A human hand gesture recognition system and recognition method based on time reversal technology

Abstract

The present invention provides a system for recognizing human gestures based on time reversal technology, comprising a PC terminal forming a closed loop, a transmitter, a transmitting antenna r T , metal cavity, receiving antenna R and receiver, wherein the PC is connected to the transmitter and receiver at the same time, and the transmitter is connected to the transmitting antenna r T , the receiver is connected to the receiving antenna r R , the metal cavity is also connected to the transmitting antenna r T and receiving antenna r R The human hand gesture is located within the interior space of the metal cavity. The present invention also provides a method for recognizing human hand gestures using a time reversal technique. The solution of the present invention achieves super-resolution gesture recognition, accurately recognizing gestures using low-frequency electromagnetic waves. This approach is both cost-effective and highly accurate, and its robustness to interference in complex environments is also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of human gesture recognition, and in particular to a human gesture recognition system and method based on time reversal technology. Background Art

[0002] Human motion recognition, especially human gesture recognition technology, has far-reaching theoretical research significance and broad application prospects in many fields today. Current recognition methods are mainly based on video image information, sensors, and radio frequency microwaves.

[0003] Video-based recognition uses computers to analyze images and video captured by cameras to identify human actions. This traditional motion recognition method is now highly developed, highly accurate, and intuitive. However, it suffers from drawbacks such as significant impact of lighting angle, viewing angle, and background changes, relatively high energy consumption and computing resource requirements, making it difficult to use in portable devices, and privacy and security issues. Currently, there are also infrared recognition technologies that are less affected by light, provide good privacy protection, and can achieve non-line-of-sight recognition. There is also human motion recognition based on skeleton information, which can protect privacy, but these technologies are relatively costly.

[0004] Sensor-based recognition uses sensors (such as accelerometers and gyroscopes) to collect characteristic parameters such as acceleration, pressure, angle, and height, and then processes these parameters through subsequent algorithms to identify human movements. While this approach offers excellent accuracy, stability, and immediacy, it often suffers from expensive and bulky equipment, requires specialized locations and limited range, offers a poor user experience, and requires the subject to actively cooperate.

[0005] Radio frequency microwave-based recognition relies on changes in radio signals caused by human motion. Since different motions correspond to different signal changes, training can be used to map different human motions to these changes, building corresponding models and identifying the corresponding motions. This approach includes two main approaches: using Universal Software Radio Peripheral (USRP) technology and using WiFi technology. Currently, WiFi technology, due to its low frequencies and limitations imposed by the Rayleigh diffraction limit, offers limited accuracy and difficulty recognizing delicate gestures. USRP technology, on the other hand, often involves high frequencies, resulting in complex equipment, high power consumption, and high cost.

[0006] From the above, it can be seen that there is an urgent need for a low-cost, high-precision gesture recognition method. Summary of the Invention

[0007] In view of the shortcomings of the prior art, the present invention provides a human hand gesture recognition method based on time reversal technology, which solves the problems of high cost and low accuracy of the current hand gesture recognition methods in the prior art.

[0008] The above technical objectives of the present invention are achieved through the following technical solutions:

[0009] A system for recognizing human gestures based on time reversal technology, comprising a PC terminal, a transmitter, and a transmitting antenna forming a closed loop. T , metal cavity, receiving antenna R and receiver, wherein the PC is connected to the transmitter and receiver at the same time, and the transmitter is connected to the transmitting antenna r T , the receiver is connected to the receiving antenna r R , the metal cavity is also connected to the transmitting antenna r T and receiving antenna r R ; The human body gestures are located in the inner space of the metal cavity.

[0010] The present invention also provides a method for recognizing human gestures based on a time reversal technique, comprising the following steps:

[0011] Step 1: The PC outputs signal S1(t) to the transmitter, which transmits signal S1(t) to the transmitting antenna r T , transmitting antenna r T The transmitted signal S1(t) passes through the metal cavity environment h0(t) and is received by the receiving antenna r R and the receiver receives it, which is recorded as signal S2(t). At this time, the metal cavity environment h0(t) does not contain human gestures;

[0012] Step 2: Time-reverse the signal S2(t) at the PC to obtain the signal S3(t) = S2(-t);

[0013] Step 3: The PC outputs signal S3(t), and the transmitter transmits signal S3(t) to the transmitting antenna r T , transmitting antenna r T Send signal S3(t), passing through the metal cavity environment h i (t) is then received by antenna r R and the receiver receives it, which is recorded as signal S4(t). At this time, the metal cavity environment h i (t) contains human gestures;

[0014] Step 4, erase the focus peak of S4(t) to obtain S5(t);

[0015] Step 5, perform short-time Fourier transform on S5(t) to obtain F(w, t);

[0016] Step 6, extract the phase Arg[F(w, t)] of F(w, t) and draw a grayscale image;

[0017] Step 7: Use the trained convolutional neural network to recognize the grayscale image.

