Remote awakening method and device for low-power-consumption monitoring equipment

Abstract

The invention relates to a remote wakeup method and device for low-power-consumption monitoring equipment. The remote wakeup method and device are used for solving the problem of resource waste caused by fixed-point wakeup monitoring or continuous monitoring in the prior art. According to the method, the mainboard module enters the low-power-consumption mode by controlling the low-power-consumption monitoring equipment to be powered on, and the video signal processing board is not powered at the moment, so that the standby power consumption is reduced. In the low power consumption mode, the device can receive a remote signal issued by the server and identify whether the remote signal is a wake-up signal. And if the signal is identified as a wake-up signal, executing wake-up processing, and controlling the equipment to turn on a power switch of the video signal processing board to ensure that the onboard circuit sends a related operation instruction after being electrified, so that the onboard circuit executes a real-time video monitoring function. When the video signal processing board does not need to monitor the task, the equipment turns off the power switch to reduce unnecessary energy consumption. According to the method, the power supply state of the video signal processing board is dynamically managed while the real-time monitoring requirement is ensured, and energy consumption and functions are effectively balanced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a remote wake-up method and device for low-power monitoring equipment. Background Art

[0002] With the development of Internet of Things technology, low-power monitoring devices have been widely used in remote scenarios. However, the existing technology mainly adopts two methods: fixed-point wake-up or continuous monitoring, which has obvious resource waste problems. Fixed-point wake-up starts the device by preset time, but it cannot flexibly respond to sudden monitoring needs, resulting in untimely monitoring or frequent operation of the device during unnecessary periods, reducing energy utilization efficiency. Although continuous monitoring achieves real-time performance, the high power consumption caused by long-term operation puts higher requirements on device endurance and maintenance costs, especially in scenarios where battery power is supplied or power supply is inconvenient. Therefore, how to effectively reduce device power consumption while meeting real-time monitoring needs has become a problem that needs to be solved urgently in existing technologies. Summary of the invention

[0003] The main purpose of the present invention is to provide a remote wake-up method and device for a low-power monitoring device, so as to solve the technical problem of how to effectively reduce the power consumption of the device while meeting the real-time monitoring requirements.

[0004] To achieve the above-mentioned purpose, the present invention provides a remote wake-up method for a low-power monitoring device, comprising the following steps: controlling the low-power monitoring device to perform power-on processing so that the mainboard module of the low-power monitoring device enters a low-power mode, wherein when the mainboard module is in the low-power mode, the video signal processing board is not powered; when the low-power monitoring device is in the low-power mode, receiving a remote signal sent by a server, and identifying whether the remote signal is a wake-up signal for waking up the low-power monitoring device, and if so, performing wake-up processing on the low-power monitoring device; controlling the low-power monitoring device to turn on the power switch of the video signal processing board, and after the video signal processing board is powered on, sending relevant operation instructions to the video signal processing board, so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions; when the video signal processing board does not need to perform real-time video monitoring processing, controlling the low-power monitoring device to turn off the power switch of the video signal processing board.

[0005] The present invention provides a remote wake-up device for a low-power monitoring device, comprising: a control unit, used for controlling the low-power monitoring device to perform power-on processing so that a mainboard module of the low-power monitoring device enters a low-power mode, wherein when the mainboard module is in the low-power mode, a video signal processing board is not powered; a receiving unit, used for receiving a remote signal sent by a server when the low-power monitoring device is in the low-power mode, and identifying whether the remote signal is a wake-up signal for waking up the low-power monitoring device, and if so, performing wake-up processing on the low-power monitoring device; a monitoring unit, used for controlling the low-power monitoring device to turn on a power switch of a video signal processing board, and after the video signal processing board is powered on, sending relevant operation instructions to the video signal processing board, so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions; and a shut-down unit, used for controlling the low-power monitoring device to turn off the power switch of the video signal processing board when the video signal processing board does not need to perform real-time video monitoring processing.

[0006] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0007] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0008] The remote wake-up method and device for low-power monitoring equipment provided by the present invention effectively reduces the overall power consumption of the equipment by flexibly managing the power status of each module in the low-power mode, while ensuring the responsiveness to real-time monitoring needs. The core of the method and device is to use the remote wake-up signal to trigger the equipment to switch from the low-power mode to the working mode, and to turn off the power switch of the video signal processing board in time after the monitoring task is completed, thereby avoiding unnecessary energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a schematic diagram of the steps of a remote wake-up method for a low-power monitoring device in one embodiment of the present invention; Figure 2 is a device-related schematic diagram of a low-power monitoring device in one embodiment of the present invention; Figure 3 is a cross-sectional schematic diagram of a low power consumption monitoring device in one embodiment of the present invention; Figure 4 It is a schematic diagram of the structure of a remote wake-up device for a low-power monitoring device in one embodiment of the present invention; Figure 5is a multi-angle view of a remote wake-up device for a low-power monitoring device in one embodiment of the present invention; Figure 6 is a partial exploded diagram of a remote wake-up device for a low-power monitoring device according to an embodiment of the present invention; Figure 7 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0010] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0012] Reference Figure 1 The embodiment of the present invention provides a remote wake-up method for a low-power monitoring device, comprising the following steps: S1, controlling the low power consumption monitoring device to perform power-on processing so that the mainboard module of the low power consumption monitoring device enters a low power consumption mode, wherein when the mainboard module is in the low power consumption mode, the video signal processing board is not powered.

[0013] S2, when the low-power monitoring device is in the low-power mode, receiving a remote signal sent by a server, and identifying whether the remote signal is a wake-up signal for waking up the low-power monitoring device, and if so, waking up the low-power monitoring device.

[0014] S3, controlling the low-power monitoring device to turn on the power switch of the video signal processing board, and after the video signal processing board is powered on, sending relevant operation instructions to the video signal processing board, so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions.

[0015] S4, when the video signal processing board does not need to perform real-time video monitoring processing, controlling the low-power monitoring device to turn off the power switch of the video signal processing board.

[0016] The embodiment of the present invention achieves the goal of effectively reducing the power consumption of the device while meeting the real-time monitoring requirements through the collaborative work of multiple steps. First, the low-power monitoring device is controlled to perform power-on processing so that the mainboard module enters a low-power mode. In this mode, the mainboard module can maintain basic communication capabilities, but the video signal processing board is not powered, thereby significantly reducing the standby power consumption of the device. For example, in a field monitoring scenario, the device maintains a low-power mode during non-monitoring periods, which can reduce energy consumption to a minimum while ensuring that it still has the ability to receive external commands. For example, the mainboard module can run using a low-power microprocessor, and the power supply status of different modules can be accurately controlled through an optimized power management chip.

