search is performed to identify short term motion peaks above a predetermined STM threshold and long term motion peaks above a predetermined LTM threshold. One or more targets are identified based on the STM peaks and the LTM peaks.

[0039] Figure 1 A radar system 100 according to embodiments of the invention is shown. The radar system 100 includes a millimeter wave radar 102 and a processor 104. In some embodiments, the millimeter wave radar 102 includes the processor 104.

[0040] During normal operation, the millimeter wave radar 102 transmits a plurality of radiation pulses 106, such as a chirp signal, toward a scene 108 using a transmitter (TX) circuit 120. In some embodiments, the chirp signal is a linear chirp signal (i.e., the instantaneous frequency of the chirp signal varies linearly with time).

[0041] The transmitted radiation pulses 106 are reflected by objects in the scene 108. The reflected radiation pulses (not shown), also referred to as echo signals, are received by the millimeter wave radar 102 using a receiver (RX) circuit 122 and processed by the processor 104 to, for example, detect and track targets, such as human bodies. Figure 1

[0042] The objects in the scene 108 can include stationary human bodies, such as a lying human body 110, human bodies exhibiting low and infrequent motion, such as a standing human body 112, and moving human bodies, such as a walking human 114 and a running human 116. The objects in the scene 108 can also include stationary objects (not shown), such as furniture, walls, and periodic motion devices. Other objects can also be present in the scene 108.

[0043] ​The processor 104 uses signal processing techniques to analyze the echo data to determine the location of the human body. For example, in some embodiments, the processor 104 performs target detection by a moving investigation of each range bin of the range data. In some embodiments, the processor 104 performs a moving investigation for a particular range bin by a micro-Doppler evaluation in the IQ plane of the particular range bin. In some embodiments, the processor 104 tracks a detected target(s), for example, using multiple states. In some embodiments, a tracking algorithm such as an alpha-beta filter can be used to track the target(s). In some embodiments, other tracking algorithms can be used, for example, an algorithm using a Kalman filter.

[0044] The processor 104 can be implemented as a general purpose processor, a controller, or a digital signal processor (DSP), including a combination of circuitry coupled to memory, for example. In some embodiments, the processor 104 can be implemented in an ARM architecture, for example. In some embodiments, the processor 104 can be implemented as a custom application specific integrated circuit (ASIC). Some embodiments can be implemented as a combination of a hardware accelerator and software running on a DSP or general purpose microcontroller. Other implementations are possible.

[0045] The millimeter wave radar 102 operates as an FMCW radar, including millimeter wave radar sensor circuitry and one or more antennas. The millimeter wave radar 102 transmits (using TX 120) and receives (using RX 122) signals via the one or more antennas (not shown) in the range of 20 GHz to 122 GHz. For example, in some embodiments, the millimeter wave radar 102 has a 200 MHz bandwidth when operating in the frequency range of 24.025 GHz to 24.225 GHz. Some embodiments can use frequencies outside of this range, for example, frequencies between 1 GHz and 20 GHz, or frequencies between 122 GHz and 300 GHz.

[0046] In some embodiments, the echo signals received by the millimeter wave radar 102 are processed in the analog domain using band pass filters (BPFs), low pass filters (LPFs), mixers, low noise amplifiers (LNAs), and intermediate frequency (IF) amplifiers in a manner known in the art. The echo signals are then digitized for further processing using one or more ADCs. Other implementations are possible.

[0047] Figure 2 A sequence of radiation pulses 106 transmitted by the TX circuitry 120 is shown in accordance with an embodiment of the application. As shown, the radiation pulses 106 are organized in a plurality of frames, and can be implemented as up-chirp signals. Some embodiments can use down-chirp signals or a combination of up-chirp signals and down-chirp signals. Figure 2 As shown, the radiation pulses 106 are organized in a plurality of frames, and can be implemented as up-chirp signals. Some embodiments can use down-chirp signals or a combination of up-chirp signals and down-chirp signals.

[0048] The time between the chirp signals of a frame is often referred to as the pulse repetition time (PRT). In some embodiments, the PRT is 5 ms. Different PRTs can also be used, for example less than 5 ms, such as 4 ms, 2 ms or less, or more than 5 ms, such as 6 ms or more.

[0049] A frame of chirp signals 106 comprises a plurality of chirp signals. For example, in some embodiments, each frame of chirp signals contains 16 chirp signals. Some embodiments can comprise more than 16 chirp signals per frame, such as 20 chirp signals, 32 chirp signals, or more, or less than 16 chirp signals per frame, such as 10 chirp signals, 8 chirp signals, or less. In some embodiments, each frame of chirp signals comprises a single chirp signal.

[0050] The frames repeat every FT time. In some embodiments, the FT time is 50 ms. Different FT times can also be used, for example more than 50 ms, such as 60 ms, 100 ms, 200 ms or more, or less than 50 ms, such as 45 ms, 40 ms or less.

[0051] In some embodiments, the FT time is chosen such that the time between the start of the last chirp signal of frame n and the start of the first chirp signal of frame n+1 is equal to the PRT. Other embodiments can use or result in different timing.

[0052] Figure 3 A flowchart of an embodiment method 300 for detecting and tracking a human target according to embodiments of the application is shown. The method 300 can be performed, for example, by the processor 104.

[0053] During step 302, the millimeter wave radar 102 transmits linear chirp signals, for example organized in frames, using the TX circuit 120 (as shown in Figure 2 The frequency of a transmitted chirp signal, for example having a bandwidth B and a duration T, can be represented as

[0054]

[0055] where f c is the ramp start frequency.

[0056] After reflection from the object, the RX circuit 122 receives the reflected chirp signals during step 304.

[0057] During step 306, the reflected chirp signals received during step 304 are processed in the analog domain in a conventional manner to generate an IF signal. For example, the reflected chirp signals are mixed with a copy of the transmitted signal, thereby generating a beat signal.

[0058] During step 308, the IF signal is converted to the digital domain (using an ADC) to generate raw data for further processing.

[0059] During step 310, a range discrete Fourier transform (DFT), such as a range FFT, is performed on the raw data to generate range data. For example, in some embodiments, the raw data is zero-padded and a fast Fourier transform (FFT) is applied to generate range data that includes range information for all targets. In some embodiments, the maximum unambiguous range of the range FFT is based on the PRT, the number of samples per chirp signal, the chirp signal time, and the sampling rate of an analog-to-digital converter (ADC). In some embodiments, the ADC has 12 bits. An ADC with a different resolution (e.g., 10 bits, 14 bits, or 16 bits) can also be used.

[0060] In some embodiments, the range FFT is applied to all samples of the chirp signal.

[0061] During step 312, target detection is performed by a moving investigation of each range bin. Target detection is based on short-term motion (STM) detection and / or long-term motion (LTM) detection. Thus, step 312 includes step 314 and / or step 316. Step 314 includes steps 314a and 314b. Step 316 includes steps 316a and 316b.

[0062] During step 314a, the STM motion for each range bin R r is determined by the following equation:

[0063]

[0064] where R represents the complex output value of the range FFT, R r is the complex value at a particular range bin r, M STM,r represents the short-term motion of the current frame for range bin r, PN represents the number of chirp signals per frame, and c represents the chirp signal index such that R r,c is the complex (with I and Q components) associated to the range bin R r for chirp signal c, and R r,c+1 is the complex (with I and Q components) associated to the range bin R r for chirp signal c+1. In some embodiments, PN can be a value equal to or higher than 2, such as 8 or 16.

[0065] Equation 2 can also be understood as a summation of all edges of the I-Q plot generated using the chirp signals of the current frame (e.g., as Figure 13 , 15, 17, 19, 21, 23, 25, 27). In this I-Q plot, the edges are straight lines connecting the nodes, where each of the nodes represents the (I, Q) components of a particular chirp signal c of the current frame. The value M STM,r may also be referred to as an intensity value and indicates short-term movement (the higher the value, the more movement detected in the distance bin r in fast time for the current frame).

[0066] During step 314b, a peak search is performed on all short-term movement values M STM,r and (local) peaks above a predetermined STM threshold T M,STM are identified. Since the intensity of the peaks identified during step 314b relates to short-term movement, stationary objects are typically associated to intensity values below the predetermined STM threshold T M,STM (e.g. as shown in Figures 29-34 ).

