Adaptive multi-source level wake-up method, system and medium based on hisilicon platform usb

Abstract

This invention provides an adaptive multi-source level wake-up method, system, and medium based on the HiSilicon platform USB. It synchronously captures the first level information, first square wave information, and first voltage information of the USB Hub, and generates a time-aligned trigger signal matrix through delay compensation. Based on the trigger signal matrix, feature signals are extracted, and a wake-up score is generated using a weighted fusion algorithm. An adaptive wake-up threshold is dynamically calculated based on a real-time noise baseline and historical false wake-up records. When only the first slope information exceeds the limit, it is determined to be a Level 1 wake-up, and the wake-up score is increased. When both the first slope information and the first frequency information exceed the limit, it is determined to be a Level 2 wake-up, and a wake-up action is triggered. When the first slope information does not exceed the limit and the wake-up score exceeds the wake-up threshold, it is determined to be a Level 3 wake-up, triggering a wake-up action and initiating wake-up diagnosis. Noise interference is reduced through multi-source signal cross-validation, and a dynamic threshold algorithm is used to adapt to complex industrial environments, improving the success rate and reliability of wake-up.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of level wake-up, more particularly, to a self-adaptive multi-source level wake-up method, system and medium based on HiSilicon platform USB. BACKGROUND

[0002] In the field of industrial control and embedded systems, HiSilicon platform is widely used due to its high performance and low power consumption. However, in the USB2.0 wake-up circuit of this platform, there is a significant technical bottleneck. The CPU does not support native USB 2.0 device wake-up function in S3 (suspend to RAM) and S4 (suspend to disk) deep sleep states, including keyboard or mouse, which requires an external USB_Hub to indirectly achieve the purpose of wake-up. The traditional solution relies on single signal detection, which has serious defects in complex electromagnetic environments. Single PGANG pin level or DRV pin square wave detection is easily disturbed by environmental noise, leading to high false wake-up rate. In addition, in industrial scenarios, voltage fluctuations, power frequency interference, etc. cause signal distortion, and single threshold detection has failure phenomenon.

[0003] Therefore, there is an urgent need for a self-adaptive multi-source composite level wake-up technology based on HiSilicon platform USB.

[0004] Wake-up: refers to the process of waking up a computer system from a sleep state through specific devices or events.

[0005] ‌S3 (suspend to RAM): In this mode, all devices except memory stop working, but the data in memory is saved to the hard disk. The system can quickly recover from S3 state because the data in memory is not cleared, and only needs to load the saved data.

[0006] ‌S4 (suspend to disk): In this mode, the data in memory is saved to the hard disk, and all devices stop working. When the system recovers from S4 state, it needs to read data from the hard disk to the memory, so the recovery time is longer.

[0007] EC: Full name is Embedded Controller, embedded controller. It is actually a single-chip microcomputer commonly used in laptops for keyboard control, touchpad, power management, fan control, notebook battery management, etc. SUMMARY

[0008] In view of the above problems, the purpose of the present application is to provide a HiSilicon platform USB-based adaptive multi-source level wake-up method, system and medium, which synchronously analyzes the PGANG level, DRV square wave and VSENSE voltage stroke multi-source composite level; extracts the characteristic signal and generates a wake-up score by combining a preset weighting algorithm; dynamically adjusts the threshold based on real-time noise and historical false positive rate; triggers a three-level wake-up mechanism according to the composite signal and the wake-up score; and reduces noise interference through multi-source signal cross-validation, adapts to complex industrial environments by using a dynamic threshold algorithm, and improves the success rate and reliability of wake-up.

[0009] The first aspect of the present application provides a HiSilicon platform USB-based adaptive multi-source level wake-up method, which comprises:

[0010] Parallelly collecting a trigger signal, including first level information, first square wave information and first voltage information;

[0011] Based on a preset delay compensation, a time-aligned trigger signal matrix is generated according to the trigger signal;

[0012] According to the trigger signal matrix, a characteristic signal is extracted, including first slope information, first frequency information and first energy information;

[0013] Based on a preset weighting fusion algorithm, a wake-up score is obtained according to the characteristic signal;

[0014] According to the first voltage information, a noise baseline is determined, and a wake-up threshold is obtained by combining historical wake-up records;

[0015] It is judged whether the first slope information exceeds a preset first slope threshold;

[0016] If yes, it is judged whether the first frequency information is within a preset first frequency range;

[0017] If no, it is determined as a first-level wake-up, and the wake-up score is adjusted upward;

[0018] If yes, it is determined as a second-level wake-up, and a wake-up action is triggered;

[0019] If the wake-up score exceeds the wake-up threshold, it is determined as a third-level wake-up, and a wake-up action and a wake-up diagnosis are triggered.

[0020] In the present scheme, based on the preset delay compensation, the time-aligned trigger signal matrix is generated according to the trigger signal, specifically:

[0021] Based on a preset sampling circuit or sampling channel, first delay information and second delay information are determined;

[0022] applying forward time shift compensation of the first delay information to the first level information;

[0023] applying forward time shift compensation of the second delay information to the first square wave information;

[0024] combining the compensated first level information and the first square wave information with the first voltage information to integrate into a trigger signal matrix according to a time axis, wherein row vectors of the trigger signal matrix correspond to a first level sequence, a first square wave sequence and a first voltage sequence respectively.

[0025] In the scheme, the feature signal is extracted according to the trigger signal matrix, specifically:

[0026] extracting the first level sequence based on a preset sliding window, and extracting a slope value as first slope information according to a first-order differential operation;

[0027] extracting the first square wave sequence based on a preset time window, and obtaining first frequency information according to zero-crossing point detection;

[0028] performing a root mean square calculation on the first voltage sequence, and obtaining first energy information by performing a preset normalization processing on a root mean square result.

[0029] In the scheme, the noise baseline is determined according to the first voltage information, and a wake-up threshold is obtained by combining a historical wake-up record, specifically:

[0030] recording the first voltage information when there is no wake-up event to obtain a second voltage sequence;

[0031] determining a noise baseline by calculating a root mean square value of a difference between voltage values of the second voltage sequence and a standard voltage value;

[0032] generating a historical false wake-up rate based on a preset statistical window according to a proportion of false wake-up events to total wake-up events in the historical wake-up record;

[0033] obtaining a wake-up threshold based on a pre-trained wake-up threshold association model according to the noise baseline and the historical false wake-up rate.