[0017] When the device is in low-power mode, the server can wake up the device through a remote signal, which is one of the key links of this method. After receiving the remote signal sent by the server, the device needs to identify whether the signal is a wake-up signal. This step is usually implemented through a specific communication protocol, such as a customized data packet format, to ensure that only signals that meet the preset format and verification rules can trigger the wake-up operation. For example, when the server detects abnormal activity or receives a remote monitoring instruction from the user, it can send a signal containing a unique wake-up identifier to the device. After the device captures the signal, it uses the built-in verification algorithm to identify it and confirm the authenticity of the signal and the purpose of the wake-up. If the signal passes the verification, the device switches from low-power mode to working mode. This design not only improves the security of the device, but also ensures that the wake-up operation does not waste energy due to false triggering.

[0018] After the device is successfully awakened, the next step is to control the power supply status of the video signal processing board. The core is to turn on the power switch of the video signal processing board to put it into normal working state. To achieve this, an efficient power management circuit needs to be integrated inside the device to ensure that it can be powered on smoothly when the power is turned on, avoiding hardware damage or system instability that may be caused by current shock. For example, when the power switch is turned on, the mainboard module can send a startup signal to the video signal processing board, and gradually increase the voltage and current through the relevant power management module until the video processing board is fully started. Once the video signal processing board is powered on, the mainboard module will send corresponding operation instructions, such as starting video acquisition, encoding or transmission functions, so as to realize real-time video monitoring. This design ensures that the device can quickly enter the working state after receiving the wake-up signal, meeting the user's demand for real-time monitoring. For example, in a security scenario, when a user issues a monitoring request through a remote application, the device can start the camera and start transmitting the video stream within a few seconds to ensure that emergencies are handled in a timely manner.

[0019] The last step is that when the video signal processing board no longer needs to perform real-time video monitoring tasks, the device will control its power switch through the shutdown unit to restore to low-power mode. This process not only reduces the overall power consumption of the device, but also extends the battery life. In order to achieve this function, the device needs to have the ability to judge the task status in real time. For example, when the server detects that the monitoring task has been completed or the user closes the monitoring screen, the device will receive a task end signal through the communication module. According to the signal instruction, the mainboard module performs the shutdown operation, including stopping the acquisition and transmission of video data and disconnecting the power supply of the video signal processing board. This dynamic power management mechanism greatly improves the energy efficiency of the device and ensures that there will be no energy waste after the task is completed. For example, in an environmental monitoring scenario, when the device completes monitoring of a certain area, it can automatically shut down the video signal processing board and enter a low-power standby state, while the mainboard module continues to maintain the ability to monitor remote signals so that it can respond to new task instructions at any time.

[0020] Furthermore, since the device usually works in low-power mode, the received signal power is relatively low, which will cause the signal to attenuate during transmission due to distance and environmental influences. In addition, devices in low-power mode are more susceptible to external noise interference, which makes it more difficult to decode and analyze the received signal. Conventional data processing methods, such as traditional filtering techniques or simple threshold detection, often have difficulty ensuring the integrity and accuracy of the signal when faced with noise interference, which directly affects the recognition effect of the wake-up signal. Therefore, how to effectively receive and decode the wake-up signal in a low-power environment has become a daunting task.

[0021] Secondly, in order to save power, low-power devices usually adopt a periodic sleep and short-time reception working mode, that is, they monitor the wake-up signal by setting a "listening window". However, due to the short time interval of the listening window, the device needs to quickly determine whether the signal meets the wake-up condition within a limited time, which puts higher requirements on real-time processing capabilities. Especially when the network environment fluctuates greatly or there are noisy signals, the device may miss or misjudge due to insufficient processing speed or insufficient algorithm accuracy, which may cause the device to miss the wake-up signal and cause unnecessary power consumption due to misjudgment of the wake-up signal. Therefore, how to efficiently process the signal and accurately identify the wake-up condition within the limited listening window is another important issue in low-power remote wake-up technology.

[0022] In summary, how to effectively identify wake-up signals to ensure that the device responds in a timely manner while avoiding wasting energy due to misjudgment is also a core issue that needs to be urgently addressed in this field.

[0023] Based on this, in one example, when the low-power monitoring device is in the low-power mode, a remote signal sent by a server is received, and it is identified whether the remote signal is a wake-up signal for waking up the low-power monitoring device. If so, the low-power monitoring device is woken up, including: when the low-power monitoring device enters the wake-up listening mode, determining the remote signal received by the low-power monitoring device; determining the boundary frequency corresponding to the remote signal, and performing passband filtering on the remote signal based on the boundary frequency to obtain a pre-processed valid remote signal; using a sparse representation algorithm to convert the remote signal into a wake-up signal; An effective remote signal is decomposed into a small number of basis vectors containing only important features, and a sparse matrix is ​​constructed based on the small number of basis vectors; the sparse matrix is ​​reconstructed to obtain a reconstructed signal corresponding to the sparse matrix; wake-up feature points exceeding a threshold are detected in the reconstructed signal, and candidate wake-up signal segments are screened out in the reconstructed signal based on the wake-up feature points; the candidate wake-up signal segments are compared with multiple noise models in terms of spectral features and time domain features to identify and eliminate pseudo wake-up signal segments in the candidate wake-up signal segments; and the low-power monitoring device is waked up based on the retained wake-up signal.

[0024] The remote wake-up method for low-power monitoring equipment provided by the embodiment of the present invention effectively overcomes the problems of weak signal reception, large noise interference, short processing time, etc. in the low-power state through a series of signal processing steps. Its main implementation method involves the mutual cooperation of multiple links to accurately identify the remote wake-up signal while minimizing misjudgment.

[0025] When the low-power monitoring device enters the wake-up listening mode, the first thing to do is to determine the remote signal received by the device. This step ensures that the device can identify and obtain the external signal source in a low-power state. The reception of the remote signal is the starting point of the entire wake-up process, so accurately obtaining the signal is the key. Because the power of the signal received by the device in the low-power state is low, the signal is often mixed with high-frequency and low-frequency noise in the environment, which affects the decoding of the signal. Therefore, after receiving the signal, its boundary frequency must be determined as soon as possible. This boundary frequency is set according to the specific frequency range of the wake-up signal so that the effective frequency band of the wake-up signal can be accurately retained during filtering.

[0026] After the signal boundary frequency is determined, the device filters the signal through a bandpass filter, retaining only the effective frequency components within the frequency range of the wake-up signal, thereby filtering out noise that is not within the desired frequency band. For example, if the effective frequency band of the wake-up signal is between 300-800 Hz, the upper and lower boundary frequencies of the bandpass filter are set to 300 Hz and 800 Hz, so that the device can only pass the signal in this frequency band and exclude other high-frequency or low-frequency interference. This bandpass filtering process greatly reduces noise interference, retains the core characteristics of the wake-up signal, and makes the subsequent feature extraction and analysis of the signal more reliable.