[0067] During step 316a, the LTM movement is determined along the first chirp signal of the last W frames by the following equation

[0068]

[0069] where M LTM,r represents the short-term movement of the current frame of distance bin r, w represents the frame index such that R r,1,w is the complex number (with I and Q components) of distance bin R r associated to the first chirp signal of frame w, and R r,1,w+1 is the complex number (with I and Q components) of distance bin R r associated to the first chirp signal of frame w+1. Some embodiments can use a different index of chirp signal than the first chirp signal to compute M LTM,r . In some embodiments, W can be a value equal to or higher than 2, e.g. 10 or 20.

[0070] Equation 3 can also be understood as the addition of all edges of an I-Q plot (e.g. shown in Figure 14 , 16 , 18, 20, 22, 24, 26, 28) generated using a single chirp signal (e.g. the first chirp signal) from each of the last W frames. In this I-Q plot, the edges are straight lines connecting the nodes, where each of the nodes represents the (I, Q) components of a single chirp signal c of a particular frame w. The value M LTM,r may also be referred to as an intensity value and indicates long-term movement (the higher the value, the more movement detected in the distance bin r in slow time over the last W frames).

[0071] During step 316b, a peak search is performed on all long-term movement values MLTM,r The peak search is performed above and local peaks above a predetermined LTM threshold T M,LTM are identified. Since the intensity of the peaks identified during step 316b is related to long term motion, in some cases (e.g. due to shadowing effects) a stationary object can be associated to an intensity value above the predetermined LTM threshold T M,LTM (as shown in Fig. 3B). Figures 29-34

[0072] During step 314b and / or 316b, an ordered statistics (OS) constant false alarm rate (CFAR) (OS-CFAR) detector can be used to identify local peaks (peaks above a predetermined STM threshold T M,STM or LTM threshold T M,LTM . Other search algorithms can also be used.

[0073] In some embodiments, T M,LTM is different than T M,STM . In other embodiments, T M,LTM is equal to T M,STM . As a non-limiting example, in an embodiment, T M,LTM is equal to 50 and T M,STM is equal to 200.

[0074] The peaks identified during step 314a and / or 316b represent potential or actual targets. During step 318, some or all of the potential or actual targets are tracked.

[0075] In some embodiments, a state machine (e.g. implemented in processor 104) can be used to track the targets (e.g. during step 318). For example, Figure 4 Fig. 4 shows a state diagram 400 for tracking a human target according to an embodiment of the present application. In some embodiments, the target state is evaluated at each frame n.

[0076] State diagram 400 includes a stationary state 402, an uncertain state 404, a moving state 406 and a stationary state 408. Stationary state 402 is associated to a human target that is not tracked (e.g. because the corresponding trajectory has been terminated or has not been created yet). Uncertain state 404 is associated to a potential human target. Moving state 406 is associated to an actual human target that is moving. Stationary target 408 is associated to an actual human target that is stationary.

[0077] ​As will be described in more detail later, in some embodiments, when the target is first transitioned from the uncertain state 404 to the moving state 406, the target is activated (and thus transitioned from a potential target to an actual target). Thus, in some embodiments, a target cannot transition from the dormant state 402 to the uncertain state 404 without first being activated, and then directly to the stationary state 408. As will be described in more detail later, since a target is in the moving state 406 before being in the stationary state 408, some embodiments advantageously prevent actively tracking stationary targets (e.g., such as a wall) that can sometimes appear to move (e.g., due to a shadow effect).

[0078] As Figure 4 shown, in some embodiments, a target cannot directly transition from the moving state 406 to the dormant state 402, thus advantageously allowing tracking of actual targets that can temporarily disappear (e.g., during step 312, the target becomes undetected) due to noise or due to the target stopping moving.

[0079] Figures 5A-5D Embodiment methods 500, 520, 550, and 570 for transitioning between the states of the state diagram 400 are shown. The methods can be understood in accordance with Figures 5A-5D . Figure 4 .

[0080] As Figures 5A-5D shown, the method 500 shows a flowchart for transitioning from the dormant state 402; the method 520 shows a flowchart for transitioning from the uncertain state 404; the method 550 shows a flowchart for transitioning from the moving state 406; and the method 570 shows a flowchart for transitioning from the stationary state 408.

[0081] As Figure 5A shown, when the target is in the dormant state 402 and the target is detected during step 504 (e.g., which corresponds to step 312), if it is determined during step 506 that the peak associated to the detected target is an STM peak (e.g., determined during step 314b), a track is created and the target is transitioned from the dormant state 402 to the uncertain state 404 (step 510).

[0082] As will be described in more detail later, LTM peaks and STM peaks that are close to each other can be associated to the same target. Thus, if it is determined during step 506 that the peak is an LTM peak (e.g., determined during step 316b), a track is created, and if it is determined during step 508 that the LTM peak is not associated to any STM peak, the target is transitioned from the dormant state 402 to the uncertain state 404 (step 510).

[0083] AsFigure 5B As shown, when the target is in an uncertain state 404, during step 524, it is determined whether a target peak detected in the current frame is associated with the trajectory. If no peak is associated with the trajectory, it is determined during step 526 whether the timer has expired. In some embodiments, the timer counts the time (e.g., frame number) during which the tracked target's trajectory has not been associated with any peak (or any STM peak) of the trajectory.

[0084] If it is determined during step 526 that the timer has expired, the trajectory is terminated during step 528. Otherwise, the target remains in an uncertain state 404 during step 530. By waiting before terminating the trajectory (e.g., by using a timer), some embodiments advantageously allow the trajectory to remain valid temporarily, and thus allow the tracking of a real target that may temporarily disappear, for example, due to noise or because the target stops moving (e.g., the target becomes undetectable during step 312).

[0085] If a peak associated with the trajectory is determined during step 524, the type of the peak is determined during step 532. If the peak associated with the trajectory is an STM peak, the trajectory is activated during step 538 (thus switching from tracking a potential target to tracking an actual target), and the state transitions from an uncertain state 404 to a moving state 406 during step 540. In some embodiments, the trajectory may be activated only after multiple frames of STM peaks associated with it have been presented.

[0086] If the peak associated with the trajectory is an LTM peak, then if it is determined during step 534 that the trajectory has been activated, the state transitions from uncertain state 404 to static state 408 during step 536. Otherwise, step 526 is executed. In some embodiments, the trajectory may transition from uncertain state 404 to static state 408 only after multiple frames of the trajectory have been in uncertain state 404.

[0087] like Figure 5C As shown, when the target is in motion state 406, during step 554, it is determined whether the target peak detected in the current frame is associated with the trajectory. If no peak is associated with the trajectory, during step 556, the target transitions from motion state 406 to uncertain state 404.

[0088] If a peak is determined to be associated with a trajectory during step 554, the type of the peak is determined during step 558. If the peak associated with the trajectory is determined to be an STM peak during step 558, the target remains in moving state 408 during step 560. Otherwise, if the peak associated with the trajectory is an LTM peak, the target transitions from moving state 406 to uncertain state 404 during step 556.

[0089] like Figure 5D As shown, when the target is in a stationary state 408, during step 574, it is determined whether a target peak detected in the current frame is associated with the trajectory. If no peak is associated with the trajectory, during step 576, it is determined whether a timer has expired (e.g., in a manner similar to that in step 526). If it is determined during step 576 that the timer has expired, the trajectory is terminated during step 578. Otherwise, during step 580, the target remains in a stationary state 404.

[0090] If a peak is determined to be associated with a trajectory during step 574, the type of the peak is determined during step 582. If the peak associated with the trajectory is determined to be an STM peak during step 582, the target transitions from a stationary state 408 to an uncertain state 404 during step 584. Otherwise, if the peak associated with the trajectory is an LTM peak, the target remains in a stationary state 408 during step 580.

[0091] Figure 6 A flowchart of an embodiment of a method 600 for tracking a human body using a state machine 400, according to an embodiment of the present invention, is shown. Step 318 can be implemented as method 600.

[0092] like Figure 6 As shown, method 600 includes steps 602 for updating all valid trajectories, 612 for terminating invalid trajectories, 614 for assigning new trajectories, and 620 for generating estimated distance and velocity for each tracked target. Step 602 is performed for each valid (not terminated) trajectory and includes steps 604, 606, 608, and 610. Step 614 includes steps 616 and 618.

[0093] During step 604, the distance to the target tracked by the trajectory is predicted, for example, using the following formula.

[0094] R pred =R w-1 -FT·S w-1 (4)

[0095] Among them, R pred Let R be the prediction distance of the current frame, FT be the frame time, and w be the frame index, such that R w-1 S represents the distance to the target in the previous (last) frame (e.g., 704). w-1 This indicates the velocity of the target in the previous (last) frame (e.g., 708).