[0034] In the scheme, the wake-up action specifically includes:

[0035] outputting a high-level pulse signal through a preset GPIO interface;

[0036] controlling a power management unit to switch a supply voltage from an S3 / S4 maintenance voltage to a rated voltage;

[0037] generating and storing a wake-up log according to a feature signal and a wake-up level when a wake-up is triggered.

[0038] In this scheme, further comprising:

[0039] If the wake-up action is triggered, the default is marked as the first device wake-up event;

[0040] If the first slope information is greater than the preset second slope threshold, and the first frequency information is within the preset second frequency range, it is determined and recorded as a second device wake-up event;

[0041] If the first slope information is greater than the preset second slope threshold, and the first energy information exceeds the preset first energy threshold, it is determined and recorded as a third device wake-up event.

[0042] The second aspect of the present application provides a self-adaptive multi-source level wake-up system based on HiSilicon platform USB, comprising a self-adaptive multi-source level wake-up method program based on HiSilicon platform USB, which is executed by the processor to realize the following steps:

[0043] Parallelly collect trigger signals, including first level information, first square wave information and first voltage information;

[0044] Based on the preset delay compensation, generate a time-aligned trigger signal matrix according to the trigger signal;

[0045] According to the trigger signal matrix, extract feature signals, including first slope information, first frequency information and first energy information;

[0046] Based on the preset weighted fusion algorithm, obtain a wake-up score according to the feature signals;

[0047] According to the first voltage information, determine a noise baseline, and combine historical wake-up records to obtain a wake-up threshold;

[0048] Determine whether the first slope information exceeds a preset first slope threshold;

[0049] If yes, determine whether the first frequency information is within a preset first frequency range;

[0050] If no, determine it as a first-level wake-up, and increase the wake-up score;

[0051] If yes, determine it as a second-level wake-up, and trigger a wake-up action;

[0052] If the wake-up score exceeds the wake-up threshold, determine it as a third-level wake-up, trigger a wake-up action and start a wake-up diagnosis.

[0053] In this scheme, based on the preset delay compensation, the time-aligned trigger signal matrix is generated according to the trigger signal, specifically:

[0054] Based on a preset sampling circuit or sampling channel, determine the first delay information and the second delay information;

[0055] Apply positive time-shift compensation of the first delay information to the first level information;

[0056] Apply positive time-shift compensation of the second delay information to the first square wave information;

[0057] The compensated first level information and first square wave information are combined with the first voltage information and integrated into a trigger signal matrix according to the time axis. The row vectors of the trigger signal matrix correspond to the first level sequence, the first square wave sequence and the first voltage sequence, respectively.

[0058] In this scheme, the step of extracting feature signals based on the trigger signal matrix specifically involves:

[0059] The first level sequence is extracted based on a preset sliding window, and the slope value is extracted as the first slope information according to the first-order differential operation.

[0060] The first square wave sequence is extracted based on a preset time window, and the first frequency information is obtained by zero-crossing detection.

[0061] The root mean square (RMS) calculation is performed on the first voltage sequence, and the RMS result is normalized according to a preset method to obtain the first energy information.

[0062] A third aspect of the present invention provides a computer-readable storage medium comprising an adaptive multi-source level wake-up method program based on the HiSilicon platform USB, wherein when the adaptive multi-source level wake-up method program based on the HiSilicon platform USB is executed by a processor, the steps of the adaptive multi-source level wake-up method based on the HiSilicon platform USB as described in any of the preceding claims are implemented.

[0063] This invention provides an adaptive multi-source level wake-up method, system, and medium based on the HiSilicon platform USB. It synchronously captures the first level information, first square wave information, and first voltage information of the USB Hub, and generates a time-aligned trigger signal matrix through delay compensation. Based on the trigger signal matrix, feature signals are extracted, and a wake-up score is generated using a weighted fusion algorithm. An adaptive wake-up threshold is dynamically calculated based on a real-time noise baseline and historical false wake-up records. When only the first slope information exceeds the limit, it is determined to be a Level 1 wake-up, and the wake-up score is increased. When both the first slope information and the first frequency information exceed the limit, it is determined to be a Level 2 wake-up, triggering a wake-up action. When the wake-up score exceeds the wake-up threshold, it is determined to be a Level 3 wake-up, triggering a wake-up action and initiating wake-up diagnosis. Noise interference is reduced through multi-source signal cross-validation, and a dynamic threshold algorithm is used to adapt to complex industrial environments, improving the success rate and reliability of wake-up. Attached Figure Description

[0064] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope.

[0065] Figure 1 The flowchart of an adaptive multi-source level wake-up method based on the HiSilicon platform USB according to the present invention is shown;

[0066] Figure 2 A flowchart illustrating a method for generating a trigger signal matrix according to an embodiment of the present invention is shown;

[0067] Figure 3 This diagram illustrates a flowchart of a feature signal extraction method provided by an embodiment of the present invention.

[0068] Figure 4 A block diagram of an adaptive multi-source level wake-up system based on the HiSilicon platform USB is shown. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Unless otherwise defined, all terms (including technical and scientific terms) used in embodiments of this invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as being interpreted in an idealized or highly formalized sense, unless expressly defined in this embodiment of the invention.

[0071] The terms "first," "second," and similar words used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "an," "a," or "the" do not indicate a quantity limitation, but rather indicate the presence of at least one. Similarly, terms such as "including" or "comprising" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The steps preceding or following the steps in the method of the embodiments of this invention are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0072] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0073] Figure 1 The flowchart of an adaptive multi-source level wake-up method based on the HiSilicon platform USB is shown.

[0074] like Figure 1 As shown, the first aspect of this invention discloses an adaptive multi-source level wake-up method based on the HiSilicon platform USB, the method comprising:

[0075] Parallel acquisition of trigger signals, including first level information, first square wave information, and first voltage information;

[0076] Based on the preset delay compensation, a time-aligned trigger signal matrix is ​​generated according to the trigger signal;

[0077] Based on the trigger signal matrix, feature signals are extracted, including first slope information, first frequency information, and first energy information;

[0078] Based on a preset weighted fusion algorithm, a wake-up score is obtained according to the feature signals;

[0079] Based on the first voltage information, the noise baseline is determined, and the wake-up threshold is obtained by combining the historical wake-up records.

[0080] Determine whether the first slope information exceeds a preset first slope threshold;

[0081] If so, determine whether the first frequency information is within a preset first frequency range;

[0082] If not, it is determined to be a Level 1 wake-up, and the wake-up score is increased;

[0083] If so, it is determined to be a level 2 wake-up, triggering the wake-up action;

[0084] If the wake-up score exceeds the wake-up threshold, it is determined to be a level 3 wake-up, triggering a wake-up action and initiating wake-up diagnosis.

