A power supply method and device for a multimodal smart pen
By linking the timing of multi-channel sensor signals and preparing power supply status, the problems of power supply control delay and confusion in multimodal smart pens are solved, achieving efficient and stable multi-level power supply switching, and improving user experience and battery life.
Patent Information
- Application Number
- CN202611277975.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing multimodal smart pens suffer from issues such as passive response delay, chaotic single-channel wake-up signal processing, voltage interference during gear switching, and coarse power supply control granularity, resulting in a choppy user experience and power consumption loss.
By correlating the timing of multi-channel sensor signals, the system can predict scene transitions in advance and prepare the power supply status. It can also achieve multi-level gradient power supply by using adjustable DC/DC conversion modules and branch power supply modules. Combined with timing arbitration and behavior prediction, it can ensure that the power supply status is stable before the load changes, eliminating voltage fluctuations and power surges at the moment of switching.
It achieves zero-lag response in high-frequency scene switching of the multimodal smart pen, ensures the accuracy of sensor acquisition, reduces the average standby power consumption by more than 20%, and improves the smoothness of use and battery life.
Smart Images

Figure CN123122234A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power supply control technology for smart pens, and specifically relates to a power supply method and device for a multimodal smart pen. Background Technology
[0002] Multimodal smart pens integrate multiple functional modules such as pressure sensing, voice acquisition, visual recognition, and proximity sensing. The power consumption and voltage regulation requirements of each module vary significantly under different interaction scenarios. Existing mainstream power supply controls all adopt an event-triggered passive response architecture: the controller only executes module switching and output voltage level adjustment operations after detecting actual changes in load current and voltage. This control logic has the following problems under the high-frequency switching conditions of smart pens: 1. Passive response inherent delay: Existing solutions all need to wait for the actual load change before starting power supply adjustment. In high-frequency scenario switching such as writing start and voice activation of multimodal smart pens, the response delay of 50ms to 200ms causes a perceptible functional lag for users. 2. Single-channel wake-up signal: When multiple wake-up signals, such as those from proximity sensors, voice VAD (Voice Activity Detection), and touch sensing, arrive sequentially within a short time window, existing solutions respond one by one in order, leading to chaotic power supply timing of multiple functional modules and power surges. Existing wake-up solutions only process a single touch signal and do not utilize the timing correlation between multiple signals such as proximity, voice, and pressure sensing, making it impossible to predict upcoming scene transitions. For example, the invention patent with publication number CN114167971A discloses a low-power sleep wake-up device for electronic devices, which relies solely on a single touch signal to trigger the unified power-on of the entire device, failing to achieve differentiated pre-power supply for different modules.
[0003] 3. Voltage interference during gear switching: When the DC / DC converter switches gears after a load change, the voltage fluctuation during the voltage regulation transition phase will generate voltage fluctuations in the range of microseconds to milliseconds, interfering with the acquisition accuracy of the pressure sensor and the low-noise voice ADC (Analog-to-Digital Converter). For example, the invention patent with publication number CN119356509A discloses a load real-time feedback type adaptive power supply system that relies on real-time sampling current for closed-loop passive adjustment. The adjustment action lags behind the load change, and the voltage fluctuation during gear switching cannot be eliminated, resulting in a persistent detection response delay.
[0004] 4. Coarse power supply control granularity: Existing modular control schemes use binary on / off control, which cannot achieve a smooth transition from deep sleep to full power operation. Frequent full-scale switching leads to unnecessary power loss and power surges. For example, the invention patent with publication number CN109256826A discloses a modular overvoltage and overcurrent protection power supply device, which uses fixed threshold binary switching control logic. It can only realize module on / off and does not support multi-level gradient power supply adjustment based on continuous distance.
[0005] In summary, the existing technology has the following problems: passive response with response delay, only supports single-channel wake-up recognition, voltage interference during gear switching, and coarse power supply control granularity. Summary of the Invention
[0006] To address the above problems, this invention provides a power supply method for a multimodal smart pen, comprising the following steps: The signal acquisition unit synchronously sends the acquired multi-channel sensor signals to the behavior prediction unit and gradient mapping unit, and sends the multi-channel raw wake-up signals to the timing arbitration unit; The gradient mapping unit extracts the continuous distance of multiple sensor signals, determines the current power supply level based on the multi-level gradient power supply mode and the continuous distance, and sends the current power supply level to the pre-power supply control unit. The behavior prediction unit determines the scene prediction confidence and advance preparation time T_pre based on the time series of multi-channel sensor signals and the scene precursor pattern library. After the confidence meets the requirements, it sends a preparation command to the pre-power supply control unit. The timing arbitration unit generates a time-division activation sequence within the configured fusion time window based on multiple original wake-up signals, and sends the orderly power-on command to the pre-power supply control unit. Based on the current power supply level, preparation command, and orderly power-on command, the pre-power supply control unit sends a voltage pre-switching command to the adjustable DC / DC conversion module within the advance preparation time T_pre to adjust the output level in advance, and sends an LDO pre-activation command to the branch power supply module to preheat the corresponding LDO regulator in advance. The adjustable DC / DC converter module maintains the base output voltage according to the current power supply level, and switches to the target voltage level in advance when it receives a pre-switching command.
[0007] Furthermore, it also includes the following steps: If the actual scene signal does not arrive within the advance preparation time T_pre, the preparation instruction is canceled, the current state is returned, and the misjudgment data is uploaded to the main control module for subsequent updates to the scene precursor mode library.