[0027] The signal after bandpass filtering enters the feature extraction stage, and the sparse representation algorithm is used to decompose the effective signal into a small number of basis vectors. The role of the sparse representation algorithm is to simplify the complex signal structure into a sparse matrix containing only important features. The "important features" here refer to the features that can represent the core information of the wake-up signal. For example, if the wake-up signal shows a significant amplitude peak or frequency mutation at a specific moment, the sparse representation algorithm will use these peaks and frequency components as important features, thereby ignoring irrelevant or repeated signal components. By constructing a sparse matrix, the device only needs to process a small amount of core feature information, thereby achieving data dimensionality reduction. This processing method based on sparsity not only reduces the complexity of subsequent data processing, but also maintains the integrity of the signal at extremely low power consumption.

[0028] Next, the sparse matrix is ​​reconstructed to obtain the reconstructed signal. The reconstruction process relies on compressed sensing technology, which is a method that can reconstruct high-quality signals at low sampling rates. By downsampling and sparse solving the sparse matrix, the reconstructed signal restores the core features of the original signal. For example, if the sparse matrix of the wake-up signal contains several amplitude peaks and specific frequency points, compressed sensing technology will restore a reconstructed signal that is close to the original signal structure based on this sparse information. This reconstructed signal has a high degree of restoration and provides a clear data basis for the subsequent wake-up signal recognition of the device.

[0029] After the reconstructed signal is generated, the device detects the wake-up feature points in the reconstructed signal. These feature points are signal parts that exceed the set threshold, usually manifested as significant amplitude or frequency changes. Taking amplitude as an example, the device can set a threshold value to determine whether each signal data point exceeds the threshold. The data point that exceeds the threshold is the wake-up feature point. By indexing these feature points, the device selects candidate signal segments that meet the wake-up characteristics to form a set of candidate signal segments.

[0030] Before entering the wake-up analysis, the candidate signal segment set still needs to be further processed to eliminate pseudo wake-up signal segments. To this end, the device compares the candidate signal segments with the preset noise model library. The noise model library is composed of typical interference signals, which contains the spectral characteristics and time domain characteristics of common interference signals. By comparing the spectral characteristics and time domain characteristics of the candidate segments with the noise model, the device can filter out signal segments that are highly similar to the noise characteristics. For example, if the frequency distribution of a candidate segment is similar to the known background noise frequency distribution, the segment will be regarded as a pseudo signal and eliminated. Conversely, only those signal segments whose spectral characteristics and time domain characteristics are different from those of the noise model library will be retained as true wake-up signal segments.

[0031] Finally, based on these multiple screened real wake-up signal fragments, the device performs wake-up processing to switch the monitoring device from low-power mode to working state. Through this refined processing flow, the device can accurately identify the wake-up signal under low-power conditions, ensure the accuracy and timeliness of the wake-up, and reduce the waste of electricity caused by misjudgment. This technical solution is not only suitable for the wake-up of remote control monitoring equipment, but also can achieve effective signal recognition and response in other low-power application scenarios, providing a new technical path for realizing intelligent low-power remote control.

[0032] In one example, a boundary frequency corresponding to the remote signal is determined, and a passband filtering process is performed on the remote signal based on the boundary frequency to obtain a preprocessed effective remote signal, including: sampling the received remote signal to obtain a discrete signal sequence; analyzing the spectrum of the wake-up signal to determine the core frequency range of the wake-up signal, and setting the upper and lower boundary frequencies of the bandpass filter based on the core frequency range; calculating the filter coefficients of the bandpass filter according to the upper and lower boundary frequencies of the bandpass filter to obtain a filter coefficient sequence; and convolving the discrete signal sequence according to the filter coefficient sequence to obtain an effective remote signal after bandpass filtering.

[0033] In the example of the embodiment of the present invention, the boundary frequency determination and passband filtering preprocessing in the remote wake-up method are key steps to identify effective remote signals. First, the device receives the remote signal in low power mode. This signal usually has environmental noise and low signal power. Therefore, after receiving the signal, it is necessary to perform preliminary sampling processing on it to convert the analog signal into a discrete digital signal sequence. Through this process, a series of discrete sampling points are obtained to form a signal sequence that can be digitally processed. This signal sequence contains all the frequency components of the wake-up signal.

[0034] Next, in order to determine the core frequency range of the remote wake-up signal, the system needs to perform spectrum analysis on the discrete signal sequence. By studying the spectral characteristics of the signal, the core frequency components of the wake-up signal can be determined. Usually the wake-up signal will have a preset frequency range to distinguish it from the background noise. For example, if the frequency of the wake-up signal fluctuates between 300 Hz and 800 Hz, and the noise is mainly distributed in the frequency band below 300 Hz or above 800 Hz, then this frequency band can be set as the core frequency range of the wake-up signal. In order to ensure that the device only responds to this frequency band and filters out irrelevant signals in other frequency bands, the system needs to set the upper and lower boundary frequencies of the bandpass filter to 300 Hz and 800 Hz respectively. By clarifying the upper and lower frequency boundaries, the bandpass filter will only allow signals in the frequency band of 300 Hz to 800 Hz to pass through, while shielding other frequency signals outside this range, thereby reducing the interference of environmental noise on remote signal recognition.

[0035] After determining the upper and lower boundary frequencies of the bandpass filter, the next step is to calculate the filter coefficients of the bandpass filter based on these frequency boundaries. The calculation method of the filter coefficients is usually based on the window function method, especially for the finite impulse response (FIR) filter, which can obtain a more ideal frequency response within the set frequency range. For example, different types of window functions such as Hanning window or Kaiser window can be selected to enable the filter to pass frequency components within the specified frequency band while suppressing frequency components outside the frequency band. The calculated filter coefficient sequence will be used for subsequent filtering operations to ensure that the device can accurately extract the core features of the wake-up signal.

[0036] After calculating the filter coefficient sequence, the next step is to apply the coefficient sequence to the discrete signal sequence to achieve bandpass filtering. The filter coefficient sequence and the discrete signal sequence are multiplied point by point and summed through convolution operation. Each new output point represents the effective signal component that has been filtered in the frequency band of 300 Hz to 800 Hz. For example, in actual operation, if the discrete signal sequence y[n] represents the sampled signal, and the filter coefficient sequence h[n] represents the characteristics of the bandpass filter, when the two are convolved, each signal point y[n] is multiplied by the corresponding coefficient in the filter, and the calculation results are added in turn to obtain a new filtered signal point. The whole process effectively filters out irrelevant signals below 300 Hz and above 800 Hz through the superposition and integration of convolution, thereby obtaining an effective signal after bandpass filtering.