[0018] The present invention has the following advantages: the solution of the present invention realizes super-resolution gesture recognition, that is, gestures can be accurately recognized using low-frequency electromagnetic waves, with low cost and high accuracy, and its anti-interference ability in complex environments is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the present invention;

[0020] Figure 2 This is a schematic diagram of a scenario for recognizing human gestures in a metal cavity according to the present invention;

[0021] Figure 3 (a) is signal S1(t); (b) is S2(t); (c) is S3(t); (d) is S4(t); and (e) is S5(t).

[0022] Figure 4 is the grayscale image finally obtained in the embodiment of the present invention;

[0023] Figure 5 These are the five basic gestures to be recognized in the embodiment of the present invention. DETAILED DESCRIPTION

[0024] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0025] Example:

[0026] A system for recognizing human gestures based on time reversal technology, comprising a PC terminal, a transmitter, and a transmitting antenna forming a closed loop. T , metal cavity, receiving antenna R and receiver, where the PC is connected to both the transmitter and receiver, and the transmitter is connected to the transmitting antenna r T , the receiver is connected to the receiving antenna r R , the metal cavity is connected to the transmitting antenna r T and receiving antenna r R ; The human body gestures are located in the inner space of the metal cavity.

[0027] The transmitter plays the role of inputting the transmitting signal, the receiver plays the role of outputting the receiving signal, and the PC side both inputs the signal to the transmitter and receives the signal from the receiver.

[0028] The five basic gestures to be recognized are Figure 5As shown, experiments were conducted using the recognition method proposed in the present invention to recognize human hand gestures in actual situations.

[0029] Step 1: PC outputs signal S1(t) (e.g. Figure 3 (a) to the transmitter, which transmits the signal S1(t) to the transmitting antenna r T , transmitting antenna r T The transmitted signal S1(t) passes through the metal cavity environment h0(t) and is received by the receiving antenna r R and the receiver receives it, recorded as signal S2(t) (as Figure 3 As shown in (b), the metal cavity environment h0(t) does not contain human gestures;

[0030] Step 2: Time-reverse the signal S2(t) at the PC to obtain the signal S3(t)=S2(-t) (e.g. Figure 3 (as shown in (c));

[0031] Step 3: The PC outputs signal S3(t), and the transmitter transmits signal S3(t) to the transmitting antenna r T , transmitting antenna r T Send signal S3(t), passing through the metal cavity environment h i (t) is then received by antenna r R and the receiver receives it, which is recorded as signal S4(t) (as Figure 3 (d)), the metal cavity environment h i (t) contains human gestures;

[0032] Step 4: erase the focus peak of S4(t) to obtain S5(t) (as Figure 3 (e)

[0033] Step 5, perform short-time Fourier transform on S5(t) to obtain F(w, t);

[0034] Step 6, extract the phase Arg[F(w, t)] of F(w, t) and draw a grayscale image;

[0035] Step 7: Use the trained convolutional neural network to recognize the grayscale image.

[0036] Table 1

[0037]

[0038] Table 2

[0039]

[0040] The final recognition results are shown in Table 1 and Table 2:

[0041] Table 1 shows the combined recognition results for gestures two and three, while Table 2 shows the combined recognition results for all five gestures. As can be seen, the recognition results for the two gestures are excellent, and the average recognition score for the five gestures reaches 87%. This demonstrates the feasibility of this system for super-resolution human gesture recognition. Therefore, with improvements and optimizations to the hardware and software systems, such as adopting more complex neural networks, collecting more gesture data, replacing high-performance antennas, and improving the metal cavity environment to reduce noise, the recognition rate can be further improved.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A system for recognizing human gestures based on time reversal technology, characterized by: Including the PC end, transmitter, transmitting antenna r that form a closed loop T , metal cavity, receiving antenna R and receiver, wherein the PC is connected to the transmitter and receiver at the same time, and the transmitter is connected to the transmitting antenna r T , the receiver is connected to the receiving antenna r R , the metal cavity is also connected to the transmitting antenna r T and receiving antenna r R ;The human hand gesture is located in the inner space of the metal cavity; The identification method of the system includes the following steps: Step 1: Output signal from PC S 1 （t） To the transmitter, which transmits the signal S 1 （t） To the transmitting antenna r T , transmitting antenna r T Sending a signal S 1 （t） , through the metal cavity environment h 0 （t） The received antenna r R and the receiver receives it, recorded as signal S 2 （t） , at this time the metal cavity environment h 0 （t） It does not contain human gestures; Step 2: Transmit the signal S 2 （t） Get the signal by time reversal on the PC side S 3 （t） = S 2 （-t） ; Step 3: Output signal from PC S 3 （t） , the transmitter transmits the signal S 3 （t） To the transmitting antenna r T , transmitting antenna r T Sending a signal S 3 （t） , through the metal cavity environment h i （t） The received antenna r R and the receiver receives it, recorded as signal S 4 （t） , at this time the metal cavity environment h i （t） To contain human gestures; Step 4: S 4 （t） The focus peak is erased to obtain S 5 （t） ; Step 5, S 5 （t） Perform short-time Fourier transform to obtain F(w,t) ; Step 6, Extract F(w,t) Phase Arg [ F(w,t) ], draw a grayscale image; Step 7: Use the trained convolutional neural network to recognize the grayscale image.

Images