[0008] Furthermore, the advance preparation time includes: the advance preparation time is equal to the sum of the time required for the adjustable DC / DC converter module to switch from the current operating level to the target level, the target LDO setup time, and the preset safety margin.
[0009] Furthermore, the scenario precursor pattern library includes: Writing start pre-processor: When the distance value decreases monotonically and the pressure sensor is not triggered, the prediction is a writing scenario, and the confidence level increases monotonically as the distance decreases. Voice interaction precursor: When the energy value of VAD frames exceeds the silence threshold for multiple consecutive frames, it is predicted to be a voice scene; Hibernation Precursor: When the distance value is continuously greater than the distance threshold and the VAD energy is less than the silence threshold for a duration greater than the preset duration, a hibernation scenario is predicted.
[0010] Furthermore, the current power supply level is determined based on the multi-level gradient power supply mode and the continuous distance, including: When the continuous distance is greater than 10cm, it is determined to be in deep sleep mode, and only the timing buffer circuit of the proximity sensor and behavior prediction unit is powered. When the continuous distance is greater than 5cm and less than or equal to 10cm, it is determined to be in the light preparation mode. The main control module switches from deep sleep to low power standby mode, the adjustable DC / DC conversion module maintains PFM mode, and the branch power supply module is in the preheating state. When the continuous distance is greater than 1cm and less than or equal to 5cm, it is determined to be the preheating activation level. The power supply module and the LDO regulator that powers the pressure sensor complete the pre-activation. The output voltage of the adjustable DC / DC converter module is pre-adjusted to the first target voltage, and the pressure sensor is in a low-power standby state. When the continuous distance is less than or equal to 1cm, it is determined to be in full power mode. The adjustable DC / DC conversion module switches to PWM mode and outputs the second target voltage. When the second target voltage is greater than the first target voltage, the pressure sensor enters the full power sampling state.
[0011] Furthermore, after the adjustable DC / DC converter module maintains the base output voltage according to the current power supply level and switches to the target voltage level in advance upon receiving a pre-switching command, it also includes: The branch power supply module maintains a basic low power consumption state according to the current power supply level and performs independent power-off control on idle modules.
[0012] Furthermore, it also includes the following steps: When the pre-power supply control unit receives a fault signal from the protection module, it immediately terminates all pre-power supply operations and resets all power supply outputs.
[0013] The present invention also provides a power supply device for a multimodal smart pen, including a signal acquisition unit, a behavior prediction unit, a pre-power supply control unit, a gradient mapping unit, a timing arbitration unit, an adjustable DC / DC conversion module, a branch power supply module, and a main control module; The signal acquisition unit is electrically connected to the input terminal of the behavior prediction unit, the output terminal of the behavior prediction unit is electrically connected to the input terminal of the pre-power supply control unit, and the parameter configuration terminal of the behavior prediction unit is electrically connected to the main control module; the output terminal of the pre-power supply control unit is electrically connected to the control terminal of the adjustable DC / DC conversion module and the control terminal of the branch power supply module. The input terminal of the gradient mapping unit is electrically connected to the output terminal of the signal acquisition unit, and the output terminal of the gradient mapping unit is electrically connected to the pre-power supply control unit and the main control module. The input terminal of the timing arbitration unit is electrically connected to the output terminal of the signal acquisition unit, and the output terminal of the timing arbitration unit is electrically connected to the pre-power supply control unit and the main control module. The control terminal of the adjustable DC / DC converter module is also electrically connected to the main control module, the output terminal of the adjustable DC / DC converter module is electrically connected to the input terminal of the branch power supply module, and the control terminal of the branch power supply module is also electrically connected to the main control module.
[0014] Furthermore, the behavior prediction unit includes a timing buffer register group, a scene matching logic circuit, a preparation time calculation unit, and a scene confidence estimator; Among them, the timing buffer register group is used to continuously receive multiple signals output by the signal acquisition unit and maintain the time series of the multiple signals; The scene matching logic circuit has a built-in multimodal scene precursor pattern library. The scene matching logic circuit, combined with the scene confidence evaluator, compares and matches the real-time cached time series with the pattern library, and synchronously outputs the predicted scene code and the corresponding scene prediction confidence. The preparation time calculation unit is used to calculate the advance preparation time T_pre.
[0015] Furthermore, the pre-power supply control unit includes a pre-command decoder, a graded power supply timing generator, and a T_pre countdown timer; Among them, the preparatory instruction decoder is used to receive the preparatory instructions output by the behavior prediction unit; The graded power supply timing generator is used to send voltage level pre-switching commands to the adjustable DC / DC conversion module and LDO pre-activation commands to the branch power supply module in sequence according to the activation priority sequence before the T_pre countdown timer is triggered. The gradient mapping unit incorporates a distance-state mapping table, a hysteresis comparator, and a gradient state machine. Among them, the distance state mapping table is used to map the continuous distance output by the proximity sensor to a multi-level gradient power supply mode; The hysteresis comparator is used to set upper and lower threshold hysteresis intervals based on the continuous analog distance output by the proximity sensor, and output a distance determination signal. The gradient state machine is used to receive the distance determination signal output by the hysteresis comparator and manage the switching and holding logic of the power supply level in the multi-level gradient power supply mode, ensuring that the state switching is orderly and mutually exclusive, and that only one power supply level is maintained at any given time.