[0037] In the signal after bandpass filtering, only the signal components within the preset frequency range are retained, and this signal is called the "effective remote signal". The result of this bandpass filtering process realizes the frequency screening of the wake-up signal, removes a large amount of background noise, and obtains a relatively pure wake-up signal, which provides a reliable signal basis for subsequent sparse representation and feature extraction. In this way, the device can focus on processing the signal fragments containing the wake-up features without being distracted by irrelevant frequency interference and dispersing computing resources, thereby improving the signal processing efficiency of the system in a low-power state.

[0038] For example, if the wake-up signal of the remote monitoring device is received in an environment with severe factory noise, the unfiltered signal will contain a lot of low-frequency mechanical noise and high-frequency electromagnetic interference. By setting a bandpass filter from 300 Hz to 800 Hz, the low-frequency mechanical noise and high-frequency electromagnetic interference can be effectively filtered out, and only the wake-up signal within this range is retained. In the process of passing through the filter, even if the factory environment generates vibration noise with a frequency below 300 Hz or electromagnetic signals above 800 Hz, these interferences will be shielded during the filtering process, ensuring that the filtered signal can accurately represent the characteristics of the wake-up signal.

[0039] This bandpass filtering method provides a basic signal processing solution for remote wake-up of low-power monitoring devices. By gradually removing irrelevant noise, it significantly improves the device's recognition accuracy of the target signal and ensures that subsequent signal processing steps can function effectively in a low-power state. This process is not only applicable to a specific wake-up frequency range, but also can adjust the filter boundary frequency according to different application requirements, thereby flexibly adapting to different wake-up conditions and environmental noise conditions.

[0040] In one example, a sparse representation algorithm is used to decompose the effective long-range signal into a small number of basis vectors containing only important features, and a sparse matrix is ​​constructed based on the small number of basis vectors, including: dividing the effective long-range signal into a plurality of signal segments of equal length to form a set of signal segments of equal length; using a basis dictionary to represent each of the signal segments of equal length as a sparse coefficient vector, wherein the sparse coefficient vector includes only a small number of non-zero values; merging the sparse coefficient vectors of all the signal segments of equal length column by column to form a sparse coefficient basic matrix; setting a threshold for small non-zero elements in the sparse coefficient basic matrix to perform cropping to obtain a cropped sparse coefficient matrix; and reconstructing the signal segments of equal length using the cropped sparse coefficient matrix and a basis dictionary to obtain a final sparse matrix.

[0041] In the specific implementation of the embodiment of the present invention, in order to effectively identify important features in the wake-up signal and eliminate redundant information, a sparse representation algorithm is used to decompose and sparsely process the effective remote signal after filtering. First, the effective remote signal is divided into multiple signal segments of equal length for further analysis and sparse decomposition based on the local characteristics of the signal. Each signal segment of equal length represents a small segment of the signal, so that the device can accurately analyze and extract signal features at the segment scale. This division step not only simplifies the local processing of the signal, but also facilitates real-time feature extraction under limited computing resources.

[0042] Next, the system builds a sparse representation for each signal segment. By selecting an appropriate basis dictionary, each signal segment can be represented as a sparse coefficient vector. This basis dictionary consists of a set of predefined basis vectors, each of which represents a different signal pattern. The core of sparse representation is to find a sparse coefficient vector so that the signal segment can be approximately represented by a linear combination of a small number of basis vectors. In other words, for each signal segment, we represent it as a sparse coefficient vector through sparse decomposition, which contains only a small number of non-zero elements. For example, assuming that each signal segment contains several features of amplitude mutation or frequency jump, then these significant features will be represented by a small number of basis vectors in the basis dictionary, while other irrelevant information will be discarded.

[0043] After obtaining the sparse coefficient vector of each signal segment, the sparse coefficient vectors of all signal segments are merged column by column to form a sparse coefficient basic matrix. The basic matrix records the sparse representation of each signal segment, so that the entire signal can be simplified into a set of a small number of important features. Since most of the elements in the sparse matrix are close to zero, this sparse characteristic greatly reduces the storage and processing requirements of the data. In order to further streamline the amount of data in the sparse matrix and improve its representation efficiency, the system introduces threshold clipping processing in the sparse coefficient basic matrix. Specifically, for small non-zero elements in the sparse coefficient matrix, a threshold is set, and those non-zero elements less than the threshold are set to zero, and only sparse coefficients above the threshold are retained. This clipping operation can effectively eliminate noise and irrelevant information, and the retained sparse coefficients are closer to the core features of the wake-up signal.

[0044] After completing the threshold clipping, we get the clipped sparse coefficient matrix, which contains the main features of the wake-up signal and has eliminated as much useless data as possible. Based on the sparse matrix, the system uses the clipped sparse coefficient matrix and the base dictionary to reconstruct the signal segments of equal length in order to recover the signal containing important features. Through the reconstruction of the sparse matrix, the system obtains a set of feature matrices that are similar to the original signal but with a greatly reduced amount of data. This sparse feature matrix retains the main information of the wake-up signal and provides reliable data support for the wake-up recognition of the device in a low-power state.

[0045] For example, assume that the effective remote signal is divided into several signal segments of equal length after filtering, and assume that one of the segments contains significant features in the wake-up signal, such as the amplitude mutation point or frequency jump point of the signal. In the sparse representation process, these significant feature points will be identified by a few basis vectors in the basis dictionary and represented as a small number of non-zero elements in the sparse coefficient vector. Through the clipping process, the small non-zero coefficients below the threshold are set to zero, the noise components are removed, and the retained sparse matrix only contains the core feature information of the wake-up signal. Finally, by reconstructing the matrix, the sparse feature matrix obtained by the device not only represents the important components of the signal, but also greatly reduces data redundancy, so that the wake-up signal can be quickly and accurately identified in low power mode. This sparse signal representation method effectively reduces the computational complexity and provides a reliable signal basis for subsequent wake-up processing and noise removal.

[0046] In one example, the sparse matrix is ​​reconstructed to obtain a reconstructed signal corresponding to the sparse matrix, including: constructing a compressed sensing observation matrix according to the structure of the sparse matrix and the sparsity characteristics of the wake-up signal; using the observation matrix to downsample the sparse matrix to obtain downsampled data; using the L1 norm minimization algorithm to perform sparse solution on the downsampled data to solve the reconstructed sparse matrix; using the basis dictionary (such as the scattered cosine transform basis dictionary) to perform linear combination processing on the reconstructed sparse matrix to obtain an approximate reconstructed signal matrix; performing difference calculation on the approximate reconstructed signal matrix and the downsampled data to obtain an error matrix, wherein the error matrix is ​​used to evaluate the accuracy of compressed sensing reconstruction and the signal restoration quality; performing error correction on the approximate reconstructed signal matrix according to the error matrix to obtain a corrected reconstructed signal matrix; extracting key features from the reconstructed signal matrix, and combining the extracted key features into a final reconstructed signal.