[0096] During step 606, appropriate STM peaks are associated with valid trajectories. For example, in some embodiments, when the distance R associated with the STM peak is...STM (For example, the distance R identified in step 314b) is greater than the predetermined STM distance. STM_th The predicted distance R closer to the trajectory pred When (i.e., if R) STM With R pred The deviation between them is lower than R STM_th If the STM peak is then correlated with the trajectory, in some embodiments, the prediction distance R closest to the trajectory is... pred The STM peak value is associated with this trajectory.

[0097] During step 608, appropriate LTM peaks are associated with valid trajectories. For example, in some embodiments, when the distance R associated with the LTM peak is... LTM (For example, the distance R identified in step 316b) is greater than the predetermined LTM distance. LTM_th The predicted distance R closer to the trajectory pred When (i.e., if R) LTM With R pred The deviation between them is lower than R LTM_th If the STM peak is then associated with the trajectory, in some embodiments, the prediction distance R closest to the trajectory is... pred The LTM peak value is associated with this trajectory.

[0098] In some embodiments, relative to R associated with the trajectory STM Instead of relative to R pred To measure the deviation. In some embodiments, the threshold R... STM_th Equal to threshold R LTM_th In other embodiments, the threshold R STM_th With threshold R LTM_th different.

[0099] During step 610, the trajectory status is updated based on the associated STM peak and LTM peak. For example, if an STM peak is associated with the trajectory, steps 506, 532, 558, and 582 output "STM", regardless of whether an LTM peak is associated with the trajectory. If an LTM peak is associated with the trajectory, but no STM peak is associated with it, steps 506, 532, 558, and 582 output "LTM". If the trajectory has no associated peaks, steps 524, 554, and 574 output "No".

[0100] During step 612, failed tracks are terminated. For example, during step 612, for each valid track, steps 528 and 578 are performed, if applicable.

[0101] During step 616, a new track is created for each STM peak that is not associated to any track (e.g., during step 510). Similarly, during step 618, a new track is created for each LTM peak that is not associated to any track (e.g., during step 510). In some embodiments, when an STM peak is assigned to a new track during step 616, the corresponding LTM peak (e.g., the LTM peak closest to the R STM of the STM peak) is also assigned to the same new track during step 616. After all STM peaks and corresponding LTM peaks are assigned to respective tracks, any remaining unassociated LTM peaks are assigned to new tracks during step 618.

[0102] During step 620, for each valid track, an estimated distance and velocity for the current frame is generated. For example, in some embodiments, the distance R w for the current frame can be calculated by the following equation

[0103] R w = β · R est +(1 - β) · R w-1 (5)

[0104] where

[0105] R est = α · R meas +(1 - α) · R pred (6)

[0106] where α and β are factors that can be predetermined, where R pred is calculated using equation 4, and where R meas is determined using equation 7 if there is an STM peak associated to the target (step 606), R meas is determined using equation 8 if there is no STM peak associated to the target but there is an LTM peak associated to the target (step 608), and R meas is determined using equation 9 if the target has no STM peak or LTM peak associated with it.

[0107] R meas = R STM (7)

[0108] R meas = R LTM (8)

[0109] R meas = R pred (9)

[0110] In some embodiments, the current frame S wThe speed of the target can be calculated using the following formula

[0111]

[0112] Where FT is the frame time, and SL represents the number of frames used for rate determination. In some embodiments, SL is 10. Other values ​​of SL may also be used, such as those below 10 (e.g., 9, 8 or lower), or those above 10, such as 11, 12 or higher.

[0113] As shown in Equation 10, the derivative of the distance is used instead of the Doppler FFT to determine the velocity of the tracked target.

[0114] In some embodiments, the actual distance and velocity generated during step 620 are filtered versions of the distance and velocity calculated using Equations 5 and 10. For example, in some embodiments, a median filter is used on the last I frames to determine the actual distance and velocity generated during step 620, where I is greater than 1, for example, 3 or 10. In some embodiments, I is equal to SL.

[0115] Figure 7 A block diagram 700 illustrates the parameters tracked by each trajectory using method 600 according to an embodiment of the present invention.

[0116] like Figure 7 As shown, each valid trajectory (e.g., one created during step 510 and not terminated) has a trajectory identification code 702. Each valid trajectory tracks either a potential target or an actual target, and this trajectory state is tracked by parameter 714. Trajectories tracking actual targets are called active trajectories (A=1), and trajectories tracking potential targets are called inactive trajectories (A=0).

[0117] like Figure 7 As shown, each valid trajectory tracks the last distance (704) and velocity (708) of the tracked target (e.g., determined during step 620). Each trajectory also has a distance history (706) of the tracked target that can be used in Equation 10, and a velocity history (710) of the tracked target. In some embodiments, the distance history 706 and / or the velocity history 710 may also be used during step 620 to generate filtered versions of the distance and velocity.

[0118] Each track also tracks the current state (712) of the tracked target, which is one of states 402, 404, 406, and 408. Each track also has a counter (716), which may be used to implement a timer (e.g., as used in steps 526 and 576). Each track also has an alpha factor (718), for example, as used in Equation 6.

[0119] Figures 8A-8D The transitions between states of state diagram 400 using parameter 700 are shown according to an embodiment of the present invention. Figure 8A The transition from stagnant state 402 according to an embodiment is shown, and possible implementations of method 500 are also shown. Figure 8B The transition from the uncertain state 404 according to an embodiment is shown, and possible implementations of method 520 are also shown. Figure 8C The transition from moving state 406 according to an embodiment is shown, and a possible implementation of method 550 is also shown. Figure 8D The transition from stagnant state 408 according to an embodiment is shown, and a possible implementation of method 570 is also shown.

[0120] like Figure 8A As shown, when an STM peak is associated with the target (STM == 1), for example, as shown in step 506, the target transitions from a stagnant state 402 to an uncertain state 404 (step 510), regardless of whether there is an LTM peak associated with the target (LTM == X). During this transition, the alpha factor α (718) is set to 1, and the counter (716) is set to 1 (C = 1). Figure 8A As shown, the trajectory is not activated (A=0).

[0121] For example Figure 8A As shown, when the LTM peak is associated with a target that does not have an associated STM peak (LTM == 1), for example, as shown in step 506, the target transitions from a stagnant state 402 to an uncertain state 404 (step 510). During this transition, the alpha factor α (718) is set to 0.5, and the counter (716) is set to 2 (C = 2). Figure 8A As shown, the trajectory is not activated (A=0).

[0122] like Figure 8B As shown, when the counter is greater than 0 (C > 0) and no STM peak is associated with the target (STM == 0), and no LTM peak is associated with the target (LTM = 0) or the trajectory is not activated (A == 0), the counter is decremented (C = C - 1). When the counter reaches 0 (output "Yes" from step 526) and there is no LTM peak associated with the trajectory (LTM == 0), the trajectory is terminated (step 528). If when the counter reaches 0 (C == 0), there is an LTM peak associated with the trajectory (LTM == 1) (output "LTM" in step 532), and if the trajectory has been activated (A == 1), the counter is set to count T. SC (C=T) SC), alpha factor a is set to 0.5 (a = 0.5), and the target transitions from the uncertain state 404 to the stationary state 408 (step 536).

[0123] If there is an STM peak associated to the trajectory (STM == 1), alpha factor a is set to 1 (a = 1) and the counter is incremented (C = C + 1) until the counter reaches a predetermined count T SC When the counter reaches the count T SC , the trajectory is activated (A = 1), alpha factor a is set to 0.8 (a = 0.8), and the target transitions from the uncertain state 404 to the moving state 406, as shown in step 538.

[0124] In some embodiments, the count T SC is equal to 5. Different values can also be used for the count T SC , such as 6, 7 or higher, or 4, 3 or lower.

[0125] As shown in step 542, the target remains a potential target for at least T SC frames before becoming the actual target (A = 1). In some embodiments, waiting for a number of frames (e.g., 3 frames) before activating the trajectory advantageously allows terminating trajectories associated to non-human targets, such as ghost targets. Figure 8B As shown in step 558, when the target is in the moving state 406, the target will remain in the moving state with the associated STM peak (output of step 558 equal to "STM"). When the target no longer has an associated STM peak (output of step 558 equal to "LTM" or output of step 554 equal to "No"), the counter is decremented (C = C - 1) and alpha factor a is set to 0.2, and the target transitions from the moving state 406 to the uncertain state 404 (step 556). As shown in step 560, when the target transitions between the moving state 406 and the uncertain state 404, the counter is set to T SC - 1 (because the target enters the moving state 406 where the counter is set to T SC , and the counter value does not change while the target is in the moving state 406). Figure 8C As shown in step 562, when the target is in the stationary state 408, the target will remain in the stationary state with the associated STM peak (output of step 562 equal to "STM"). When the target no longer has an associated STM peak (output of step 562 equal to "LTM" or output of step 554 equal to "No"), the counter is decremented (C = C - 1) and alpha factor a is set to 0.2, and the target transitions from the stationary state 408 to the uncertain state 404 (step 556). As shown in step 564, when the target transitions between the stationary state 408 and the uncertain state 404, the counter is set to T SC - 1 (because the target enters the stationary state 408 where the counter is set to T SC , and the counter value does not change while the target is in the stationary state 408).