[0016] The beneficial effects of this invention are: 1. The present invention determines the scene prediction confidence and advance preparation time T_pre based on the time series of multi-channel sensor signals and the scene precursor mode library. The power supply state preparation is completed in advance before the scene switching occurs, thus solving the problem of response delay.
[0017] 2. Because the power supply status of this invention is prepared before the load changes, the adjustable DC / DC conversion module is in a stable output state when the target load arrives, eliminating the voltage fluctuation at the moment of switching in the passive solution and ensuring the acquisition accuracy of sensitive modules such as pressure sensing and voice ADC.
[0018] 3. The timing arbitration unit of the present invention generates a time-division activation sequence for multiple wake-up signals within the configured fusion time window, and activates the corresponding modules in sequence and differentiates them to prevent surge impact caused by multiple modules being powered on at the same time.
[0019] 4. The gradient mapping unit of the present invention extracts the continuous distance of multiple sensor signals and determines the current power supply level according to the multi-level gradient power supply mode and the continuous distance, replacing the coarse-grained binary switch control, realizing a smooth transition from deep sleep to full power operation, and reducing the average standby power consumption by more than 20% compared with the existing solution.
[0020] 5. The advance preparation time T_pre of the present invention is dynamically calculated based on the current gear, target gear and output capacitor parameters of the adjustable DC / DC conversion module, ensuring that the preparation action can be accurately completed before scene switching under different working conditions.
[0021] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description and the drawings. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 A schematic diagram of the power supply device for a multimodal smart pen according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of the main flow of a power supply method for a multimodal smart pen according to an embodiment of the present invention is shown.
[0024] In the diagram: 1. Power input module; 2. Signal acquisition unit; 3. Behavior prediction unit; 4. Pre-power supply control unit; 5. Gradient mapping unit; 6. Timing arbitration unit; 7. Adjustable DC / DC conversion module; 8. Branch power supply module; 9. Main control module; 10. Protection module. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.
[0026] It should be noted that the terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein.
[0027] To address the shortcomings of existing passive power management solutions for multimodal smart pens, this invention proposes a power supply device for multimodal smart pens. Through real-time analysis of the timing correlation of multiple sensor signals, it anticipates upcoming interaction scenarios and pre-configures the power supply status before load switching, thereby achieving the following core objectives: 1. Eliminate the perceptible delay caused by power supply switching and achieve zero-lag interaction.
[0028] 2. Multi-channel wake-up signal timing fusion arbitration, time-sharing start module, avoiding power-on surge impact; 3. Suppress instantaneous voltage fluctuations during DC / DC range switching to ensure the sampling stability of high-precision sensors.
[0029] 4. By relying on the continuous measurement of distance sensing, multi-level gradient fine power supply is achieved, reducing the average standby power consumption of the whole machine.
[0030] In view of at least one problem existing in the prior art, such as Figure 1As shown, this embodiment of the invention provides a power supply device for a multimodal smart pen, including a power input module 1, a signal acquisition unit 2, a behavior prediction unit 3, a pre-power supply control unit 4, a gradient mapping unit 5, a timing arbitration unit 6, an adjustable DC / DC conversion module 7, a branch power supply module 8, and a main control module 9.
[0031] The output of the power supply input module 1 is electrically connected to the input of the adjustable DC / DC converter module 7; the signal acquisition unit 2 is electrically connected to the input of the behavior prediction unit 3; the output of the behavior prediction unit 3 is electrically connected to the input of the pre-power supply control unit 4; and the parameter configuration terminal of the behavior prediction unit 3 is electrically connected to the main control module 9. The output of the pre-power supply control unit 4 is electrically connected to the control terminal of the adjustable DC / DC converter module 7 and the control terminal of the branch power supply module 8.
[0032] The input terminal of gradient mapping unit 5 is electrically connected to the output terminal of signal acquisition unit 2, and the output terminal of gradient mapping unit 5 is electrically connected to pre-power supply control unit 4 and main control module 9. The input terminal of timing arbitration unit 6 is electrically connected to the output terminal of signal acquisition unit 2, and the output terminal of timing arbitration unit 6 is electrically connected to pre-power supply control unit 4 and main control module 9.
[0033] The control terminal of the adjustable DC / DC converter module 7 is also electrically connected to the main control module 9, the output terminal of the adjustable DC / DC converter module 7 is electrically connected to the input terminal of the branch power supply module 8, and the control terminal of the branch power supply module 8 is also electrically connected to the main control module 9.
[0034] For example, the power supply device also includes a protection module 10, which is connected in series between the power input module 1 and the adjustable DC / DC conversion module 7, and the signal terminal of the protection module 10 is electrically connected to the main control module 9.
[0035] The protection module is used to implement overvoltage, overcurrent, overtemperature and reverse connection protection of the power supply device. The protection action signal is fed back to the main control module 9. It is linked with the pre-power supply control unit 4. When a protection event is detected, all preparatory commands are immediately revoked and the power supply is cut off.
[0036] For example, the power input module 1 includes a Type-C charging interface, a lithium battery (100~300mAh), and a charging management chip (such as BQ25100). The Type-C charging interface is electrically connected to the lithium battery via the charging management chip, and the output terminal of the lithium battery is electrically connected to the input terminal of the adjustable DC / DC conversion module 7.