[0047] In an embodiment of the present invention, the reconstruction process of the sparse matrix is ​​a key step. Through a series of orderly data processing processes, the sparse matrix is ​​converted into a usable reconstructed signal to achieve effective restoration of the wake-up signal. In this example, it is first necessary to construct a compressed sensing observation matrix based on the structure of the sparse matrix and the sparsity characteristics of the wake-up signal. The construction of the observation matrix is ​​based on the theory of compressed sensing. The purpose is to retain the main features of the signal while minimizing the amount of data through sparse sampling methods. To this end, the observation matrix can be constructed using a Gaussian random matrix or a partial discrete Fourier transform matrix to ensure that the signal can still maintain a high reconstruction quality at a low sampling rate. The design of the observation matrix is ​​very critical because it directly affects the accuracy of subsequent data reconstruction.

[0048] With the observation matrix, the system can use it to downsample the sparse matrix to obtain downsampled data. Downsampling the sparse matrix through the observation matrix means retaining only a small but important sample point in the signal, which greatly reduces the amount of data while retaining the main features of the signal. For example, assuming that the wake-up signal represented by the sparse matrix is ​​sparse, after downsampling, only a small number of key samples need to be processed without completely restoring the entire signal, which greatly reduces the demand for computing resources.

[0049] After the downsampled data is acquired, the next step is to use the L1 norm minimization algorithm to perform sparse solution on the downsampled data in order to solve the reconstructed sparse matrix. The L1 norm minimization algorithm, such as orthogonal matching pursuit (OMP) or least absolute shrinkage selection operator (LASSO), is used to reconstruct the signal in this step, and the signal is effectively restored by finding the sparsest solution. The algorithm processes each sample of the downsampled data to ensure that the restored sparse matrix retains the key features of the signal. In this process, the constraint of the L1 norm ensures that the sparse matrix obtained by the solution has as many zero elements as possible, reducing unnecessary data redundancy while accurately restoring the important information of the signal.

[0050] Subsequently, the reconstructed sparse matrix is ​​linearly combined using a basis dictionary (such as a scattered cosine transform basis dictionary) to obtain an approximate reconstructed signal matrix. The basis dictionary plays a core role in signal restoration at this stage. Through linear combination, the non-zero elements of the sparse matrix are multiplied and accumulated with the basis vectors of the basis dictionary to form a reconstructed signal matrix that is close to the original signal. Although the approximate reconstructed signal matrix after such processing retains most of the characteristics of the wake-up signal, it may still have a certain gap with the original signal. Therefore, in order to further improve the quality of signal restoration, it is necessary to perform error evaluation on the approximate reconstructed signal matrix.

[0051] Error assessment is completed by calculating the difference between the approximate reconstructed signal matrix and the downsampled data. Specifically, the downsampled data is compared with the approximate reconstructed signal one by one, and the difference of each sample is calculated to obtain the error matrix. The error matrix contains the deviation information between the reconstructed signal and the downsampled signal at each point, which represents the restoration accuracy of the current reconstructed signal. By analyzing the error matrix, we can understand the error distribution between the approximate reconstructed signal and the original signal, which provides a basis for subsequent error correction.

[0052] Next, the system uses the error matrix to correct the error of the approximate reconstructed signal matrix, thereby obtaining the corrected reconstructed signal matrix. The error correction process is achieved by adjusting the values ​​in the error matrix with the corresponding points of the approximate reconstructed signal to reduce the deviation and improve the accuracy of signal reconstruction. For example, if the error value of some points in the error matrix is ​​large, the value of the point in the reconstructed signal matrix is ​​increased or decreased accordingly during correction, so that the corrected reconstructed signal matrix is ​​closer to the original wake-up signal.

[0053] The last step is to extract key features from the corrected reconstructed signal matrix, and then combine the extracted key features into the final reconstructed signal. Key feature extraction mainly focuses on the core parts of the signal, such as amplitude peaks, frequency components, etc., which can provide the core information of the wake-up signal. In the process of reconstructing the signal, only the most representative characteristic components are retained, which can reduce data redundancy and improve the wake-up response efficiency of the device. Taking a specific example, assuming that the corrected reconstructed signal contains multiple frequency components and different amplitude peaks, the key feature extraction process will only retain those components with higher amplitudes and frequencies that meet the characteristics of the wake-up signal. The final combined reconstructed signal therefore only contains the most important features of the wake-up signal, thereby achieving efficient and accurate remote wake-up signal recognition under low power consumption conditions.

[0054] In one example, a wake-up feature point exceeding a threshold is detected in the reconstructed signal, and candidate wake-up signal segments are screened out in the reconstructed signal based on the wake-up feature point, including: performing signal feature analysis on the reconstructed signal according to a preset feature of the wake-up signal to obtain a signal amplitude sequence and a signal frequency sequence; calculating an average amplitude value and a standard deviation of the signal amplitude sequence, and taking the sum of the average amplitude value and the standard deviation as a first threshold value of the wake-up signal; traversing the signal amplitude sequence to determine whether each data point exceeds the first threshold value, recording the data points exceeding the first threshold value, and indexing them to a first feature point index set; segmenting the reconstructed signal according to the first feature point index set to obtain a set of candidate signal segments; performing a length analysis on each segment in the set of candidate signal segments to screen out segments that meet a preset time length range to obtain a set of valid candidate signal segments; performing a frequency analysis on each signal segment in the set of valid candidate signal segments to determine a signal frequency sequence corresponding to each signal segment; screening out segments whose signal frequency sequences meet a preset wake-up frequency range to obtain candidate wake-up signal segments.

[0055] In an example of an embodiment of the present invention, in order to accurately select candidate signal segments with wake-up features from the reconstructed signal, the signal feature analysis of the reconstructed signal is first performed. According to the preset features of the wake-up signal, the amplitude and frequency features of the reconstructed signal are extracted to obtain a signal amplitude sequence and a signal frequency sequence, respectively. These two sequences can effectively characterize the intensity and frequency distribution of the reconstructed signal, and provide basic data for subsequent threshold detection and feature screening.

[0056] Once the amplitude sequence of the signal is obtained, the mean and standard deviation of the amplitude sequence are calculated. These two values ​​are used to determine the first threshold value. Specifically, the first threshold value is obtained by adding the average amplitude value to the standard deviation. This threshold value reflects the typical characteristics of the wake-up signal. In practical applications, it can avoid misjudgment caused by accidental amplitude fluctuations. The first threshold value is set to distinguish between background noise and possible wake-up signals. Only those points with amplitudes higher than the threshold value are regarded as potential wake-up signal feature points. Next, the system traverses each data point in the signal amplitude sequence, records the data points that exceed the first threshold value, and stores their indexes in the first feature point index set. In this way, the first feature point index set contains all points that may contain wake-up features, laying the foundation for the next step of segment division.