[0085] It should be noted that the USB Hub includes a PGANG pin, a DRV pin, and a VSENSE pin; the first level information is the level value of the PGANG pin; the first square wave information is the square wave value of the DRV pin; the first voltage information is the voltage value of the VSENSE pin; the first slope information is the slope of the PGANG level change; the first frequency information is the frequency of the square wave on the DRV pin; and the first energy information is the fluctuation energy reflected by the voltage on the VSENSE pin.

[0086] In this embodiment, when the system enters the S3 / S4 sleep state, the processor initiates three-channel parallel sampling. The PGANG pin level signal is stabilized by a 100kΩ pull-down resistor and then input to the ADC channel. The DRV pin square wave signal is captured by a timer, and the VSENSE pin voltage is amplified by a dual-stage operational amplifier before sampling. Since the three signals pass through different hardware circuits and detection structures, delay compensation is performed based on the corresponding delay conditions. As one implementation, the PGANG level signal is delayed by 0.2ms, and the DRV square wave signal is delayed by 1.5ms. Subsequently, based on the delay-compensated sampled data, a time-axis-aligned two-dimensional matrix is ​​constructed in memory. The two-dimensional matrix includes the PGANG level sequence, the DRV square wave sequence, and the VSENSE voltage sequence. Next, based on preset feature extraction logic, the two-dimensional matrix is ​​analyzed in real time. The PGANG level sequence's differential derivative or difference value is calculated as a slope feature to obtain the PGANG slope. The DRV square wave sequence's frequency is derived by counting the number of zero-crossing points within a set time window. The VSENSE energy value is obtained by performing root mean square (RMS) calculation on the VSENSE sequence and then normalizing it. Then, a wake-up score is calculated based on the weighted weights corresponding to the PGANG slope, DRV frequency, and VSENSE energy value in a preset weighted fusion algorithm. An adaptive threshold is generated by fusing the environmental noise baseline and historical false wake-up rate according to a dynamic decision-making mechanism. The wake-up level is determined based on a preset hierarchical wake-up triggering mechanism. Level 1 wake-up is initiated only when the PGANG slope exceeds a preset first slope threshold, and the wake-up score is increased. Level 2 wake-up is initiated when the PGANG slope exceeds the preset first slope threshold and the DRV frequency is within a set first frequency range, waking up the EC controller and switching the power supply voltage. If neither of the above conditions is met, but the wake-up score exceeds the wake-up threshold, Level 3 wake-up is activated, performing wake-up actions and diagnosing wake-up parameters. This embodiment reduces the problem of false wake-up judgment by using multi-source composite signals; and reduces the false judgment rate and improves the reliability of wake-up operation based on three-level wake-up decision.

[0087] Figure 2 A flowchart illustrating a method for generating a trigger signal matrix according to an embodiment of the present invention is shown.

[0088] According to embodiments of the present invention, such as Figure 2 As shown, the step of generating a time-aligned trigger signal matrix based on the preset delay compensation and the trigger signal specifically involves:

[0089] Based on a preset sampling circuit or sampling channel, determine the first delay information and the second delay information;

[0090] Apply positive time-shift compensation of the first delay information to the first level information;

[0091] Apply positive time-shift compensation of the second delay information to the first square wave information;

[0092] The compensated first level information and first square wave information are combined with the first voltage information and integrated into a trigger signal matrix according to the time axis. The row vectors of the trigger signal matrix correspond to the first level sequence, the first square wave sequence and the first voltage sequence, respectively.

[0093] It should be noted that the first delay information is the delay caused by the PGANG level sampling circuit and logic; the second delay information is the delay caused by the DRV square wave sampling circuit and logic. This embodiment provides a generation process for a trigger signal matrix. To eliminate signal transmission delay differences, the system allocates independent ADC channels for the PGANG level and VSENSE voltage, and allocates a timer for the DRV square wave. As one implementation method, after the PGANG signal is input, a 0.2ms time shift compensation is applied to offset the delay of its internal sampling circuit; a 1.5ms compensation is applied to the DRV signal to solve the response lag of the square wave generation circuit; the VSENSE signal is processed by pass-through. The timestamps of the three compensated data channels are aligned to construct a trigger signal matrix with a fixed row vector dimension, wherein the first row is the PGANG level sequence, the second row is the DRV square wave sequence, and the third row is the VSENSE voltage sequence. This embodiment, based on delay compensation, ensures the timing consistency of wake-up event analysis and avoids feature extraction distortion caused by signal asynchrony.

[0094] Figure 3 A flowchart illustrating a feature signal extraction method provided by an embodiment of the present invention is shown.

[0095] According to embodiments of the present invention, such as Figure 3 As shown, the step of extracting feature signals based on the trigger signal matrix specifically involves:

[0096] The first level sequence is extracted based on a preset sliding window, and the slope value is extracted as the first slope information according to the first-order differential operation.

[0097] The first square wave sequence is extracted based on a preset time window, and the first frequency information is obtained by zero-crossing detection.

[0098] The root mean square (RMS) calculation is performed on the first voltage sequence, and the RMS result is normalized according to a preset method to obtain the first energy information.

[0099] It should be noted that this embodiment provides a feature extraction process for wake-up signals. As one implementation, each sequence is separated from the trigger signal matrix. The PGANG level sequence operates based on a preset first-order derivative, derivative, or difference, calculating the instantaneous slope through an adjustable sliding window to lock the maximum positive slope value as a wake-up sensitivity indicator. The DRV square wave sequence uses a zero-crossing detection mechanism to count the number of rising and falling edge transitions within a preset time window, calculating the DRV square wave frequency characteristics based on the number of transitions and the time window length. The VSENSE voltage sequence first performs a square operation to eliminate negative values, then calculates the average value within the window, and finally takes the square root to obtain the root mean square energy value; the root mean square energy value is normalized, mapping it to the [0,1] range. The processing result is encapsulated as a feature vector for use by the decision-making layer. This embodiment accurately captures PGANG level transitions through differential operations, enhances the anti-amplitude interference of the DRV square wave through zero-crossing detection, and eliminates individual device differences through energy standardization; thereby improving the accuracy of subsequent analysis and decision-making.

[0100] According to an embodiment of the present invention, the step of determining the noise baseline based on the first voltage information and obtaining the wake-up threshold by combining it with historical wake-up records specifically includes:

[0101] When there is no wake-up event, the first voltage information is recorded to obtain the second voltage sequence;

[0102] The root mean square value of the difference between the voltage values ​​of the second voltage sequence and the standard voltage value is calculated to determine the noise baseline.