[0037] The Type-C charging interface connects to an external Type-C power source to charge the built-in lithium battery, while also providing a stable initial input voltage (3.7V~4.2V) for the entire power supply device; the charging management chip implements constant current and constant voltage charging control and charging protection.
[0038] For example, signal acquisition unit 2 includes a distance ADC acquisition circuit, a voice VAD energy detection circuit, a pressure-sensitive signal edge detection circuit, and a touch signal level detection circuit.
[0039] Each circuit signal line of the signal acquisition unit 2 is electrically connected to the corresponding signal input terminal of the behavior prediction unit 3. The distance ADC acquisition circuit is used to acquire the distance analog quantity output by the proximity sensor in real time (accuracy ≤1mm, sampling period ≤5ms) and input it to the behavior prediction unit 3. The distance analog quantity is converted into a digital quantity by the distance ADC acquisition circuit (12-bit ADC) and input to the behavior prediction unit 3.
[0040] The voice VAD energy detection circuit is used to collect the frame energy value output by the voice VAD module and input it to the behavior prediction unit 3; the pressure-sensitive signal edge detection circuit is used to collect the trigger edge signal of the pressure sensor and input it to the behavior prediction unit 3; the touch signal level detection circuit is used to collect the level state of the touch sensor and input it to the behavior prediction unit 3.
[0041] For example, the parameter configuration terminal of behavior prediction unit 3 is electrically connected to the I2C interface of main control module 9. Behavior prediction unit 3 includes a timing cache register group, scene matching logic circuit, preparation time calculation unit, and scene confidence evaluator.
[0042] The timing buffer register group stores the signal sequences of each sensor within the last 200ms. The timing buffer register group is used to continuously receive multiple signals output by the signal acquisition unit 2 and maintain the time sequence of the multiple signals to ensure that the timestamp accuracy of the time sequence is ≤1ms.
[0043] The scene matching logic circuit has a built-in multimodal scene precursor pattern library. The pattern library pre-stores four types of scenes: writing start, voice interaction, visual scanning, and sleep, as well as the corresponding sensing timing features of each scene. The scene matching logic circuit, combined with the scene confidence evaluator, compares and matches the real-time cached timing sequence with the pattern library, and synchronously outputs the predicted scene code and the corresponding scene prediction confidence.
[0044] The preparation time calculation unit is used to calculate the advance preparation time T_pre.
[0045] For example, the scene matching logic circuit can also use a low-power main control chip (MCU) as the carrier, and burn the program code (prediction algorithm) for scene matching, timing analysis and advance duration calculation into the firmware of the MCU; the chip reads multi-channel sensor data in real time and runs the algorithm to complete scene prediction.
[0046] For example, the pre-power supply control unit 4 includes a pre-instruction decoder, a graded power supply timing generator, and a T_pre countdown timer.
[0047] The preparatory instruction decoder is used to receive the preparatory instructions output by the behavior prediction unit 3. The preparatory instructions include the prediction scene code, the T_pre value, and the activation priority sequence of each module.
[0048] The graded power supply timing generator is used to send voltage level pre-switching instructions to the adjustable DC / DC conversion module 7 and LDO pre-activation instructions to the branch power supply module 8 in sequence according to the activation priority sequence before the T_pre countdown timer is triggered, so that each target module is in a stable power supply state when the actual scenario arrives.
[0049] The pre-activation timing interval of the adjustable DC / DC conversion module 7 and the branch power supply module 8 is configurable (default 2ms) to prevent surge impact caused by multiple modules being powered on at the same time; when the actual scenario signal arrives, the pre-power supply control unit 4 confirms that the power supply status has stabilized and sends a pre-completion confirmation signal to the main control module 9.
[0050] For example, the gradient mapping unit 5 has a built-in distance state mapping table (fixed ROM), a hysteresis comparator (anti-jitter), and a gradient state machine.
[0051] The distance state mapping table is used to map the continuous distance quantity D output by the proximity sensor to a multi-level gradient power supply mode (4 levels), and each level is defined as follows: Deep sleep mode (D>10cm): Only the proximity sensor and the timing buffer circuit of behavior prediction unit 3 are powered, and the standby current of the whole machine is <5μA.
[0052] Light preparatory mode (5cm < D ≤ 10cm): The main control module 9 switches from deep sleep to low power standby mode, the adjustable DC / DC conversion module 7 maintains PFM mode, and all LDO regulators of the branch power supply module 8 are in preheating state (low power standby, setup time < 0.5ms). Preheating activation level (1cm<D≤5cm): The LDO regulator of the branch power supply module 8, which supplies power to the pressure sensor, completes pre-activation. The output voltage of the adjustable DC / DC conversion module 7 is pre-adjusted to the first target voltage (1.8V). The pressure sensor is in a low-power standby state and can switch to full-power sampling within <2ms.
[0053] Full power mode (D≤1cm): Adjustable DC / DC conversion module 7 switches to PWM mode and outputs the second target voltage (e.g., 3.3V). All working modules are powered and the pressure sensor enters the full power sampling state.
[0054] The hysteresis comparator is used to set upper and lower threshold hysteresis intervals based on the continuous analog distance output by the proximity sensor, and output a distance judgment signal to avoid frequent switching of power supply status (jitter, frequent power skipping) when the distance value fluctuates slightly at the graded critical point.