[0057] Based on the first feature point index set, the reconstructed signal is segmented, and points that continuously exceed the threshold value are classified as a signal segment to form a candidate signal segment set. The candidate signal segment set is a preliminary screening of the reconstructed signal, removing the signal parts that do not meet the threshold requirements and retaining only the signal segments that may contain wake-up information. This step of segmentation enables the system to effectively focus on the time region of potential wake-up signals, thus providing a convenient starting point for further screening and analysis.

[0058] For each segment in the set of candidate signal segments, the system will then perform a length analysis. This step is to ensure that the duration of the candidate segment meets the preset wake-up signal duration range, thereby avoiding misjudgment and redundancy. For example, if the typical duration of the wake-up signal is between 100 milliseconds and 200 milliseconds, those segments that are lower or higher than this range will be regarded as invalid candidate segments and eliminated, ensuring that only signal segments with a duration within this range are retained. Through this time-length screening process, the resulting set of valid candidate signal segments only contains those segments that may meet the wake-up characteristics, thereby further optimizing the accuracy of the signal.

[0059] After obtaining a set of valid candidate signal fragments, the frequency distribution of each fragment needs to be analyzed. By performing frequency analysis on each fragment in the set of valid candidate signal fragments, the signal frequency sequence corresponding to each fragment can be obtained. This frequency sequence can reveal the frequency characteristics of the signal fragments, thereby helping to further confirm whether these fragments have the spectral characteristics of the wake-up signal. Next, the system compares the signal frequency sequence with the preset wake-up frequency range to filter out the fragments that meet the wake-up frequency range. For example, if the frequency range of the preset wake-up signal is between 300 Hz and 800 Hz, the system will only retain signal fragments that meet this frequency range, and those fragments whose frequency components are mainly distributed below 300 Hz or above 800 Hz will be eliminated. This step of screening ensures that the candidate wake-up signal fragments meet the preset requirements in terms of spectrum, thereby further reducing the possibility of misjudgment.

[0060] Finally, through the above multi-step screening process, the candidate wake-up signal segments obtained are those signal segments that meet the preset requirements of the wake-up signal in terms of amplitude, duration and frequency characteristics. These segments can significantly represent the core characteristics of the wake-up signal, ensuring that the device can respond quickly when receiving a valid wake-up signal. In practical applications, for example, in an industrial environment, the monitoring device receives a wake-up signal in a large amount of background noise, and initially selects points with significant fluctuations through amplitude detection, and then identifies segments that meet the wake-up conditions through duration and frequency characteristics. This series of processing can help the device accurately identify valid signals and ensure efficient response to wake-up signals in a low-power state.

[0061] In one example, the candidate wake-up signal segment is compared with multiple noise models in terms of spectral features and time domain features to identify and eliminate pseudo wake-up signal segments in the candidate wake-up signal segment, including: obtaining historical remote signals received by the low-power monitoring device, and extracting multiple interference signals from the historical remote signals; performing spectral analysis on the multiple interference signals to obtain a typical noise spectral feature set; performing cluster analysis on the typical noise spectral feature set to identify noise categories with similar spectral features, and grouping similar noise spectral features into a group to produce a noise model library; performing fast Fourier transform on the candidate wake-up signal segment; The spectral features are extracted by performing conversion processing to obtain a spectral feature set of candidate wake-up signal segments; the spectral feature set and the noise model library are compared for similarity using a cosine similarity algorithm to obtain a similarity score for each candidate wake-up signal segment; the candidate wake-up signal segments are screened according to the similarity score to obtain a set of candidate signal segments with low similarity; each candidate signal segment in the candidate signal segment set is analyzed in the time domain to generate a time feature set; the time feature set is compared with the time domain features of the noise model library, and the segments that do not meet the noise features are retained to obtain a final set of true wake-up signal segments.

[0062] In the implementation example of the present invention, in order to ensure that the low-power monitoring device can accurately distinguish between the real wake-up signal and the noise interference when receiving the remote signal, the system will compare the candidate wake-up signal fragments with multiple noise models, and gradually eliminate the false wake-up signals. First, the system obtains the historical remote signal data received by the low-power monitoring device, and extracts multiple interference signals from it. The historical records of these interference signals can reflect the common background noise characteristics of the device, such as electromagnetic interference from power lines, low-frequency noise from environmental mechanical vibrations, and high-frequency radio interference. The acquisition and extraction of these noise signals provides a rich data foundation for the system to build a noise model library.

[0063] After extracting multiple interference signals, the system performs spectrum analysis on these signals to generate a typical noise spectrum feature set. Spectral analysis can reveal the frequency distribution characteristics of noise signals. For example, some electromagnetic noise has high intensity at specific frequencies, while mechanical vibration noise is concentrated in a lower frequency range. This spectrum feature set provides a detailed frequency basis for identifying different types of noise. Next, the system performs cluster analysis on the typical noise spectrum feature set to identify noise categories with similar spectrum characteristics. For example, electromagnetic noise with similar frequency distribution will be classified into one category, while mechanical noise will be classified into another category. Through this clustering method, the system can group and classify similar noise spectrum features, and finally generate a noise model library. The noise model in each category represents a specific type of noise characteristics, providing a data basis for subsequent noise comparison.

[0064] After building the noise model library, the system performs fast Fourier transform (FFT) processing on the candidate wake-up signal fragments to extract the spectral features of the candidate wake-up signal fragments and form a spectral feature set of the candidate fragments. This set contains the distribution of each candidate signal fragment at different frequencies and is the main feature description of the wake-up signal fragment in the frequency domain. Then, the system uses the cosine similarity algorithm to compare the spectral feature set of the candidate signal fragments with the noise model library. By calculating the similarity score between each candidate signal fragment and different noise categories in the noise model library, the similarity of the candidate signal fragments in spectral features is obtained. The higher the similarity score of the fragment, the more similar its spectral features are to a certain type of noise. Conversely, the fragment with a lower similarity score is more likely to be a valid wake-up signal. For example, if the amplitude of the spectral distribution of a candidate signal fragment in a certain frequency band is similar to the electromagnetic interference noise in the noise model library, the similarity score of the candidate signal fragment will be high, and it will be regarded as a false wake-up signal.