[0103] Based on a preset statistical window, a historical false wake-up rate is generated according to the proportion of false wake-up events in the historical wake-up records to the total number of wake-up events.

[0104] Based on a pre-trained wake-up threshold association model, the wake-up threshold is obtained according to the noise baseline and the historical false wake-up rate.

[0105] It should be noted that this embodiment provides an adaptive setting mechanism for dynamic thresholds. Noise baseline calculation is automatically performed during system idle periods, based on the recorded VSENSE voltage when there are no wake-up events, recorded as a second voltage sequence. The deviation between the second voltage sequence and the standard voltage value is calculated, and the root mean square value of the deviation is used as the environmental noise benchmark. As one implementation, a 24-hour statistical window is used to retrieve all wake-up records; and the percentage of records marked as false wake-up events out of the total number of wake-ups is calculated to obtain the historical false wake-up rate. The environmental noise benchmark and the historical false wake-up rate are input into a pre-trained model, and their combined effect is adjusted through a negative feedback coefficient to dynamically generate the wake-up threshold. As one implementation, the wake-up threshold association model is a neural network model; as another implementation, the wake-up threshold association model is a mapping relationship. This embodiment uses a dynamically adaptable working voltage fluctuation environment in conjunction with the threshold association model to improve the accuracy of wake-up judgment.

[0106] According to an embodiment of the present invention, the wake-up action specifically includes:

[0107] Output a high-level pulse signal through the preset GPIO interface;

[0108] The control power management unit switches the supply voltage from the S3 / S4 sustaining voltage to the rated voltage;

[0109] Based on the characteristic signals and wake-up level triggered when wake-up is activated, wake-up logs are generated and stored.

[0110] It should be noted that, as one implementation method, the wake-up action specifically includes pulse output, voltage switching, and log recording. When a valid wake-up is determined, the GPIO_CTRL pin outputs a high-level pulse of a set duration. In one implementation method, this pulse is transmitted to the power management unit after opto-isolation. Based on the pulse signal, the S3 / S4 energy-saving power supply circuit is first cut off, then the 3.3V rated voltage converter is activated, completing the CPU core power supply switching within the safe switching time. Simultaneously, the log engine records wake-up event characteristics, including a timestamp, storing PGANG slope peak, DRV dominant frequency, VSENSE energy surge increment, and wake-up level code. This embodiment uses log recording for later analysis and tracing.

[0111] According to an embodiment of the present invention, it further includes:

[0112] If a wake-up action is triggered, it will be marked as the first device wake-up event by default;

[0113] If the first slope information is greater than the preset second slope threshold, and the first frequency information is within the preset second frequency range, then it is determined and recorded as a second device wake-up event;

[0114] If the first slope information is greater than the preset second slope threshold, and the first energy information exceeds the preset first energy threshold, then it is determined and recorded as a third device wake-up event.

[0115] It should be noted that the first device wake-up event indicates an event triggered by an unknown device; the second device wake-up event indicates an event triggered by a mouse; and the third device wake-up event indicates an event triggered by a keyboard. This embodiment provides a wake-up device classification mechanism. When a wake-up is triggered, it is defaulted to an unknown device event. As one implementation, if the PGANG slope is greater than 0.8V / ms and the DRV frequency is in the range of 950-1050Hz, it is marked as a mouse device; if the slope exceeds 0.8V / ms and the VSENSE energy is greater than 50mV, it is marked as a keyboard device. After being marked as a mouse or keyboard device, it can be used to select a communication protocol after the wake-up operation, shortening the device activation time.

[0116] It's worth mentioning that after being determined to be a Level 3 wake-up, the following also applies:

[0117] Based on the first level sequence, the slope change rate is obtained;

[0118] If the slope change rate is lower than the preset change rate threshold, the wake-up action is rejected and determined to be an interference event.

[0119] Perform a fast Fourier transform on the first square wave sequence and statistically analyze the energy proportion of a preset frequency band based on the frequency domain data;

[0120] If the energy percentage exceeds a preset percentage threshold, the wake-up action is rejected and determined to be an interference event.

[0121] If the first energy information remains below a preset second energy threshold, the wake-up action is rejected and determined to be an interference event.

[0122] It should be noted that this embodiment provides a triple interference detection mechanism. In this embodiment, after the three-level wake-up is triggered, the triple anti-interference detection is automatically started. First, the PGANG slope change rate is analyzed. If the slope change rate is lower than the change rate threshold, it is judged as slow-varying interference, such as temperature drift causing level shift, and therefore, the wake-up is rejected. Then, the DRV square wave is subjected to a 512-point Fast Fourier Transform (FFT) to calculate the energy ratio in the 50Hz-60Hz frequency band. When the energy ratio exceeds the set 30% threshold, it is judged as power frequency interference, and the wake-up is rejected. Finally, the VSENSE energy is continuously monitored. If the normalized energy value identified in the VSENSE sequence is continuously lower than 0.3, it is judged as transient interference, and the wake-up is rejected. If any of the above steps is rejected, the wake-up process is terminated, and the event is recorded as "interference rejection". This embodiment improves anti-interference capability through triple verification; among them, frequency domain analysis solves the impact of power frequency interference, and continuous energy monitoring prevents contact jitter and accidental touch.

[0123] It is worth mentioning that it also includes:

[0124] Establish a wake-up event database to record the characteristic signals, wake-up levels, and wake-up device types;

[0125] When a false wake-up event is identified, the current event is recorded in the interference pattern feature database.

[0126] Adjust the weight coefficients in the weighted fusion algorithm based on the characteristic signals and wake-up level of the current event;

[0127] Write the updated weight coefficients and interference mode feature library into non-volatile memory.

[0128] It should be noted that this embodiment provides a self-learning mechanism to adaptively adjust the weight coefficients in the weighted fusion algorithm. In this embodiment, the adaptive learning mechanism includes two stages: online learning and offline storage. After each wake-up event, the slope, frequency, and energy value as feature vectors, the decision result, and the device type are packaged into a data packet and stored in the RAM cache. If the event is rejected by triple verification, the data packet is appended with a false wake-up tag and migrated to the interference feature library; if the wake-up is successful, it is marked as a positive sample and added to the positive sample library. Based on a preset self-learning cycle, the frequency of occurrence of similar feature patterns in the false wake-up library is counted, and the corresponding signal weights are reduced. As one implementation method, if the PGANG slope determines frequent false touches, the PGANG slope weight coefficient is lowered; furthermore, the positive sample library is analyzed to optimize the recognition threshold. Finally, the updated weight coefficients and the interference pattern feature library are written to local non-volatile memory to achieve offline storage.