[0055] For example, a hysteresis interval of ±0.5cm is set between each adjacent level to prevent the distance value from fluctuating near the threshold and causing frequent state machine flips.
[0056] The gradient state machine serves as the timing control logic for the graded power supply state. It receives the distance determination signal output by the hysteresis comparator and manages the switching and holding logic of the power supply levels (deep sleep level, light preparation level, preheating activation level, and full power level) of the multi-level gradient power supply mode. This ensures that the state switching is orderly and mutually exclusive, and only one power supply level is maintained at any given time.
[0057] For example, timing arbitration unit 6 includes a multi-channel signal timestamp latch, a priority weight register, and arbitration logic circuitry.
[0058] The multi-signal timestamp latch is used to perform timing latching and priority arbitration of multiple wake-up signals (proximity, voice VAD, touch) that arrive simultaneously or sequentially within a configurable time fusion window T_fuse (default 50ms).
[0059] The priority weight register is used to preset priority weights. The preset priority weights include prioritizing the activation of the pressure-sensitive and visual modules in the writing scenario corresponding to proximity signals, prioritizing the activation of the voice module in the voice signal scenario, and prioritizing the activation of the display and main control modules 9 in the touch signal scenario.
[0060] The arbitration logic circuit is used to generate a time-sharing activation sequence according to the preset priority weights and send the orderly power-on command to the pre-power supply control unit 4 to ensure that only one module is activated at a time, with an interval of ≥2ms, to prevent surge impact caused by multiple modules being powered on at the same time.
[0061] For example, the control terminal of the adjustable DC / DC converter module 7 is electrically connected to the pre-powered control unit 4 and the main control module 9 via an I2C interface. The adjustable DC / DC converter module 7 uses RT6160 and supports automatic PWM / PFM switching, including a gear control interface.
[0062] The adjustable DC / DC converter module 7 is used to convert the lithium battery voltage of the power input module 1 into a multi-level adjustable output voltage (0.7V, 1.2V, 1.8V, 3.3V). According to the instructions of the pre-power supply control unit 4 or the main control module 9, it switches to the target output voltage level in advance. When the target load arrives, the output of the adjustable DC / DC converter module 7 is already in a stable state, eliminating voltage fluctuations at the moment of switching.
[0063] For example, the branch power supply module 8 includes multiple independent LDO regulators (such as RT9193, which supports low power enable control) and MOSFET power switches. The LDO regulators can be RT9193, which supports low power enable control, and the MOSFET power switches can be AW9163.
[0064] Each LDO regulator provides an independent secondary regulated power supply for one functional module (pressure sensing, voice, vision, proximity) of the multimodal smart pen.
[0065] The MOSFET power switch is used to preheat the LDO regulator corresponding to the target function module in advance according to the pre-activation command of the pre-power supply control unit 4, so that it can enter a stable working state without waiting for the LDO establishment time when the scenario arrives; it is also used to perform independent power-off control on idle modules.
[0066] For example, the main control module 9 uses a low-power MCU (RISC-V architecture, such as CH32V103), and the main control module 9 has built-in multimodal behavior prediction algorithm firmware.
[0067] The main control module 9 serves as the core of system coordination, maintaining the global power supply state machine, receiving the outputs of the behavior prediction unit 3, gradient mapping unit 5, and timing arbitration unit 6, and coordinating the power supply strategies of each module; it also carries the storage and dynamic updates of the multimodal scenario precursor mode library (the scenario prediction model parameters can be upgraded via OTA).
[0068] For example, the protection module 10 includes an overvoltage protection chip, an overcurrent detection resistor, an overtemperature sensor, and a reverse connection protection MOSFET.
[0069] The overvoltage protection chip monitors the output voltage of the power input module 1 and the adjustable DC / DC converter module 7 in real time. When the output voltage of the lithium battery or the adjustable DC / DC converter module 7 exceeds the chip's set safety threshold, it immediately cuts off the power supply path to the downstream stage and simultaneously outputs a fault feedback signal to the main control module 9. This prevents high-precision sensitive chips such as pressure sensors and voice ADCs from burning out due to excessive voltage, and avoids component breakdown and damage.
[0070] An overcurrent detection resistor is connected in series between the power input module 1 and the adjustable DC / DC conversion module 7. It collects the circuit current in real time by relying on the voltage divider of the resistor. When a short circuit or module overload occurs in the circuit, the voltage difference across the resistor increases sharply. The system identifies the overcurrent fault and links the pre-power supply control unit 4 to cancel all pre-activation commands and cut off the power supply, thereby suppressing the risk of heat generation and device burnout caused by high current.
[0071] The over-temperature sensor is used to collect the temperature of the charging management chip and lithium battery area in real time. When the temperature exceeds the rated operating limit of the device, protection logic is triggered. High temperatures can accelerate battery aging and cause voltage regulator chip performance failure. This device can promptly report overheating faults, cut off the power supply circuit, and avoid safety hazards such as fire and cell bulging.
[0072] A reverse connection protection MOSFET is connected in series at the front end of the Type-C charging interface. If the user inserts the charging cable with the positive and negative terminals reversed, the MOSFET will automatically turn off, blocking the reverse current from flowing into the internal circuit. This prevents the charging management chip and lithium battery from being directly damaged by the reverse voltage, ensuring circuit safety when plugging and unplugging the charging interface.