[0127] Figure 8D ​As shown, when no STM peak is associated to the target (STM == 0), the target remains in the stationary state 408. However, when no LTM peak is also associated to the target (LTM == 0), the counter is decremented (C = C - 1). When the counter reaches 0 (C == 0; output of step 576 equals "Yes"), the trajectory is terminated (step 578). Since the target entered the stationary state 408 with the counter equal to T SC , the trajectory is not terminated for at least T SC frames. In some embodiments, avoiding termination of the trajectory over a number of frames (e.g., 3) advantageously allows temporarily keeping the trajectory valid, and thus allows keeping track of an actual target that can temporarily stop moving.

[0128] As shown, when an STM peak is associated to the target (output of step 582 equals "STM"), the counter is decremented (C = C - 1), the alpha factor a is set to 0.8 (a = 0.8), and the target is transitioned from the stationary state 408 to the uncertain state 404 (step 584). Figure 8D

[0129] A flowchart of an embodiment method 900 for generating range data according to an embodiment of the application is shown. Step 310 can be implemented as method 900. Figure 9 During step 902, data is calibrated. In some embodiments, the calibration data are stored raw data having the size of one chirp signal. These data can be generated by recording only one chirp signal or fusing several chirp signals of one frame, etc. During step 902, these calibration data are subtracted from the acquired raw data.

[0130] During step 904, DC offset is removed by a DC offset compensation step (also called mean removal). In some embodiments, the DC offset compensation advantageously allows removing the DC offset caused by RF non-idealities.

[0131] During step 906, a windowing operation (e.g., using a Blackman window) is performed to, for example, increase the signal-to-noise ratio (SNR).

[0132] During step 908, zero padding is performed to, for example, improve the accuracy of the range FFT output, and thus the accuracy of the range estimate. In some embodiments, a factor of 4 is used for the zero padding operation.

[0133] During step 910, a range FFT is performed by applying an FFT to this zero-padded data to generate the range data. The range FFT is applied to all samples of the chirp signal. Other implementations are also possible.

[0134]

[0135] ​It should be appreciated that some of the steps disclosed, e.g., steps 902, 904, 906, and / or 908, can be optional and can not be implemented.

[0136] Figures 10-48 Experimental results according to embodiments of the application are shown. Unless otherwise noted, the measurement data associated with Figures 10-48 were obtained with a frame time FT of 50 ms.

[0137] Figure 10 A map chart 1000 according to embodiments of the application is shown, showing the short-term movement of each distance bin per frame of the person 114 walking towards and away from the mmWave radar 102. Figure 10 The short-term movement M STM,r is determined using Equation 2.

[0138] Figure 11 A map chart 1100 according to embodiments of the application is shown, showing the long-term movement of each distance bin per frame of the same person 114 walking towards and away from the mmWave radar 102. Figure 11 The long-term movement M LTM,r is determined using Equation 3.

[0139] As shown in Figure 10 and Figure 11 the long-term movement of the person 114 is delayed relative to the short-term movement of the person 114, e.g., as shown by the LTM and STM distances of distance bins 79, 96, and 115. Figure 10 and Figure 11 A stationary object, e.g., a wall at a distance of approximately 10.5 m from the mmWave radar 102, is also shown, captured by LTM Figure 11 but not by STM Figure 10 .

[0140] As shown in Figure 10 when the human target 114 turns around and thus partially stationary, the human target 114 temporarily disappears from the STM distance at frame 253. However, the human target 114 is captured by the LTM during frame 253.

[0141] Figure 12 A map chart 1200 is shown, showing the amplitude of the maximum distance of the same human 114 walking towards and away from the mmWave radar 102. The map chart 1200 can be generated based on the distance data generated during step 310 or step 910. Figures 10-12 is generated based on the same raw data generated during step 308.

[0142] As shown in Figure 12As shown, the distance data from map chart 1200 includes information about the movement of human target 114 as well as information about stationary objects such as a wall.

[0143] Figures 13-28 I-Q plots for different frames of map charts 1000 and 1100 are shown in accordance with embodiments of the application. Figures 29-34 Amplitude plots of distance FFT, STM, and LTM for different frames of map charts 1000, 1100, and 1200 are shown in accordance with embodiments of the application. Figures 13-34 may be taken together and with reference to Figures 10-12 are understood.

[0144] Figures 13-16 Frame 79 of map charts 1000, 1100, and 1200 corresponds to frames 28 and 29. At frame 79, human target 114 is walking towards millimeter wave radar 102 and is at a distance of about 7 m from millimeter wave radar 102. The wall is at a distance of about 10.5 m from millimeter wave radar 102.

[0145] As Figure 29 shown, the distance FFT amplitude plot (which corresponds to Figure 12 ) includes a peak 2902 corresponding to RF leakage, a peak 2904 corresponding to human target 114, and a peak 2906 corresponding to the wall. The STM amplitude plot (which corresponds to Figure 10 and is calculated using Equation 2) includes a peak 2924 corresponding to human target 114. The LTM amplitude plot (which corresponds to Figure 11 and is calculated using Equation 3) includes a peak 2944 corresponding to human target 114 and a peak 2946 corresponding to the wall.

[0146] As Figure 29 shown, the leakage peak is not present in the STM or LTM plots, as indicated by locations 2922 and 2942.

[0147] As Figure 29 also shown, the wall peak present in the distance FFT plot (peak 2906) and the LTM plot (peak 2946) is not present in the STM plot, as indicated by location 2926. For example, Figure 15 An I-Q plot for the STM at frame 79 at location 2926 is shown. Figure 16 An I-Q plot for the LTM at frame 79 at peak 2946 is shown. As Figure 15 and Figure 16 shown, the amount of movement exhibited by the wall is much smaller in the STM plot than in the LTM plot. This smaller amount of movement results in the STM strength M STM,r being lower than the predetermined STM threshold T M,STMAnd therefore not identified as a peak. The amount of movement exhibited by the walls in the LTM curve results in the LTM intensity M. LTM,r Above the predetermined LTM threshold T M,LTM And thus it was identified as a peak of 2946.

[0148] like Figure 13 and Figure 14 As shown, the amount of movement exhibited by the walking person 114 is sufficient to cause the STM intensity M STM,r Above the predetermined STM threshold T M,STM And LTM strength M LTM,r Above the predetermined LTM threshold T M,LTM Therefore, positions 2924 and 2944 were identified as peaks.

[0149] Figure 29 The delay between peak 2944 and peak 2924 is also shown. This delay is the result of calculating the LTM curve using information from the previous W frames (as shown in Equation 3) and calculating the STM curve using information from the current frame w (as shown in Equation 2).

[0150] As from Figure 29 It can be seen that the output of step 314b is a peak value of 2924, and the output of step 316b is a peak value of 2944 and 2946. From... Figure 29 As can be seen from the distance FFT amplitude curve, conventional target detection methods that rely on the amplitude peaks of the distance FFT will detect, for example, three targets associated with peaks 2902, 2904 and 2906 (or, for example, only two targets associated with peaks 1902 and 1906 and miss peak 2904 corresponding to the walking person 114).

[0151] Figures 17-20 30 corresponds to frame 96 of map charts 1000, 1100, and 1200. At frame 96, human target 114 is approaching millimeter-wave radar 102 and is approximately 5m away from millimeter-wave radar 102.

[0152] like Figure 30 As shown, the distance FFT amplitude curve (which corresponds to) Figure 12 This includes peak value 3002 corresponding to RF leakage, peak value 3004 corresponding to human target 114, and peak value 3006 corresponding to the wall. STM amplitude curve (corresponding to...) Figure 10 (Calculated using Formula 2) This includes a peak value of 3024 corresponding to the human target 114. LTM amplitude curve (corresponding to...) Figure 11 (and calculated using Formula 3) This includes the peak value 3044 corresponding to the human target 114 and the peak value 3046 corresponding to the wall.