[0129] It is worth mentioning that it also includes:

[0130] When the ambient temperature exceeds the preset temperature threshold, the range between the first frequency range and the second frequency range is expanded and a safety event log is recorded.

[0131] When an environmental surge is determined based on the first energy information, the wake-up function is turned off based on the preset shutdown delay and a safety event log is recorded.

[0132] It should be noted that this implementation also provides an anti-environmental interference mechanism. For high-temperature environments, when the temperature sensor detects an ambient temperature exceeding 85°C, the frequency tolerance is automatically extended to ±10%, and a high-temperature compensation log is recorded. This log includes the temperature value, compensation time, and original frequency data. As one implementation, when the temperature exceeds 85°C, the first frequency range is extended from 1kHz ±5% to 1kHz ±10% to compensate for crystal oscillator temperature drift. For power surge conditions, as one implementation, if the normalized energy value of the first energy information exceeds 0.8 for 100ms, it is determined to be an environmental surge. The wake-up function is then automatically disabled for 200ms, and a surge event log is generated. This log includes the peak voltage, duration, and spectral characteristics.

[0133] Figure 4 A block diagram of an adaptive multi-source level wake-up system based on the HiSilicon platform USB is shown.

[0134] like Figure 4 As shown, the second aspect of the present invention discloses an adaptive multi-source level wake-up system 4 based on the HiSilicon platform USB, including a memory 41 and a processor 42. The memory includes an adaptive multi-source level wake-up method program based on the HiSilicon platform USB. When the processor executes the adaptive multi-source level wake-up method program based on the HiSilicon platform USB, it implements the following steps:

[0135] Parallel acquisition of trigger signals, including first level information, first square wave information, and first voltage information;

[0136] Based on the preset delay compensation, a time-aligned trigger signal matrix is ​​generated according to the trigger signal;

[0137] Based on the trigger signal matrix, feature signals are extracted, including first slope information, first frequency information, and first energy information;

[0138] Based on a preset weighted fusion algorithm, a wake-up score is obtained according to the feature signals;

[0139] Based on the first voltage information, the noise baseline is determined, and the wake-up threshold is obtained by combining the historical wake-up records.

[0140] Determine whether the first slope information exceeds a preset first slope threshold;

[0141] If so, determine whether the first frequency information is within a preset first frequency range;

[0142] If not, it is determined to be a Level 1 wake-up, and the wake-up score is increased;

[0143] If so, it is determined to be a level 2 wake-up, triggering the wake-up action;

[0144] If the wake-up score exceeds the wake-up threshold, it is determined to be a level 3 wake-up, triggering a wake-up action and initiating wake-up diagnosis.

[0145] It should be noted that the USB Hub includes a PGANG pin, a DRV pin, and a VSENSE pin; the first level information is the level value of the PGANG pin; the first square wave information is the square wave value of the DRV pin; the first voltage information is the voltage value of the VSENSE pin; the first slope information is the slope of the PGANG level change; the first frequency information is the frequency of the square wave on the DRV pin; and the first energy information is the fluctuation energy reflected by the voltage on the VSENSE pin.

[0146] In this embodiment, when the system enters the S3 / S4 sleep state, the processor initiates three-channel parallel sampling. The PGANG pin level signal is stabilized by a 100kΩ pull-down resistor and then input to the ADC channel. The DRV pin square wave signal is captured by a timer, and the VSENSE pin voltage is amplified by a dual-stage operational amplifier before sampling. Since the three signals pass through different hardware circuits and detection structures, delay compensation is performed based on the corresponding delay conditions. As one implementation, the PGANG level signal is delayed by 0.2ms, and the DRV square wave signal is delayed by 1.5ms. Subsequently, based on the delay-compensated sampled data, a time-axis-aligned two-dimensional matrix is ​​constructed in memory. The two-dimensional matrix includes the PGANG level sequence, the DRV square wave sequence, and the VSENSE voltage sequence. Next, based on preset feature extraction logic, the two-dimensional matrix is ​​analyzed in real time. The PGANG level sequence's differential derivative or difference value is calculated as a slope feature to obtain the PGANG slope. The DRV square wave sequence's frequency is derived by counting the number of zero-crossing points within a set time window. The VSENSE energy value is obtained by performing root mean square (RMS) calculation on the VSENSE sequence and then normalizing it. Then, a wake-up score is calculated based on the weighted weights corresponding to the PGANG slope, DRV frequency, and VSENSE energy value in a preset weighted fusion algorithm. An adaptive threshold is generated by fusing the environmental noise baseline and historical false wake-up rate according to a dynamic decision-making mechanism. The wake-up level is determined based on a preset hierarchical wake-up triggering mechanism. Level 1 wake-up is initiated only when the PGANG slope exceeds a preset first slope threshold, and the wake-up score is increased. Level 2 wake-up is initiated when the PGANG slope exceeds the preset first slope threshold and the DRV frequency is within a set first frequency range, waking up the EC controller and switching the power supply voltage. If neither of the above conditions is met, but the wake-up score exceeds the wake-up threshold, Level 3 wake-up is activated, performing wake-up actions and diagnosing wake-up parameters. This embodiment reduces the problem of false wake-up judgment by using multi-source composite signals; and reduces the false judgment rate and improves the reliability of wake-up operation based on three-level wake-up decision.

[0147] According to an embodiment of the present invention, the step of generating a time-aligned trigger signal matrix based on the trigger signal according to the preset delay compensation specifically includes:

[0148] Based on a preset sampling circuit or sampling channel, determine the first delay information and the second delay information;

[0149] Apply positive time-shift compensation of the first delay information to the first level information;

[0150] Apply positive time-shift compensation of the second delay information to the first square wave information;

[0151] The compensated first level information and first square wave information are combined with the first voltage information and integrated into a trigger signal matrix according to the time axis. The row vectors of the trigger signal matrix correspond to the first level sequence, the first square wave sequence and the first voltage sequence, respectively.