[0073] The working process of the power supply device in this embodiment of the invention is as follows: 1. Deep sleep state: Only the proximity sensor and the timing buffer circuit of the multimodal behavior prediction unit 3 are powered, and the standby current of the whole machine is <5μA.
[0074] 2. Distance Trigger Preparation: When the user picks up the smart pen, the proximity sensor detects that the distance value begins to decrease, and the gradient mapping unit 5 sequentially triggers the gradient switching from the mild preparation level to the warm-up activation level.
[0075] 3. Advance power supply preparation: When the behavior prediction unit 3 detects the writing start pre-drive mode (distance monotonically decreases, confidence level > 70%), it controls the adjustable DC / DC conversion module 7 to complete the DC / DC range switching in advance via T_pre, and the branch power supply module 8 completes the pressure-sensitive LDO pre-activation. 4. Writing scenario: When the pen tip touches the paper, the pressure-sensitive module is already in a stable power supply state, without waiting for any setup time, achieving zero-perceptible delay response.
[0076] 5. Multimodal concurrent processing: When the user triggers voice wake-up during writing, the timing arbitration unit 6 recognizes the concurrent signals of writing (pressure sensor active) and voice (VAD energy increased). According to the arbitration priority, while maintaining the power supply to the pressure sensor, the voice module is activated after a delay of 2ms to prevent the two modules from being powered on at the same time.
[0077] 6. Hibernation Switching: After no operation for more than the preset timeout, the behavior prediction unit 3 recognizes the hibernation precursor mode, and the adjustable DC / DC conversion module 7 is controlled in advance to complete the step-down preparation. Each module is powered off in gradient sequence, and finally enters the deep hibernation state.
[0078] The power supply device of the multimodal smart pen in this invention is adapted to multimodal learning scenarios such as writing, voice interaction, and visual scanning. It focuses on solving the problems of power supply response lag and power surge during high-frequency switching of multimodal scenarios, thereby improving the smoothness of use and battery life of the smart pen.
[0079] Correspondingly, such as Figure 2As shown in the figure, this embodiment of the invention also provides a power supply method for a multimodal smart pen, realizing coordinated control of multimodal scene prediction and advanced power supply preparation, including the following steps: S1. The electrical energy output from the power input module 1 is transmitted to the adjustable DC / DC conversion module 7 after passing through the protection module 10, so as to power the proximity sensor and behavior prediction unit 3.
[0080] S2, Main Control Module 9 stores the scenario predecessor pattern library. For example, the scenario predecessor pattern library includes the following typical predecessor sequences: Writing start precondition: When the distance value decreases monotonically (slope ≥ 0.5cm / 10ms) and pressure sensitivity is not triggered, the prediction is a writing scenario, and the confidence level increases monotonically as the distance decreases.
[0081] Voice interaction precursor: When the energy value of VAD frames exceeds the silence threshold for multiple consecutive frames (e.g., 3 frames), it is predicted to be a voice scene; Hibernation Precursor: When the distance value is continuously greater than the distance threshold (10cm) and the VAD energy is less than the silence threshold, and the duration is greater than the preset duration (default 5s), the scenario is predicted to be hibernation.
[0082] S3, Signal Acquisition Unit 2 synchronously sends the acquired multi-channel sensor signals to Behavior Prediction Unit 3 and Gradient Mapping Unit 5, and sends the multi-channel raw wake-up signals to Timing Arbitration Unit 6.
[0083] For example, signal acquisition unit 2 acquires proximity sensor distance values and voice VAD frame energy values at a period of ≤5ms, and refreshes the timing buffer register group in real time (buffering the signal sequence of the most recent 200ms).
[0084] S4 and gradient mapping unit 5 determine the current power supply level based on the continuous distance of the multi-channel sensor signals and the multi-level gradient power supply mode, and transmit the current power supply level to the pre-power supply control unit 4 and the main control module 9.
[0085] For example, determining the current power supply level based on multi-level gradient power supply mode and continuous distance includes: When the continuous distance is greater than 10cm, it is determined to be in deep sleep mode, and only the proximity sensor and the timing cache circuit of behavior prediction unit 3 are powered. The standby current of the whole machine is <5μA.
[0086] When the continuous distance is greater than 5cm and less than or equal to 10cm, it is determined to be in the light preparation mode. The main control module 9 switches from deep sleep to low power standby mode, the adjustable DC / DC conversion module 7 maintains PFM mode, and all LDO regulators of the branch power supply module 8 are in the preheating state.
[0087] When the continuous distance is greater than 1cm and less than or equal to 5cm, it is determined to be the preheating activation level. The LDO regulator of the power supply module 8 for the pressure sensor completes the pre-activation. The output voltage of the adjustable DC / DC conversion module 7 is pre-adjusted to the first target voltage, and the pressure sensor is in a low-power standby state. When the continuous distance is less than or equal to 1cm, it is determined to be in full power mode. The adjustable DC / DC conversion module 7 switches to PWM mode and outputs the second target voltage. When the second target voltage is greater than the first target voltage, the pressure sensor enters the full power sampling state.
[0088] S5. The behavior prediction unit 3 determines the scene prediction confidence and advance preparation time T_pre based on the time series of multi-channel sensor signals and the scene precursor mode library. After the confidence meets the requirements, it sends a preparation command to the pre-power supply control unit 4. If the actual scene signal does not arrive within the advance preparation time T_pre, the preparation command is canceled, the current state is returned, and the misjudgment data is uploaded to the main control module 9 for subsequent updates of the scene precursor mode library.