[0153] As Figure 30 shown, there is no leakage peak in the STM or LTM plots in locations 3022 and 3042 (and similarly to Figure 29 ).

[0154] As Figure 30 also shown, the wall peak present in the range FFT plot (peak 3006) and the LTM plot (peak 3046) is not in the STM plot, as shown in location 3026. For example, Figure 19 shows the I-Q plot for the STM at frame 96 for location 3026. Figure 20 shows the I-Q plot for the LTM at frame 96 for peak 3046. As Figure 19 and Figure 20 shown, the amount of movement exhibited by the wall is much smaller in the STM plot than in the LTM plot. This smaller amount of movement results in an STM intensity M STM,r that is below a predetermined STM threshold T M,STM , and is therefore not identified as a peak. The amount of movement exhibited by the wall in the LTM plot results in an LTM intensity M LTM,r that is above a predetermined LTM threshold T M,LTM , and is therefore identified as a peak 3046.

[0155] As Figure 17 and Figure 18 shown, the amount of movement exhibited by the walking person 114 is sufficient to result in an STM intensity M STM,r that is above a predetermined STM threshold T M,STM , and an LTM intensity M LTM,r that is above a predetermined LTM threshold T M,LTM . Accordingly, locations 3024 and 3044 are identified as peaks. However, peak 3044 is delayed relative to peak 3024. Figure 30 Peak 3004 is also shown to be delayed relative to peak 3024.

[0156] As can be seen from Figure 30 , the output of step 314b is peak 3024, and the output of step 316b is peaks 3044 and 3046. As can also be seen from the range FFT amplitude plot of Figure 30 , a conventional target detection method relying on amplitude peaks of the range FFT would detect, for example, 3 targets associated with peaks 3002, 3004 and 3006, or fail to detect peak 3004.

[0157] Figures 21-2431 corresponds to frame 115 of map charts 1000, 1100, and 1200. At frame 115, the human target 114 is walking towards the millimeter-wave radar 102 and is about 3m away from the millimeter-wave radar 102.

[0158] like Figure 31 As shown, the distance FFT amplitude curve (corresponding to) Figure 12 This includes peak value 3102 corresponding to RF leakage, peak value 3104 corresponding to human target 114, peak value 3106 corresponding to the wall, and peak value 3108 corresponding to the phantom target. STM amplitude curve (corresponding to...) Figure 10 And calculated using Formula 2) including the peak value 3124 corresponding to human target 114. LTM amplitude curve (corresponding to Figure 11 (and calculated using Formula 3) including the peak value 3144 corresponding to the human target 114 and the peak value 3146 corresponding to the wall.

[0159] like Figure 31 As shown, at positions 3122 and 3142, there are no leakage peaks in the STM or LTM curves. Similarly, as... Figure 31 As shown, the peak value 3108 of the phantom target associated with the distance FFT amplitude curve does not exist in the STM or LTM curve.

[0160] For example Figure 31 As shown, the wall peaks present in the distance FFT curve (peak 3106) and LTM curve (peak 3146) are not present in the STM curve, as shown at position 3126. For example, Figure 23 The diagram shows the IQ curve of the STM at position 3126 at frame 115. Figure 24 The diagram shows the IQ curve of the LTM at peak value 3146 at frame 115. Figure 23 and Figure 24 As shown, the amount of movement exhibited by the wall is much smaller in the STM curve than in the LTM curve. This smaller amount of movement results in a smaller STM intensity M. STM,r Below the predetermined STM threshold T M,STM And therefore not identified as a peak. The amount of movement exhibited by the walls in the LTM curve results in the LTM intensity M. LTM,r Above the predetermined LTM threshold T M,LTM And thus it was identified as a peak value of 3146.

[0161] like Figure 21 and Figure 22 As shown, the amount of movement exhibited by the walking person 114 is sufficient to cause the STM intensity M STM,r Above the predetermined STM threshold T M,STM And LTM strength MLTM,r above a predetermined LTM threshold T M,LTM Thus, locations 3124 and 3144 are identified as peaks. However, peak 3144 is delayed relative to peak 3124.

[0162] As can be seen from Figure 31 the output of step 314b is peak 3124 and the output of step 316b is peaks 3144 and 3146. From Figure 31 the distance FFT magnitude plot, it can also be seen that a conventional target detection method relying on magnitude peaks of the distance FFT would detect, for example, 4 targets associated with peaks 3102, 3104, 3106, and 3108.

[0163] Figures 25-28 and 33 correspond to frame 253 of map plots 1000, 1100, and 1200. At frame 253, human target 114 is close to the wall and turning to start walking away from the wall and toward millimeter wave radar 102. Figure 32 and Figure 34 correspond to frames 243 and 263 of map plots 1000, 1100, and 1200, respectively.

[0164] As shown in Figure 33 the distance FFT magnitude plot (which corresponds to Figure 12 ) includes a peak 3302 corresponding to RF leakage and a peak 3308 corresponding to a ghost target. The distance FFT magnitude plot also includes a peak 3306 corresponding to the wall, which masks a peak 3304 corresponding to human target 114.

[0165] The STM magnitude plot (corresponding to Figure 10 and calculated using Equation 2) includes a peak 3324 corresponding to human target 114. The LTM magnitude plot (corresponding to Figure 11 and calculated using Equation 3) includes a peak 3346 corresponding to the wall, which masks a peak 3344 corresponding to human target 114.

[0166] As shown in Figure 33 there is no leakage peak present in the STM or LTM plots at locations 3322 and 3342. Also as shown in Figure 33 the peak 3308 associated with the ghost target in the distance FFT magnitude plot is not present in the STM or LTM plots.

[0167] Also as shown in Figure 33 the wall peak present in the distance FFT plot (peak 3306) and the LTM plot (peak 3346) is not present in the STM plot, as shown at location 3326. For example, Figure 27An I-Q plot at frame 253 is shown for the STM at position 3326. Figure 28 An I-Q plot at frame 253 is shown for the LTM at peak 3346. As Figure 27 and Figure 28 shown, the amount of movement exhibited by the wall is much smaller in the STM plot than in the LTM plot. This smaller amount of movement results in an STM intensity M STM,r that is below a predetermined STM threshold T M,STM and thus is not identified as a peak. The amount of movement exhibited by the wall in the LTM plot results in an LTM intensity M LTM,r that is above a predetermined LTM threshold T M,LTM and thus is identified as a peak 3346.

[0168] As Figure 26 shown, the amount of movement exhibited by the walking human 114 is sufficient to result in an LTM intensity M M,LTM that is above the predetermined LTM threshold T LTM,r . However, as Figure 33 shown, the peak 3344 is obscured by the peak 3346, and only the peak 3346 is detected (as the peak 3346 is a local maximum).

[0169] As Figure 25 shown, the amount of movement exhibited by the walking human 114 is not sufficient to result in an STM intensity M M,STM that is above the predetermined STM threshold T STM,r . Thus, the peak 3322 is not identified as a peak.

[0170] As can be seen from Figures 32-34 , the peak 3324 is smaller than the peaks 3224 and 3424, as the human target is not moving radially in frame 3324, but turning. It is thus possible that, in some embodiments (e.g., where the STM threshold T M,STM is above 300), the output of step 314b is 0 peaks (no peaks are detected for frame 253 during step 314b), and the output of step 316b is the peak 3346. It can also be seen from the range FFT magnitude plot of Figure 33 that a conventional target detection method relying on the magnitude peaks of the range FFT would detect, e.g., 3 targets associated to peaks 3302, 3306 and 3308.

[0171] As will be explained in more detail later, the failure to identify the peak 3324 does not result in termination of the tracking of the trajectory of the human target 114. For example, as Figure 32 shown (corresponding to frame 243), the STM plot has a peak 3224, which is detected as a peak in step 314b.Figure 32 The LTM curve also shows a peak of 3244, which is detected as a peak in step 316b. Therefore, when the peak of the STM curve 3324 is not detected in frame 253, the peak of the LTM curve 3346 is associated with the target because it is close to the peak of frame 243. As a result, the condition STM == 0 < M == 1 is satisfied, causing the human target 114 to move from the moving state 406 to the uncertain state 404 and remain in the uncertain state until the counter C expires (or a new corresponding STM peak is detected), as shown. Figure 8B and Figure 8C As shown. In some embodiments, even without an LTM peak associated with the trajectory, the human target 114 moves from a moving state 406 to an uncertain state 404 because, as Figure 8C As shown, the condition LTM==X is satisfied.