[0152] It should be noted that the first delay information is the delay caused by the PGANG level sampling circuit and logic; the second delay information is the delay caused by the DRV square wave sampling circuit and logic. This embodiment provides a generation process for a trigger signal matrix. To eliminate signal transmission delay differences, the system allocates independent ADC channels for the PGANG level and VSENSE voltage, and allocates a timer for the DRV square wave. As one implementation method, after the PGANG signal is input, a 0.2ms time shift compensation is applied to offset the delay of its internal sampling circuit; a 1.5ms compensation is applied to the DRV signal to solve the response lag of the square wave generation circuit; the VSENSE signal is processed by pass-through. The timestamps of the three compensated data channels are aligned to construct a trigger signal matrix with a fixed row vector dimension, wherein the first row is the PGANG level sequence, the second row is the DRV square wave sequence, and the third row is the VSENSE voltage sequence. This embodiment, based on delay compensation, ensures the timing consistency of wake-up event analysis and avoids feature extraction distortion caused by signal asynchrony.

[0153] According to an embodiment of the present invention, the step of extracting feature signals based on the trigger signal matrix specifically includes:

[0154] The first level sequence is extracted based on a preset sliding window, and the slope value is extracted as the first slope information according to the first-order differential operation.

[0155] The first square wave sequence is extracted based on a preset time window, and the first frequency information is obtained by zero-crossing detection.

[0156] The root mean square (RMS) calculation is performed on the first voltage sequence, and the RMS result is normalized according to a preset method to obtain the first energy information.

[0157] It should be noted that this embodiment provides a feature extraction process for wake-up signals. As one implementation, each sequence is separated from the trigger signal matrix. The PGANG level sequence operates based on a preset first-order derivative, derivative, or difference, calculating the instantaneous slope through an adjustable sliding window to lock the maximum positive slope value as a wake-up sensitivity indicator. The DRV square wave sequence uses a zero-crossing detection mechanism to count the number of rising and falling edge transitions within a preset time window, calculating the DRV square wave frequency characteristics based on the number of transitions and the time window length. The VSENSE voltage sequence first performs a square operation to eliminate negative values, then calculates the average value within the window, and finally takes the square root to obtain the root mean square energy value; the root mean square energy value is normalized, mapping it to the [0,1] range. The processing result is encapsulated as a feature vector for use by the decision-making layer. This embodiment accurately captures PGANG level transitions through differential operations, enhances the anti-amplitude interference of the DRV square wave through zero-crossing detection, and eliminates individual device differences through energy standardization; thereby improving the accuracy of subsequent analysis and decision-making.

[0158] According to an embodiment of the present invention, the step of determining the noise baseline based on the first voltage information and obtaining the wake-up threshold by combining it with historical wake-up records specifically includes:

[0159] When there is no wake-up event, the first voltage information is recorded to obtain the second voltage sequence;

[0160] The root mean square value of the difference between the voltage values ​​of the second voltage sequence and the standard voltage value is calculated to determine the noise baseline.

[0161] Based on a preset statistical window, a historical false wake-up rate is generated according to the proportion of false wake-up events in the historical wake-up records to the total number of wake-up events.

[0162] Based on a pre-trained wake-up threshold association model, the wake-up threshold is obtained according to the noise baseline and the historical false wake-up rate.

[0163] It should be noted that this embodiment provides an adaptive setting mechanism for dynamic thresholds. Noise baseline calculation is automatically performed during system idle periods, based on the recorded VSENSE voltage when there are no wake-up events, recorded as a second voltage sequence. The deviation between the second voltage sequence and the standard voltage value is calculated, and the root mean square value of the deviation is used as the environmental noise benchmark. As one implementation, a 24-hour statistical window is used to retrieve all wake-up records; and the percentage of records marked as false wake-up events out of the total number of wake-ups is calculated to obtain the historical false wake-up rate. The environmental noise benchmark and the historical false wake-up rate are input into a pre-trained model, and their combined effect is adjusted through a negative feedback coefficient to dynamically generate the wake-up threshold. As one implementation, the wake-up threshold association model is a neural network model; as another implementation, the wake-up threshold association model is a mapping relationship. This embodiment uses a dynamically adaptable working voltage fluctuation environment in conjunction with the threshold association model to improve the accuracy of wake-up judgment.

[0164] According to an embodiment of the present invention, the wake-up action specifically includes:

[0165] Output a high-level pulse signal through the preset GPIO interface;

[0166] The control power management unit switches the supply voltage from the S3 / S4 sustaining voltage to the rated voltage;

[0167] Based on the characteristic signals and wake-up level triggered when wake-up is activated, wake-up logs are generated and stored.

[0168] It should be noted that, as one implementation method, the wake-up action specifically includes pulse output, voltage switching, and log recording. When a valid wake-up is determined, the GPIO_CTRL pin outputs a high-level pulse of a set duration. In one implementation method, this pulse is transmitted to the power management unit after opto-isolation. Based on the pulse signal, the S3 / S4 energy-saving power supply circuit is first cut off, then the 3.3V rated voltage converter is activated, completing the CPU core power supply switching within the safe switching time. Simultaneously, the log engine records wake-up event characteristics, including a timestamp, storing PGANG slope peak, DRV dominant frequency, VSENSE energy surge increment, and wake-up level code. This embodiment uses log recording for later analysis and tracing.

[0169] According to an embodiment of the present invention, it further includes:

[0170] If a wake-up action is triggered, it will be marked as the first device wake-up event by default;

[0171] If the first slope information is greater than the preset second slope threshold, and the first frequency information is within the preset second frequency range, then it is determined and recorded as a second device wake-up event;

[0172] If the first slope information is greater than the preset second slope threshold, and the first energy information exceeds the preset first energy threshold, then it is determined and recorded as a third device wake-up event.

[0173] It should be noted that the first device wake-up event indicates an event triggered by an unknown device; the second device wake-up event indicates an event triggered by a mouse; and the third device wake-up event indicates an event triggered by a keyboard. This embodiment provides a wake-up device classification mechanism. When a wake-up is triggered, it is defaulted to an unknown device event. As one implementation, if the PGANG slope is greater than 0.8V / ms and the DRV frequency is in the range of 950-1050Hz, it is marked as a mouse device; if the slope exceeds 0.8V / ms and the VSENSE energy is greater than 50mV, it is marked as a keyboard device. After being marked as a mouse or keyboard device, it can be used to select a communication protocol after the wake-up operation, shortening the device activation time.

[0174] It's worth mentioning that after being determined to be a Level 3 wake-up, the following also applies:

[0175] Based on the first level sequence, the slope change rate is obtained;

[0176] If the slope change rate is lower than the preset change rate threshold, the wake-up action is rejected and determined to be an interference event.