[0089] For example, determining the advance preparation time T_pre includes: the advance preparation time T_pre is equal to the sum of the time required for the adjustable DC / DC converter module 7 to switch from the current operating position to the target position T_DCDC, the target LDO setup time T_LDO, and the preset safety margin T_margin, as detailed below: T_pre = T_DCDC + T_LDO + T_margin The safety margin T_margin is 5ms by default, and T_DCDC is dynamically calculated based on the output capacitor value and the gear difference, with a typical value of 5ms to 20ms.
[0090] S6 and timing arbitration unit 6 perform timestamp latching and priority sorting within the configured fusion time window based on multiple original wake-up signals, generate a time-division activation sequence, and send the orderly power-on command to the pre-power supply control unit 4 to force the power-on interval of each module to be ≥2ms, thereby eliminating power surges during synchronous power-on.
[0091] S7, the pre-power supply control unit 4, based on the current power supply level, preparation command, and orderly power-on command, within the advance preparation time T_pre, sends a voltage pre-switching command to the adjustable DC / DC conversion module 7 to adjust the output level in advance, and sends an LDO pre-activation command to the branch power supply module 8 to preheat the corresponding LDO regulator in advance; when the pre-power supply control unit 4 receives a fault signal from the protection module 10, it immediately terminates all pre-power supply operations and resets all power supply outputs.
[0092] S8. The adjustable DC / DC conversion module 7 maintains the basic output voltage according to the current power supply level; after receiving the level pre-switching command from the pre-power supply control unit 4, it switches to the target voltage level in advance, including: the adjustable DC / DC conversion module 7 switches the PWM / PFM working mode in advance, adjusts the output DC voltage, and switches to the target voltage level before the actual load change, thus completing voltage regulation and eliminating voltage switching fluctuations.
[0093] S9 and the branch power supply module 8 maintain a basic low power consumption state according to the current power supply level, and perform independent power-off control on idle modules to achieve independent power consumption management of branch circuits.
[0094] S10, the main control module 9 stores and updates the scenario predecessor mode library, configures the gradient threshold, fusion window, and T_pre margin. The main control module 9 receives the output of the behavior prediction unit 3, the gradient mapping unit 5, and the timing arbitration unit 6, and monitors the power consumption and working status of the power supply device. When a prediction misjudgment, overvoltage / overcurrent / overtemperature / reverse connection fault occurs, a global reset command is issued to force the power supply device to return to the deep sleep mode.
[0095] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A power supply method of a multi-modal smart pen, characterized by, Includes the following steps: The signal acquisition unit (2) synchronously sends the acquired multi-channel sensor signals to the behavior prediction unit (3) and the gradient mapping unit (5), and sends the multi-channel original wake-up signals to the timing arbitration unit (6); The gradient mapping unit (5) extracts the continuous distance of the multi-channel sensor signals, determines the current power supply level according to the multi-level gradient power supply mode and the continuous distance, and sends the current power supply level to the pre-power supply control unit (4). The behavior prediction unit (3) determines the scene prediction confidence and advance preparation time T_pre based on the time series of multi-channel sensor signals and the scene precursor pattern library. After the confidence meets the requirements, it sends a preparation command to the pre-power supply control unit (4). The timing arbitration unit (6) generates a time-division activation sequence within the configured fusion time window based on multiple original wake-up signals and sends the orderly power-on command to the pre-power supply control unit (4). According to the current power supply level, preparation command and orderly power-on command, the pre-power supply control unit (4) sends a voltage pre-switching command to the adjustable DC / DC conversion module (7) within the advance preparation time T_pre, adjusts the output level in advance, and sends an LDO pre-activation command to the branch power supply module (8), so that the branch power supply module (8) can preheat the corresponding LDO regulator in advance. The adjustable DC / DC converter module (7) maintains the basic output voltage according to the current power supply level, and switches to the target voltage level in advance when it receives the pre-switching command. 2.The power supply method of a multi-modal smart pen according to claim 1, characterized in that, It also includes the following steps: If the actual scene signal does not arrive within the advance preparation time T_pre, cancel the preparation instruction, return to the current state, and upload the misjudged data to the main control module (9) for subsequent updates of the scene precursor mode library. 3.The power supply method of a multi-modal smart pen according to claim 1, characterized in that, The advance preparation time includes: the advance preparation time is equal to the sum of the time required for the adjustable DC / DC converter module (7) to switch from the current working position to the target position, the target LDO establishment time, and the preset safety margin. 4.The power supply method of a multi-modal smart pen according to claim 1, wherein, The scene precursor pattern library includes: Writing start pre-processor: When the distance value decreases monotonically and the pressure sensor is not triggered, the prediction is a writing scenario, and the confidence level increases monotonically as the distance decreases. Voice interaction precursor: When the energy value of VAD frames exceeds the silence threshold for multiple consecutive frames, it is predicted to be a voice scene; Hibernation Precursor: When the distance value is continuously greater than the distance threshold and the VAD energy is less than the silence threshold for a duration greater than the preset duration, a hibernation scenario is predicted. 5.The power supply method of a multi-modal smart pen according to claim 1, wherein, The current power supply level is determined based on the multi-level gradient power supply mode and the continuous distance, including: When the continuous distance is greater than 10cm, it is determined to be in deep sleep mode, and only the timing buffer circuit of the proximity sensor and behavior prediction unit (3) is powered. When the continuous distance is greater than 5cm and less than or equal to 10cm, it is determined to be a light preparatory gear. The main control module (9) switches from deep sleep to low power standby mode, the adjustable DC / DC conversion module (7) maintains PFM mode, and the branch power supply module (8) is in preheating state. When the continuous distance is greater than 1cm and less than or equal to 5cm, it is determined to be the preheating activation level. The branch power supply module (8) and the LDO regulator that supplies power to the pressure sensor complete the pre-activation. The output voltage of the adjustable DC / DC conversion module (7) is pre-adjusted to the first target voltage, and the pressure sensor is in a low-power standby state. When the continuous distance is less than or equal to 1cm, it is determined to be the full power level. The adjustable DC / DC conversion module (7) switches to PWM mode and outputs the second target voltage. The second target voltage is greater than the first target voltage, and the pressure sensor enters the full power sampling state. 6.The power supply method of a multi-modal smart pen according to claim 1, wherein, The adjustable DC / DC converter module (7) maintains the base output voltage according to the current power supply level. After receiving the pre-switching command, it switches to the target voltage level in advance. The following steps also include: The branch power supply module (8) maintains a basic low power consumption state according to the current power supply level and performs independent power-off control on idle modules.