[0065] The system screens the candidate segments based on their similarity scores and retains the segments with lower similarity. These low-similarity segments may contain real wake-up signals. Similarity screening can significantly reduce the probability of misjudgment of noise, thereby focusing on the most likely wake-up signal segments. For low-similarity candidate signal segment sets, the system further performs time domain analysis on each segment, extracts time features and generates a time feature set. These time features describe the change patterns of the candidate signal segments in the time domain, such as the duration of the amplitude, sudden amplitude changes, etc. In real wake-up signals, there are usually characteristic patterns with specific durations or mutations, while noise signals may appear as random changes or continuous interference.

[0066] Finally, the system compares the time feature set with the time domain features in the noise model library and removes the segments that match the noise features. Through dual screening in the frequency domain and time domain, the segments that do not match the noise model features are retained to form the final set of true wake-up signal segments. For example, if the spectral features of a candidate segment are lower than the similarity threshold of a noise category and its time features meet the preset duration of the wake-up signal, then the segment will be considered a valid wake-up signal and will eventually be retained.

[0067] In order to facilitate understanding of various situations in practical applications of the embodiments of the present invention, an embodiment is now provided for understanding: The low-power monitoring equipment is designed to meet the high standards of intelligent monitoring in subway security areas, while taking into account the energy saving and long battery life of the equipment. Figure 2 , Figure 4 and Figure 5 As shown in the figure, the core configuration of the device includes a 4-megapixel, 1 / 1.8 CMOS starlight-level image sensor, which supports ultra-low illumination shooting, with a minimum illumination of 0.001 Lux in color mode and 0.0005 Lux in black and white mode, ensuring excellent imaging effects under low light conditions. Its lens supports 4x optical zoom and is compatible with H.264, H.265 and MJPEG video encoding standards. It also supports common video transmission protocols such as GB28181 and rtmp, which can meet the needs of various video applications. In addition, the device supports OSD text information overlay, with a resolution of up to 1920P, and has a wide dynamic mode to provide better picture quality for low-light environments.

[0068] like Figure 3 The cross-sectional schematic diagram of the low power consumption monitoring device shown is as follows Figure 2 The low power monitoring device is shown in a cross-sectional view along the AA section.

[0069] In terms of storage, the device supports TF card expansion storage up to 256G, which can store snapshots for more than 60 days. It also has an IP66 protection level and is suitable for harsh environments ranging from -25°C to 85°C. The main control chip uses Huawei HiSilicon processor, which has powerful and stable performance. Combined with the 4G wireless communication module, it can realize the disconnection reconnection function to ensure the reliability of real-time communication. The device is also equipped with a hardware watchdog mechanism, which can automatically recover in the event of a system crash, further improving the overall reliability.

[0070] In terms of energy management, the equipment achieves efficient energy conversion through a solar power supply system. The output power of the monocrystalline solar panel does not exceed 10W. Combined with a 20-40Ah lithium battery pack and a charging controller with a temperature compensation control algorithm, it can ensure that low-point monitoring equipment can work continuously for more than 20 days on rainy days, and high-point monitoring equipment can work for more than 7 days. The power control system provides stable voltage and current output. The overall size of the equipment is strictly controlled within 300300300mm, and the weight does not exceed 5kg, which is convenient for installation on the top of the customized warning pile.

[0071] The functional design of the device is very comprehensive. First of all, the device supports the timed snapshot function, which can capture the scene on time and transmit it back to the back-end server in real time according to the needs of the subway security zone. Secondly, it is seamlessly integrated with the existing Shenzhen Metro Security Zone information system, and the device realizes real-time remote viewing and operation on the PC and APP sides. After the built-in GPS module is integrated with GIS information, the device can display the distribution of monitoring points in the form of a map, and supports status management. When the device is offline or fails, the system will automatically warn. For low-point monitoring equipment, a sleep function is designed to turn off the power of the video signal processing board during non-working hours to extend the standby time. The device can also receive remote wake-up commands through the 4G module to activate the device. It only takes 5-8 seconds to complete the entire process from power-on, power supply to entering real-time video.

[0072] The device supports automatic zoom capture, and can dynamically adjust the focus in real-time preview to ensure that details within the monitoring range are clearly presented. In terms of storage, each device is equipped with a TF card of 64G or more, and the captured pictures can be saved locally and in the cloud for more than 60 days, which is convenient for retrieval at any time, and there is no need for video recording storage. The overall design balances high performance and low power consumption, providing a set of intelligent, efficient and long-lasting monitoring solutions for subway security areas, with extremely high application value and promotion potential.

[0073] like Figure 6 As shown, an embodiment of the present invention further provides a remote wake-up device for a low-power monitoring device, comprising: A control unit 1, used for controlling the low-power monitoring device to perform power-on processing so that the mainboard module of the low-power monitoring device enters a low-power mode, wherein when the mainboard module is in the low-power mode, the video signal processing board is not powered; A receiving unit 2 is used to receive a remote signal sent by a server when the low-power consumption monitoring device is in the low-power consumption mode, and identify whether the remote signal is a wake-up signal for waking up the low-power consumption monitoring device, and if so, perform a wake-up process on the low-power consumption monitoring device; The monitoring unit 3 is used to control the low-power monitoring device to turn on the power switch of the video signal processing board, and after the video signal processing board is powered on, send relevant operation instructions to the video signal processing board so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions; The shut-down unit 4 is used to control the low-power monitoring device to shut down the power switch of the video signal processing board when the video signal processing board does not need to perform real-time video monitoring processing.

[0074] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0075] Reference Figure 7 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0076] Those skilled in the art will understand that Figure 7 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0077] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0078] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0079] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0080] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A remote wake-up method for a low-power monitoring device, characterized in that: The following steps are involved: Controlling the low-power monitoring device to perform power-on processing so that the mainboard module of the low-power monitoring device enters a low-power mode, wherein when the mainboard module is in the low-power mode, the video signal processing board is not powered; When the low-power monitoring device is in the low-power mode, receiving a remote signal sent by a server, and identifying whether the remote signal is a wake-up signal for waking up the low-power monitoring device, and if so, performing a wake-up process on the low-power monitoring device; Controlling the low-power monitoring device to turn on the power switch of the video signal processing board, and after the video signal processing board is powered on, sending relevant operation instructions to the video signal processing board, so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions; When the video signal processing board does not need to perform real-time video monitoring processing, the low-power monitoring device is controlled to turn off the power switch of the video signal processing board.