[0177] Perform a fast Fourier transform on the first square wave sequence and statistically analyze the energy proportion of a preset frequency band based on the frequency domain data;

[0178] If the energy percentage exceeds a preset percentage threshold, the wake-up action is rejected and determined to be an interference event.

[0179] If the first energy information remains below a preset second energy threshold, the wake-up action is rejected and determined to be an interference event.

[0180] It should be noted that this embodiment provides a triple interference detection mechanism. In this embodiment, after the three-level wake-up is triggered, the triple anti-interference detection is automatically started. First, the PGANG slope change rate is analyzed. If the slope change rate is lower than the change rate threshold, it is judged as slow-varying interference, such as temperature drift causing level shift, and therefore, the wake-up is rejected. Then, the DRV square wave is subjected to a 512-point Fast Fourier Transform (FFT) to calculate the energy ratio in the 50Hz-60Hz frequency band. When the energy ratio exceeds the set 30% threshold, it is judged as power frequency interference, and the wake-up is rejected. Finally, the VSENSE energy is continuously monitored. If the normalized energy value identified in the VSENSE sequence is continuously lower than 0.3, it is judged as transient interference, and the wake-up is rejected. If any of the above steps is rejected, the wake-up process is terminated, and the event is recorded as "interference rejection". This embodiment improves anti-interference capability through triple verification; among them, frequency domain analysis solves the impact of power frequency interference, and continuous energy monitoring prevents contact jitter and accidental touch.

[0181] It is worth mentioning that it also includes:

[0182] Establish a wake-up event database to record the characteristic signals, wake-up levels, and wake-up device types;

[0183] When a false wake-up event is identified, the current event is recorded in the interference pattern feature database.

[0184] Adjust the weight coefficients in the weighted fusion algorithm based on the characteristic signals and wake-up level of the current event;

[0185] Write the updated weight coefficients and interference mode feature library into non-volatile memory.

[0186] It should be noted that this embodiment provides a self-learning mechanism to adaptively adjust the weight coefficients in the weighted fusion algorithm. In this embodiment, the adaptive learning mechanism includes two stages: online learning and offline storage. After each wake-up event, the slope, frequency, and energy value as feature vectors, the decision result, and the device type are packaged into a data packet and stored in the RAM cache. If the event is rejected by triple verification, the data packet is appended with a false wake-up tag and migrated to the interference feature library; if the wake-up is successful, it is marked as a positive sample and added to the positive sample library. Based on a preset self-learning cycle, the frequency of occurrence of similar feature patterns in the false wake-up library is counted, and the corresponding signal weights are reduced. As one implementation method, if the PGANG slope determines frequent false touches, the PGANG slope weight coefficient is lowered; furthermore, the positive sample library is analyzed to optimize the recognition threshold. Finally, the updated weight coefficients and the interference pattern feature library are written to local non-volatile memory to achieve offline storage.

[0187] It is worth mentioning that it also includes:

[0188] When the ambient temperature exceeds the preset temperature threshold, the range between the first frequency range and the second frequency range is expanded and a safety event log is recorded.

[0189] When an environmental surge is determined based on the first energy information, the wake-up function is turned off based on the preset shutdown delay and a safety event log is recorded.

[0190] It should be noted that this implementation also provides an anti-environmental interference mechanism. For high-temperature environments, when the temperature sensor detects an ambient temperature exceeding 85°C, the frequency tolerance is automatically extended to ±10%, and a high-temperature compensation log is recorded. This log includes the temperature value, compensation time, and original frequency data. As one implementation, when the temperature exceeds 85°C, the first frequency range is extended from 1kHz ±5% to 1kHz ±10% to compensate for crystal oscillator temperature drift. For power surge conditions, as one implementation, if the normalized energy value of the first energy information exceeds 0.8 for 100ms, it is determined to be an environmental surge. The wake-up function is then automatically disabled for 200ms, and a surge event log is generated. This log includes the peak voltage, duration, and spectral characteristics.

[0191] A third aspect of the present invention provides a computer-readable storage medium comprising an adaptive multi-source level wake-up method program based on the HiSilicon platform USB, wherein when the adaptive multi-source level wake-up method program based on the HiSilicon platform USB is executed by a processor, the steps of the adaptive multi-source level wake-up method based on the HiSilicon platform USB as described in any of the preceding claims are implemented.

[0192] In summary, this invention provides an adaptive multi-source level wake-up method, system, and medium based on the HiSilicon platform USB. It synchronously captures the first level information, first square wave information, and first voltage information of the USB Hub, and generates a time-aligned trigger signal matrix through delay compensation. Based on the trigger signal matrix, feature signals are extracted, and a wake-up score is generated using a weighted fusion algorithm. An adaptive wake-up threshold is dynamically calculated based on a real-time noise baseline and historical false wake-up records. When only the first slope information exceeds the limit, it is determined to be a Level 1 wake-up, and the wake-up score is increased. When both the first slope information and the first frequency information exceed the limit, it is determined to be a Level 2 wake-up, and a wake-up action is triggered. When the wake-up score exceeds the wake-up threshold, it is determined to be a Level 3 wake-up, triggering a wake-up action and initiating wake-up diagnosis. Noise interference is reduced through multi-source signal cross-validation, and a dynamic threshold algorithm is used to adapt to complex industrial environments, improving the success rate and reliability of wake-up.

[0193] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0194] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive multi-source level wake-up method based on the HiSilicon platform USB, characterized in that, The method includes: Parallel acquisition of trigger signals, including first level information, first square wave information and first voltage information, wherein the first level information is the level value of the PGANG pin, the first square wave information is the square wave value of the DRV pin, and the first voltage information is the voltage value of the VSENSE pin; Based on the preset delay compensation, a time-aligned trigger signal matrix is ​​generated according to the trigger signal; Based on the trigger signal matrix, feature signals are extracted, including first slope information, first frequency information and first energy information, wherein the first slope information is the slope of the change of the PGANG level, the first frequency information is the frequency of the square wave of the DRV pin, and the first energy information is the fluctuation energy reflected by the voltage of the VSENSE pin. Based on a preset weighted fusion algorithm, a wake-up score is obtained according to the feature signals; Based on the first voltage information, a noise baseline is determined, and a wake-up threshold is obtained by combining historical wake-up records. Specifically, the first voltage information is recorded when there are no wake-up events to obtain a second voltage sequence; the root mean square value of the difference between the voltage value of the second voltage sequence and the standard voltage value is calculated to determine the noise baseline; based on a preset statistical window, a historical false wake-up rate is generated according to the proportion of false wake-up events in the historical wake-up records to the total number of wake-up events; based on a pre-trained wake-up threshold association model, the wake-up threshold is obtained according to the noise baseline and the historical false wake-up rate. Determine whether the first slope information exceeds a preset first slope threshold; If the first slope information exceeds a preset first slope threshold, then it is further determined whether the first frequency information is within a preset first frequency range. If the first frequency information is not within the preset first frequency range, it is determined to be a level 1 wake-up, and the wake-up score is increased. If the first frequency information is within the preset first frequency range, it is determined to be a level 2 wake-up and a wake-up action is triggered. If the first slope information does not exceed the preset first slope threshold, and the wake-up score exceeds the wake-up threshold, then it is determined to be a level three wake-up, triggering a wake-up action and initiating wake-up diagnosis.