7. The power supply method of a multi-modal smart pen according to any one of claims 1-6, characterized in that, It also includes the following steps: When the pre-power supply control unit (4) receives a fault signal from the protection module (10), it immediately terminates all pre-power supply operations and resets all power supply outputs.
8. A power supply device for a multimodal smart pen, characterized in that, It includes a signal acquisition unit (2), a behavior prediction unit (3), a pre-power supply control unit (4), a gradient mapping unit (5), a timing arbitration unit (6), an adjustable DC / DC conversion module (7), a branch power supply module (8), and a main control module (9). Among them, the signal acquisition unit (2) is electrically connected to the input end of the behavior prediction unit (3), the output end of the behavior prediction unit (3) is electrically connected to the input end of the pre-power supply control unit (4), and the parameter configuration end of the behavior prediction unit (3) is electrically connected to the main control module (9); the output end of the pre-power supply control unit (4) is electrically connected to the control end of the adjustable DC / DC conversion module (7) and the control end of the branch power supply module (8); The input terminal of the gradient mapping unit (5) is electrically connected to the output terminal of the signal acquisition unit (2), the output terminal of the gradient mapping unit (5) is electrically connected to the pre-power supply control unit (4) and the main control module (9), the input terminal of the timing arbitration unit (6) is electrically connected to the output terminal of the signal acquisition unit (2), and the output terminal of the timing arbitration unit (6) is electrically connected to the pre-power supply control unit (4) and the main control module (9). The control terminal of the adjustable DC / DC converter module (7) is also electrically connected to the main control module (9), the output terminal of the adjustable DC / DC converter module (7) is electrically connected to the input terminal of the branch power supply module (8), and the control terminal of the branch power supply module (8) is also electrically connected to the main control module (9).
9. The power supply device for the multimodal smart pen according to claim 8, characterized in that, The behavior prediction unit (3) includes a timing buffer register group, a scene matching logic circuit, a preparation time calculation unit, and a scene confidence evaluator; Among them, the timing buffer register group is used to continuously receive multiple signals output by the signal acquisition unit (2) and maintain the time sequence of the multiple signals; The scene matching logic circuit has a built-in multimodal scene precursor pattern library. The scene matching logic circuit, combined with the scene confidence evaluator, compares and matches the real-time cached time series with the pattern library, and synchronously outputs the predicted scene code and the corresponding scene prediction confidence. The preparation time calculation unit is used to calculate the advance preparation time T_pre.
10. The power supply device for the multimodal smart pen according to claim 8, characterized in that, The pre-power supply control unit (4) includes a pre-instruction decoder, a graded power supply timing generator, and a T_pre countdown timer; Among them, the preparatory instruction decoder is used to receive the preparatory instruction output by the behavior prediction unit (3); The graded power supply timing generator is used to send voltage level pre-switching instructions to the adjustable DC / DC conversion module (7) and LDO pre-activation instructions to the branch power supply module (8) in sequence according to the activation priority sequence before the T_pre countdown timer is triggered. The gradient mapping unit (5) has a built-in distance-state mapping table, a hysteresis comparator, and a gradient state machine; Among them, the distance state mapping table is used to map the continuous distance output by the proximity sensor to a multi-level gradient power supply mode; The hysteresis comparator is used to set upper and lower threshold hysteresis intervals based on the continuous analog distance output by the proximity sensor, and output a distance determination signal. The gradient state machine is used to receive the distance determination signal output by the hysteresis comparator and manage the switching and holding logic of the power supply level in the multi-level gradient power supply mode, ensuring that the state switching is orderly and mutually exclusive, and that only one power supply level is maintained at any given time.
Citation Information
Patent Citations
Low power-consumption wake-up circuit, a power supply controller, a power supply device and a power tool
CN109256826A
Dormancy wake-up circuit, automatic walking device and wake-up method thereof
CN114167971A
Self-adaptive computer power supply system and method and terminal equipment
CN119356509A