2. The remote wake-up method according to claim 1, characterized in that: When the low-power monitoring device is in the low-power mode, a remote signal sent by a server is received, and whether the remote signal is a wake-up signal for waking up the low-power monitoring device is identified, and if so, a wake-up process is performed on the low-power monitoring device, including: In a case where the low power consumption monitoring device enters a wake-up listening mode, determining a remote signal received by the low power consumption monitoring device; Determine a boundary frequency corresponding to the remote signal, and perform passband filtering on the remote signal based on the boundary frequency to obtain a preprocessed effective remote signal; Decomposing the effective long-range signal into a small number of basis vectors containing only important features by using a sparse representation algorithm, and constructing a sparse matrix based on the small number of basis vectors; Reconstructing the sparse matrix to obtain a reconstructed signal corresponding to the sparse matrix; detecting a wake-up feature point exceeding a threshold in the reconstructed signal, and screening out a candidate wake-up signal segment in the reconstructed signal based on the wake-up feature point; performing spectrum feature and time domain feature comparison processing on the candidate wake-up signal segments and multiple noise models, identifying and eliminating pseudo wake-up signal segments in the candidate wake-up signal segments; and, The low power consumption monitoring device is woken up based on the reserved wake-up signal.

3. The remote wake-up method according to claim 2, characterized in that: Determining a boundary frequency corresponding to the remote signal, and performing passband filtering on the remote signal based on the boundary frequency to obtain a preprocessed effective remote signal, including: Sampling the received remote signal to obtain a discrete signal sequence; Analyze the frequency spectrum of the wake-up signal, determine the core frequency range of the wake-up signal, and set the upper and lower boundary frequencies of the bandpass filter based on the core frequency range; Calculating the filter coefficients of the bandpass filter according to the upper and lower boundary frequencies of the bandpass filter to obtain a filter coefficient sequence; The discrete signal sequence is convolved according to the filter coefficient sequence to obtain an effective long-range signal after bandpass filtering.

4. The remote wake-up method according to claim 2, characterized in that: The effective long-range signal is decomposed into a small number of basis vectors containing only important features by using a sparse representation algorithm, and a sparse matrix is ​​constructed based on the small number of basis vectors, including: Dividing the valid long-distance signal into a plurality of signal segments of equal length to form a set of signal segments of equal length; Using a base dictionary, each of the signal segments of equal length is represented as a sparse coefficient vector, wherein the sparse coefficient vector includes only a small number of non-zero values; Merging the sparse coefficient vectors of all the signal segments of equal length column by column to form a sparse coefficient basic matrix; For small non-zero elements in the sparse coefficient basic matrix, a threshold is set to perform clipping processing to obtain a clipped sparse coefficient matrix; The signal segments of equal length are reconstructed using the pruned sparse coefficient matrix and the base dictionary to obtain a final sparse matrix.

5. The remote wake-up method according to claim 4, characterized in that: Reconstructing the sparse matrix to obtain a reconstructed signal corresponding to the sparse matrix includes: Constructing a compressed sensing observation matrix according to the structure of the sparse matrix and the sparsity characteristics of the wake-up signal; Downsampling the sparse matrix using the observation matrix to obtain downsampled data; Using an L1 norm minimization algorithm, sparsely solve the downsampled data to obtain a reconstructed sparse matrix; Using the base dictionary to perform linear combination processing on the reconstructed sparse matrix to obtain an approximate reconstructed signal matrix; Performing difference calculation between the approximate reconstructed signal matrix and the downsampled data to obtain an error matrix, wherein the error matrix is ​​used to evaluate the accuracy of compressed sensing reconstruction and the quality of signal restoration; Performing error correction on the approximate reconstructed signal matrix according to the error matrix to obtain a corrected reconstructed signal matrix; Key features are extracted from the reconstructed signal matrix, and the extracted key features are combined into a final reconstructed signal.

6. The remote wake-up method according to claim 2, characterized in that: Detecting a wake-up feature point exceeding a threshold in the reconstructed signal, and screening out a candidate wake-up signal segment in the reconstructed signal based on the wake-up feature point, including: According to the preset characteristics of the wake-up signal, the reconstructed signal is subjected to signal characteristic analysis to obtain a signal amplitude sequence and a signal frequency sequence; Calculating an average amplitude value and a standard deviation of the signal amplitude sequence, and taking the sum of the average amplitude value and the standard deviation as a first threshold value of the wake-up signal; Traversing the signal amplitude sequence, determining whether each data point exceeds the first threshold value, recording the data points exceeding the first threshold value, and indexing them into the first feature point index set; Divide the reconstructed signal into segments according to the first feature point index set to obtain a set of candidate signal segments; Performing a length analysis on each segment in the candidate signal segment set, screening out segments that meet a preset time range, and obtaining a valid candidate signal segment set; Performing frequency analysis on each signal segment in the valid candidate signal segment set to determine a signal frequency sequence corresponding to each signal segment; The segments of the signal frequency sequence that meet the preset wake-up frequency range are screened out to obtain candidate wake-up signal segments.

7. The remote wake-up method according to claim 2, characterized in that: Performing spectrum feature and time domain feature comparison processing on the candidate wake-up signal segment and multiple noise models, identifying and eliminating pseudo wake-up signal segments in the candidate wake-up signal segment, includes: Acquire a historical remote signal received by the low-power monitoring device, and extract a plurality of interference signals from the historical remote signal; Performing spectrum analysis on the multiple interference signals to obtain a typical noise spectrum feature set; Performing cluster analysis on the typical noise spectrum feature set to identify noise categories with similar spectrum features, and grouping similar noise spectrum features into one group to generate a noise model library; Performing fast Fourier transform processing on the candidate wake-up signal segment to extract spectrum features and obtain a spectrum feature set of the candidate wake-up signal segment; Using a cosine similarity algorithm, the spectrum feature set and the noise model library are compared for similarity to obtain a similarity score for each candidate wake-up signal segment; Screening the candidate wake-up signal segments according to the similarity scores to obtain a set of candidate signal segments with low similarity; Performing time domain analysis on each candidate signal segment in the candidate signal segment set to generate a time feature set; The time feature set is compared with the time domain features of the noise model library, and the segments that do not meet the noise features are retained to obtain a final set of true wake-up signal segments.

8. A remote wake-up device for a low-power monitoring device, characterized in that: include: A control unit, used for controlling the low-power monitoring device to perform power-on processing so that the mainboard module of the low-power monitoring device enters a low-power mode, wherein when the mainboard module is in the low-power mode, the video signal processing board is not powered; a receiving unit, configured to receive a remote signal sent by a server when the low-power consumption monitoring device is in the low-power consumption mode, and identify whether the remote signal is a wake-up signal for waking up the low-power consumption monitoring device, and if so, perform a wake-up process on the low-power consumption monitoring device; The monitoring unit is used to control the low-power monitoring device to turn on the power switch of the video signal processing board, and after the video signal processing board is powered on, send relevant operation instructions to the video signal processing board so that the video signal processing board performs real-time video monitoring processing based on the relevant operation instructions; The shut-down unit is used to control the low-power monitoring device to shut down the power switch of the video signal processing board when the video signal processing board does not need to perform real-time video monitoring processing.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images