[0172] If counter C fails before detecting the STM peak, the condition STM == 0 is satisfied, thus making the human target 114 subject to the condition LTM == 1 (e.g., Figure 8B In the case shown, the trajectory moves from the uncertain state 404 to the stationary state 408 because the trajectory is activated (A==1) when the human target 114 transitions to the moving state 406. Figure 8D As shown. Once in a stationary state 408, the trajectory tracking the human target is not terminated, and an LTM peak is detected, as shown. Figure 8D As shown. Figure 8B As shown, if the counter fails when the conditions STM==0 and LTM==0 are met, the trajectory is terminated.

[0173] like Figure 34 As shown (corresponding to frame 263), the STM curve has a peak value of 3424, which is detected as a peak value in step 314b. Figure 34 The LTM curve also shows a peak value of 3446, which is detected as a peak value in step 316b because it is close to the peak value of 3346. Therefore, the condition STM == 1 is satisfied, and the human target 114 transitions from a static state 408 to an uncertain state 404 based on the counter C (e.g., ...). Figure 8D As shown), then based on counter C, it transitions from uncertain state 404 to moving state 406 (as shown). Figure 8B (As shown).

[0174] Figures 35-37 Each of the embodiments according to the present invention is shown as follows: Figure 10 and Figure 11Graphs 3500, 3600, and 3700 of the output of step 620 for each of the trajectories of the walking person 114 captured in FIGS. 12A-12C. Graph 3500 also shows the state of the targets (712), while graph 3600 shows these trajectories by the trajectory ID (702) of the trajectory.

[0175] As shown in FIG. 12A, trajectory 3502 corresponds to the walking person 114. As can be seen in FIG. 12B, when the walking person 114 approaches the wall, a second trajectory 3504 is generated for the wall. However, because the wall is a stationary object, the wall appears in the LTM graph but not in the STM graph. As a result, trajectory 3504 is never activated and terminated (steps 534 and 526) once the timer expires. Noise that can initially be tracked as a potential target (e.g., shown by trajectory 3516) similarly is never activated and terminated (steps 534 and 526) once the timer expires. Figures 35-37 Figure 35 Figure 36 As shown in FIG. 12A, trajectory 3502 corresponds to the walking person 114. As can be seen in FIG. 12B, when the walking person 114 approaches the wall, a second trajectory 3504 is generated for the wall. However, because the wall is a stationary object, the wall appears in the LTM graph but not in the STM graph. As a result, trajectory 3504 is never activated and terminated (steps 534 and 526) once the timer expires. Noise that can initially be tracked as a potential target (e.g., shown by trajectory 3516) similarly is never activated and terminated (steps 534 and 526) once the timer expires.

[0176] As shown in FIG. 12A, trajectory 3502 corresponds to the walking person 114. As can be seen in FIG. 12B, when the walking person 114 approaches the wall, a second trajectory 3504 is generated for the wall. However, because the wall is a stationary object, the wall appears in the LTM graph but not in the STM graph. As a result, trajectory 3504 is never activated and terminated (steps 534 and 526) once the timer expires. Noise that can initially be tracked as a potential target (e.g., shown by trajectory 3516) similarly is never activated and terminated (steps 534 and 526) once the timer expires. Figure 35

[0177] Figures 35-37 The velocity of trajectories 3502 and 3504 are also shown with graphs 3512 and 3514, respectively.

[0178] Graph 3700 is similar to graph 3600. However, graph 3700 shows only the activated trajectories. Since trajectory 3504 never transitions to the moving state 406, the trajectory is not activated.

[0179] Figure 38 Graph 3800 showing the conventional tracking of the walking person shown in FIGS. 13A-13C, where the graph is generated by identifying targets based on the peaks of the distance FFT amplitude graph (as shown in FIG. 13B), and where the Doppler FFT is used to determine the velocity of the targets. Figures 35-37 Figures 29-34 As shown in FIG. 13A, the conventional tracking results in tracking a phantom target, and can result in target splitting for higher target velocities.

[0180] As shown in FIG. 13A, the conventional tracking results in tracking a phantom target, and can result in target splitting for higher target velocities. Figure 38

[0181] Figure 37 ​​​​​​As shown, some embodiments advantageously avoid target segmentation and phantom target tracking by relying on the STM peak value used for target recognition. Additional advantages of some embodiments include improved velocity estimation, for example, as with... Figure 38 Compared to time Figure 37 As shown.

[0182] Some embodiments offer advantages such as avoiding tracking stationary objects, like walls or furniture, by activating the trajectory only after initial movement is detected within a minimal time period. By performing a time-domain-based survey on the complex distance FFT output instead of a peak search within the distance FFT amplitude, and by performing velocity determination using a time-domain-based survey instead of a Doppler FFT, some embodiments advantageously achieve successful target tracking and improved distance and velocity estimation with lower computational effort compared to conventional tracking using conventional distance and velocity estimation methods, such as peak search within the distance FFT amplitude and Doppler FFT, respectively.

[0183] Additional advantages of some embodiments include the ability to achieve smooth measurement data through the use of α-β filtering (e.g., such as Equations 5 and 6) and / or median filtering of the tracking output (distance and / or velocity).

[0184] Figures 39-41 Graphs 3900, 4000, and 4100, respectively, show the output of step 620 (tracked by trajectory ID 702) when a walking human body 113 is tracked in a stepwise manner moving away from and towards the millimeter-wave radar 102 according to an embodiment of the present invention. Graph 3900 also shows the state (712) of the target, while graph 4000 shows these trajectories via the trajectory ID (702) of the trajectory. Graph 4100 is similar to graph 4000. However, graph 4100 only shows the active trajectories.

[0185] like Figures 39-41 As shown, trajectory 3902 corresponds to the walking person 113. Figure 39 As can be seen, when the walking person 113 stops, the target state (712) transitions from the moving state 402 to the uncertain state 404 (e.g., as shown in position 3906). If the walking person 113 remains stopped for a long time, the target state transitions from the uncertain state 404 to the stationary state 408, for example, as shown in position 3910. When the walking person 113 resumes walking, the target state transitions from the stationary state 408 to the uncertain state 404, and then from the uncertain state 404 to the moving state 406, as shown in positions 3912 and 3908 respectively.

[0186] like Figure 39 and Figure 41As shown, even if the wall occasionally becomes a potential target (in an uncertain state 404), as shown by potential trajectory 3904, such trajectory is not activated and is eventually terminated. Similarly, even if noise might become a potential trajectory (as shown by potential trajectory 3938), such trajectory is not activated and is eventually terminated.

[0187] Figures 39-41 The velocities of trajectories 3902 and 3904 are also shown by curves 3914 and 3916, respectively.

[0188] Some embodiments can implement a frame skipping mode. In more skipped frames, one or more frames are skipped, for example, during the transmission of a chirped signal (e.g., in step 302). For example, in some embodiments, when the frame skipping is set to 4, frame 1 is transmitted, and then other frames are not transmitted until frame 5. In other embodiments, frame skipping is actually performed, where all frames are transmitted by the millimeter-wave radar 102, but some frames are skipped and not processed, for example, for detecting and tracking targets. For example, in some embodiments, when the frame skipping is set to 4, the millimeter-wave radar 102 transmits all frames, but only one frame out of every four is processed. By processing only a subset of frames, some embodiments achieve power savings (e.g., by increasing processor idle time).

[0189] Apart from frame skipping, all other operations remain the same as when frame skipping is not used. For example, if the frame time FT is 50ms without frame skipping, the frame time FT with 4 frame skips is 200ms. Regarding Equation 3, w refers to the actual frame used during the generation of distance data (in step 310), not the skipped frame.

[0190] Some implementations can advantageously achieve power savings without substantially degrading performance when using frame skipping modes. For example, Figure 42 This illustrates an embodiment of the invention, when using frame skipping mode to track, such as Figure 10 and Figure 11 The graph 4200 shows the output of step 620 when 114 walking people were captured. Figure 42 In this embodiment, the frame skipping is set to 4. Graph 4200 only shows the active trajectory. Figure 42 The data shown is actually from the data used to generate Figures 35-38 The same data was generated (by using only one frame out of every four frames, making the frame time FT 200ms).

[0191] like Figure 42 As shown, only 150 frames are displayed, not... Figure 37 The 600 frames shown are due to setting the frame skipping to 4. Figure 42 As shown, trajectory 4202 tracks the walking target 114, although... Figure 37The trajectory 3902 is slightly delayed compared to the previous one. As shown in curve 4212, the speed of the walking target 114 was also successfully tracked. Figure 42 As shown, no phantom targets or stationary objects such as walls are effectively tracked.