2. The adaptive multi-source level wake-up method based on the HiSilicon platform USB according to claim 1, characterized in that, The process of generating a time-aligned trigger signal matrix based on the preset delay compensation and the trigger signal specifically involves: Based on a preset sampling circuit or sampling channel, determine the first delay information and the second delay information; Apply positive time-shift compensation of the first delay information to the first level information; Apply positive time-shift compensation of the second delay information to the first square wave information; The compensated first level information and first square wave information are combined with the first voltage information and integrated into a trigger signal matrix according to the time axis. The row vectors of the trigger signal matrix correspond to the first level sequence, the first square wave sequence and the first voltage sequence, respectively.

3. The adaptive multi-source level wake-up method based on the HiSilicon platform USB according to claim 2, characterized in that, The step of extracting feature signals based on the trigger signal matrix specifically involves: The first level sequence is extracted based on a preset sliding window, and the slope value is extracted as the first slope information according to the first-order differential operation. The first square wave sequence is extracted based on a preset time window, and the first frequency information is obtained by zero-crossing detection. The root mean square (RMS) calculation is performed on the first voltage sequence, and the RMS result is normalized according to a preset method to obtain the first energy information.

4. The adaptive multi-source level wake-up method based on the HiSilicon platform USB according to claim 1, characterized in that, The wake-up action specifically includes: Output a high-level pulse signal through the preset GPIO interface; The control power management unit switches the supply voltage from the S3 / S4 sustaining voltage to the rated voltage; Based on the characteristic signals and wake-up level triggered when wake-up is activated, wake-up logs are generated and stored.

5. The adaptive multi-source level wake-up method based on the HiSilicon platform USB according to claim 1, characterized in that, Also includes: If a wake-up action is triggered, it will be marked as the first device wake-up event by default; If the first slope information is greater than the preset second slope threshold, and the first frequency information is within the preset second frequency range, then it is determined and recorded as a second device wake-up event; If the first slope information is greater than the preset second slope threshold, and the first energy information exceeds the preset first energy threshold, then it is determined and recorded as a third device wake-up event.

6. An adaptive multi-source level wake-up system based on the HiSilicon platform USB, characterized in that, The system includes a memory and a processor. The memory includes a program for an adaptive multi-source level wake-up method based on the HiSilicon platform USB. When the processor executes the program for the adaptive multi-source level wake-up method based on the HiSilicon platform USB, it performs the following steps: Parallel acquisition of trigger signals, including first level information, first square wave information and first voltage information, wherein the first level information is the level value of the PGANG pin, the first square wave information is the square wave value of the DRV pin, and the first voltage information is the voltage value of the VSENSE pin; Based on the preset delay compensation, a time-aligned trigger signal matrix is ​​generated according to the trigger signal; Based on the trigger signal matrix, feature signals are extracted, including first slope information, first frequency information and first energy information, wherein the first slope information is the slope of the change of the PGANG level, the first frequency information is the frequency of the square wave of the DRV pin, and the first energy information is the fluctuation energy reflected by the voltage of the VSENSE pin. Based on a preset weighted fusion algorithm, a wake-up score is obtained according to the feature signals; Based on the first voltage information, a noise baseline is determined, and a wake-up threshold is obtained by combining historical wake-up records. Specifically, the first voltage information is recorded when there are no wake-up events to obtain a second voltage sequence; the root mean square value of the difference between the voltage value of the second voltage sequence and the standard voltage value is calculated to determine the noise baseline; based on a preset statistical window, a historical false wake-up rate is generated according to the proportion of false wake-up events in the historical wake-up records to the total number of wake-up events; based on a pre-trained wake-up threshold association model, the wake-up threshold is obtained according to the noise baseline and the historical false wake-up rate. Determine whether the first slope information exceeds a preset first slope threshold; If the first slope information exceeds a preset first slope threshold, then it is further determined whether the first frequency information is within a preset first frequency range. If the first frequency information is not within the preset first frequency range, it is determined to be a level 1 wake-up, and the wake-up score is increased. If the first frequency information is within the preset first frequency range, it is determined to be a level 2 wake-up and a wake-up action is triggered. If the first slope information does not exceed the preset first slope threshold, and the wake-up score exceeds the wake-up threshold, then it is determined to be a level three wake-up, triggering a wake-up action and initiating wake-up diagnosis.

7. The adaptive multi-source level wake-up system based on the HiSilicon platform USB according to claim 6, characterized in that, The process of generating a time-aligned trigger signal matrix based on the preset delay compensation and the trigger signal specifically involves: Based on a preset sampling circuit or sampling channel, determine the first delay information and the second delay information; Apply positive time-shift compensation of the first delay information to the first level information; Apply positive time-shift compensation of the second delay information to the first square wave information; The compensated first level information and first square wave information are combined with the first voltage information and integrated into a trigger signal matrix according to the time axis. The row vectors of the trigger signal matrix correspond to the first level sequence, the first square wave sequence and the first voltage sequence, respectively.

8. The adaptive multi-source level wake-up system based on the HiSilicon platform USB according to claim 6, characterized in that, The step of extracting feature signals based on the trigger signal matrix specifically involves: The first level sequence is extracted based on a preset sliding window, and the slope value is extracted as the first slope information according to the first-order differential operation. The first square wave sequence is extracted based on a preset time window, and the first frequency information is obtained by zero-crossing detection. The root mean square (RMS) calculation is performed on the first voltage sequence, and the RMS result is normalized according to a preset method to obtain the first energy information.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer-readable storage medium includes an adaptive multi-source level wake-up method program based on the HiSilicon platform USB. When the adaptive multi-source level wake-up method program based on the HiSilicon platform USB is executed by a processor, it implements the steps of the adaptive multi-source level wake-up method based on the HiSilicon platform USB as described in any one of claims 1 to 5.

Images