[0192] Some embodiments can implement a low-power mode. In low-power mode, each frame contains only a single chirped signal. Therefore, the STM peak, which is identified based on multiple chirped signals per frame using Equation 2, is not used during low-power mode. Conversely, when the speed S... w Greater than a predetermined speed threshold S min In low-power mode (STM=1), the STM peak is identified. Therefore, some embodiments may use state machine 400 when operating in low-power mode.

[0193] Some implementations can advantageously achieve power savings when using low-power modes. Figure 43 This illustrates an embodiment of the invention, showing how tracking is performed using a low-power mode, such as... Figure 11 The graph 4300 is the output of step 620 when the walking person 114 is captured. Graph 4300 only shows the active trajectory. Graph 4300 uses a frame time of 50ms (FT).

[0194] like Figure 43 As shown, trajectories 4302 and 4304 track walking target 114, although target segmentation occurs when the walking target is near a wall (around frame 253). The velocity of walking target 114 is also successfully tracked, although target segmentation is also observed, as shown by curves 4312 and 4314. In some embodiments, in addition to increasing sensitivity to target segmentation, the low-power mode advantageously avoids tracking phantom targets and stationary objects with active trajectories.

[0195] By limiting the number of active trajectories output during low-power mode to a single active trajectory and associating the nearest target with the active trajectory, some embodiments can avoid target segmentation and improve performance in low-power mode (resulting in a single active trajectory). For example, some embodiments may generate more than one active trajectory during low-power mode; however, only the active trajectory closest to the millimeter-wave radar 102 is output during low-power mode.

[0196] Figure 44 This illustrates an embodiment of the invention, showing how low-power mode and frame skipping tracking are used. Figure 10 and Figure 11 The graph 4400 shows the output of step 620 when 114 walking people were captured. Figure 44 In this embodiment, the frame skipping is set to 4. Graph 4400 only shows the active trajectory.

[0197] The graph table 4400 uses a frame time FT of 200 ms, and is in fact generated from the same data used to generate the graph 4300 (by using only one frame out of every four frames).

[0198] As Figure 44 illustrated, and similar to Figure 42 , only 150 frames are shown because the frame skipping is set to 4. As Figure 44 illustrated, the trajectory 4402 tracks the walking target 114. As shown by the curve 4412, the speed of the walking target 114 is also successfully tracked. As Figure 44 illustrated, no phantom targets or stationary objects such as walls are tracked by the effective trajectory.

[0199] As Figure 44 can be seen, even with the use of the low power mode by limiting the number of trajectories to one and by associating the detected closest target to the single trajectory, target splitting is advantageously avoided.

[0200] Figure 45 A graph table 4500 showing the output of the steps 620 when tracking the human body 113 walking away from and towards the millimeter wave radar 102 in a step-wise fashion using the low power mode and frame skipping, according to an embodiment of the application, is shown. In Figure 45 the embodiment, the frame skipping is set to 4. The graph table 4500 only shows the active trajectories. The graph table 4500 is in fact generated from the same data used to generate the graph 4300 (by using only one frame out of every four frames, so that the frame time FT is 200 ms). Figures 39-41 As

[0201] illustrated, only 150 frames are shown because the frame skipping is set to 4 instead of the 600 frames shown in Figure 45 . As Figure 41 illustrated, the trajectory 4502 tracks the walking target 113. As shown by the curve 4512, the speed of the walking target 113 is also successfully tracked. As Figure 45 illustrated, no phantom targets or stationary objects such as walls are tracked by the effective trajectory. Figure 45

[0202] Although the performance of the distance estimation and speed estimation can be better without the low power mode and frame skipping, in some embodiments, the combination of the low power mode and frame skipping advantageously results in power savings, while still successfully tracking the target and successfully performing the distance estimation and speed estimation.

[0203] Figures 46-48 ​Plots 4600, 4700, and 4800 show the output of step 620 when the walking person 115 leaves the mmWave radar 102 in a step-wise fashion, according to embodiments of the application. Plot 4600 also shows the state of the target (712), while plot 4700 shows the tracks by track ID (702) of the tracks. Plot 4800 is similar to plot 4700. However, plot 4800 shows only the active tracks.

[0204] As shown in FIG. 47, the track 4602 corresponds to the walking person 115. As can be seen in FIG. 47, when the walking person 115 stops, the state of the target (712) transitions to the stationary state 408, and the track is not terminated even though the target remains in the stationary state 408 for a long time, as shown by position 4608. Figures 46-48 Figure 46 As can be seen in FIG. 47, when the walking person 115 stops, the state of the target (712) transitions to the stationary state 408, and the track is not terminated even though the target remains in the stationary state 408 for a long time, as shown by position 4608.

[0205] In some embodiments, when a human target remains in the stationary state 408 for longer than a predetermined period of time (e.g., such as 10 frames), the processor 104 can determine a vital sign (such as a heart rate and / or a respiration rate) of the human target while the target remains in the stationary state 408. The processor 104 can stop monitoring the vital sign when the target transitions out of the stationary state 408.

[0206] In some embodiments, the vital sign can be determined using the mmWave radar 102 in a manner known in the art. In some embodiments, the vital sign can be determined using the mmWave radar 102 as described in co-pending U.S. Patent Application No. 16 / 794,904, filed February 19, 2020, and entitled “Radar Life Signal Tracking Using Kalman Filter,” and / or co-pending U.S. Patent Application No. 16 / 853,011, filed April 20, 2020, and entitled “Radar-Based Vital Sign Estimation,” which are incorporated herein by reference.

[0207] Example embodiments of the application are summarized here. Other embodiments can be understood from the entirety of the description and claims submitted herewith.

[0208] Example 1. A method comprising: receiving reflected radar signals with a mmWave radar; performing a distance discrete Fourier transform (DFT) based on the reflected radar signals to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of distance bins; for each of the plurality of distance bins, determining a respective intensity value based on changes in the respective I and Q signals over time; performing a peak search over the plurality of distance bins based on the respective intensity value for each of the plurality of distance bins to identify a peak distance bin; and associating a target to the identified peak distance bin. ​

[0209] Example 2. The method of example 1, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals over time comprises determining the respective intensity value for each range bin based on changes in the respective I and Q signals over a single frame.

[0210] Example 3. The method of any one of examples 1 or 2, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals over a single frame comprises determining the respective intensity value for each range bin based on where PNrepresents the number of chirp signals per frame, R r,c+1 represents the value of the range bin R r for chirp signal c+1, and R r,c represents the value of the range bin R r for chirp signal c.

[0211] Example 4. The method of any one of examples 1-3, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals over time comprises determining the respective intensity value for each range bin based on changes in the respective I and Q signals over a plurality of frames.

[0212] Example 5. The method of any one of examples 1-4, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals over the plurality of frames comprises determining the respective intensity value for each range bin based on where W represents the number of frames, R r,i,w+1 represents the value of the range bin R r for chirp signal i of frame w+1, and R r,i,w represents the value of the range bin R r for chirp signal i of frame w.

[0213] Example 6. The method of any one of examples 1-5, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals over the plurality of frames comprises determining the respective intensity value for each range bin based on changes in the respective I and Q signals corresponding to a first chirp signal of each frame of the plurality of frames.

[0214] Example 7. The method of one of examples 1-6, wherein determining the respective intensity value for each range bin based on changes in the respective I and Q signals corresponding to the first chirp signal of each frame of the plurality of frames comprises determining the respective intensity value for each range bin based on where W represents the number of frames, Rr,1,w+1 a range bin R of a chirp signal 1 of frame w+1 r a value of R r,1,w a range bin R of a chirp signal 1 of frame w r a value of R.

[0215] Example 8. The method of any of Examples 1-7, wherein each frame of the plurality of frames includes only a single chirp signal, the method further comprising: determining a velocity of the target; and associating a peak to the target when the determined velocity is above a predetermined velocity threshold.

[0216] Example 9. The method of any of Examples 1-8, further comprising: assigning a state to the target; and updating the state based on a previous state and an identified peak range bin.

[0217] Example 10. The method of any of Examples 1-9, further comprising: identifying a second peak range bin based on the performed peak search; associating a second target to the second peak range bin; assigning a second state to the second target; and updating the second state based on a previous second state and the identified second peak range bin.

[0218] Example 11. The method of any of Examples 1-10, wherein assigning the state to the target comprises assigning the state to the target from a set of states, wherein the set of states includes an uncertain state, a moving state indicating movement of a target, and a stationary state indicating no movement of a target.

[0219] Example 12. The method of any of Examples 1-11, further comprising tracking the target with a trail, wherein the trail is activated when a target transitions into the moving state, and wherein the target transitions into the stationary state only if the trail is activated.

[0220] Example 13. The method of any of Examples 1-12, further comprising: tracking the target with a trail; and terminating the trail when a timer expires and the target is in the uncertain state.

[0221] Example 14. The method of any of Examples 1-13, wherein associating the target to the identified peak range bin comprises creating a trail and transitioning the target to the uncertain state.

[0222] Example 15. The method of any of Examples 1-14, further comprising: determining a distance of the target based on the identified peak range bin; and determining a velocity of the target based on the determined distance.

[0223] Example 16. The method of any one of examples 1-15, wherein determining the velocity of the target comprises performing a derivative of the distance of the target.

[0224] Example 17. The method of any one of examples 1-16, further comprising transmitting, with the millimeter wave radar, a radar signal, wherein the reflected radar signal is based on the transmitted radar signal, and wherein the transmitted radar signal comprises a linear chirp signal.

[0225] Example 18. The method of any one of examples 1-17, wherein the target is a human target.

[0226] Example 19. An apparatus comprising: a millimeter wave radar configured to transmit a chirp signal and receive a reflected chirp signal; and a processor configured to: perform a distance discrete Fourier transform (DFT) based on the reflected chirp signal to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of distance bins, determine, for each of the plurality of distance bins, a respective intensity value based on a change in the respective I and Q signals over time, perform a peak search over the plurality of distance bins based on the respective intensity value for each of the plurality of distance bins to identify a peak distance bin, and associate a target to the identified peak distance bin.

[0227] Example 20. A method comprising: receiving, with a millimeter wave radar, a reflected radar signal; performing a distance fast Fourier transform (FFT) based on the reflected radar signal to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of distance bins; determining, for each of the plurality of distance bins, a respective short-term motion value based on a change in the respective I and Q signals in a single frame; performing a peak search over the plurality of distance bins based on the respective short-term motion value for each of the plurality of distance bins to identify a short-term peak distance bin; and associating a target to the identified short-term peak distance bin.

[0228] Example 21. The method of example 20, further comprising: determining, for each of the plurality of distance bins, a respective long-term motion value based on a change in the respective I and Q signals over a plurality of frames; performing a peak search over the plurality of distance bins based on the respective long-term motion value for each of the plurality of distance bins to identify a long-term peak distance bin; and associating the identified long-term peak distance bin to the target.

[0229] While the application has been described with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other embodiments of the application, will be apparent to persons skilled in the art upon reference to the description. It is therefore intended that the appended claims encompass any such modifications or embodiments.

Claims

1. A method for target tracking, comprising: receiving reflected radar signals with a millimeter wave radar; performing a distance discrete Fourier transform (DFT) based on the reflected radar signals to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of range bins; for each of the plurality of range bins, determining a respective intensity value based on changes in the respective I and Q signals over time; performing a peak search across the plurality of range bins based on the respective intensity values for each of the plurality of range bins to identify a peak range bin; and associating a target to the identified peak range bin, wherein the method further comprises: assigning the target a state from a set of states, wherein the set of states includes an uncertain state associated to a potential target, a moving state indicating that an actual target is moving, and a stationary state indicating that an actual target is stationary; updating the state based on a previous state and the identified peak range bin; and tracking the target with a track, wherein: the track is activated when the target transitions into the moving state, and only while the track is activated, the target transitions into the stationary state; or the track is terminated when a timer expires and the target is in the uncertain state.

2. The method of claim 1, wherein determining a respective said intensity value for each range bin based on changes in respective said I and Q signals over time comprises: The respective intensity value for each range bin is determined based on changes in the respective I and Q signals over a single frame.

3. The method of claim 2, wherein determining a respective said intensity value for each range bin based on changes in respective said I and Q signals over said single frame comprises: The respective intensity value for each range bin is determined based on The respective intensity value for each range bin is determined based on changes in the respective I and Q signals over a plurality of frames. where PNrepresents the number of chirp signals per frame, R r,c+1 represents the value of the range bin R r for the chirp signal c+1, and R r,c represents the value of the range bin R r for the chirp signal c.

4. The method of claim 1, wherein determining a respective said intensity value for each range bin based on changes in respective said I and Q signals over time comprises: The respective intensity value for each range bin is determined based on 5. The method of claim 4, wherein determining a respective said intensity value for each range bin based on changes in the respective said I and Q signals over the plurality of frames comprises: The respective intensity value for each range bin is determined based on changes in the respective I and Q signals corresponding to a first chirp signal of each of the plurality of frames. The respective intensity value for each range bin is determined based on where W represents the number of frames, R r,i,w+1 represents the value of the range bin R r of chirp signal i of frame w+1, and R r,i,w represents the value of the range bin R r of chirp signal i of frame w.

6. The method of claim 4, wherein determining a respective said intensity value for each range bin based on changes in respective said I and Q signals over said plurality of frames comprises:

8. The method of claim 6, wherein each of the plurality of frames includes only a single chirp signal, the method further comprising:

7. The method of claim 6, wherein determining a respective said intensity value for each range bin based on variations of the respective said I and Q signals corresponding to the first chirp signal of each frame of the plurality of frames comprises: determining a velocity of the target; and where W represents the number of frames, R r,1,w+1 represents the value of the range bin R r of chirp signal 1 for frame w+1, and R r,1,w represents the value of the range bin R r of chirp signal 1 for frame w. associating a peak to the target when the determined velocity is higher than a predetermined velocity threshold.

9. The method of claim 1, further comprising: identifying a second peak range bin based on the performed peak search; associating a second target to the second peak range bin; assigning a second state to the second target; and updating the second state based on a previous second state and the identified second peak range bin. creating a track and transitioning the target to the uncertain state.

11. The method of claim 1, further comprising: determining a range of the target based on the identified peak range bin; 10. The method of claim 1, wherein associating the target to the identified peak distance bin comprises: and determining a velocity of the target based on the determined range. performing a derivative of the range of the target.

13. The method of claim 1, further comprising transmitting radar signals with the millimeter wave radar, wherein the reflected radar signals are based on the transmitted radar signals, and wherein the transmitted radar signals include a linear chirp signal. ​ 12. The method of claim 11, wherein determining the velocity of the target comprises: ​ ​ 14. The method of claim 1, wherein, The target is a human target.

15. An apparatus for target tracking, comprising: a millimeter wave radar configured to transmit a chirp signal and receive a reflected chirp signal; and a processor configured to: perform a distance Discrete Fourier Transform (DFT) based on the reflected chirp signal to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of range bins, for each of the plurality of range bins, determine a respective intensity value based on changes in the respective I and Q signals over time, perform a peak search across the plurality of range bins based on the respective intensity values for each of the plurality of range bins to identify a peak range bin, associate a target to the identified peak range bin, assign a state to the target from a set of states, wherein the set of states includes an uncertain state associated to a potential target, a moving state indicating that an actual target is moving, and a stationary state indicating that an actual target is stationary; update the state based on a previous state and the identified peak range bin; and track the target with a trajectory, wherein: the trajectory is activated when the target transitions into the moving state; and only while the trajectory is activated, the target transitions into the stationary state; or the trajectory is terminated when a timer expires and the target is in the uncertain state.

16. A method for target tracking, comprising: receiving a reflected radar signal with a millimeter wave radar; performing a distance Fast Fourier Transform (FFT) based on the reflected radar signal to generate an in-phase (I) signal and a quadrature (Q) signal for each of a plurality of range bins; for each of the plurality of range bins, determining a respective short-term movement value based on changes in the respective I and Q signals in a single frame; performing a peak search across the plurality of range bins based on the respective short-term movement values for each of the plurality of range bins to identify a short-term peak range bin; and associating a target to the identified short-term peak range bin, wherein the method further comprises: assigning a state to the target from a set of states, wherein the set of states includes an uncertain state associated to a potential target, a moving state indicating that an actual target is moving, and a stationary state indicating that an actual target is stationary; updating the state based on a previous state and the identified peak range bin; and tracking the target with a trajectory, wherein: the trajectory is activated when the target transitions into the moving state; and only while the trajectory is activated, the target transitions into the stationary state; or the trajectory is terminated when a timer expires and the target is in the uncertain state.

17. The method of claim 16, further comprising: for each of the plurality of range bins, determining a respective long-term movement value based on changes in the respective I and Q signals over a plurality of frames; performing a peak search across the plurality of range bins based on the respective long-term movement values for each of the plurality of range bins to identify a long-term peak range bin; and associating the identified long-term peak distance bin to the target.

Images