A lollipop Bluetooth speaker with NFC and AI control and a control method

Abstract

This invention belongs to the technical field of audio playback devices, specifically relating to a lollipop-shaped Bluetooth speaker with NFC and AI control, and its control method. The lollipop-shaped Bluetooth speaker includes a handle and a replaceable candy head. The handle has an internal cavity containing an NFC reading module, a pressure sensor, a tilt sensor, an AI inference and control module, a bone conduction speaker, an audio processing and amplifier module, and a power module. The NFC reading module reads the acoustic model parameters of the candy head, the pressure sensor collects the biting pressure value, the tilt sensor collects the handle tilt angle value, the AI ​​inference and control module combines the parameters and sensor data to calculate the frequency offset and generate a gain adjustment signal, and the audio processing and amplifier module adjusts the audio according to the signal and drives the bone conduction speaker to play. This invention can adaptively match the acoustic characteristics of the candy head with the user's posture, optimize the audio playback effect under oral resonance, and simultaneously achieve a balance between power consumption and sound quality under low battery conditions, improving the user experience.

Description

Technical Field

[0001] This invention belongs to the field of audio playback device technology, specifically relating to a lollipop Bluetooth speaker with NFC and AI control and its control method. Background Technology

[0002] Existing lollipop-shaped Bluetooth audio devices only offer basic Bluetooth playback and NFC information reading functions. Their audio output parameters are fixed preset values ​​and cannot adapt to the acoustic and physical characteristics of different lollipop heads. During use, changes in biting force and grip angle alter the structure of the oral cavity's resonance chambers, causing audio resonance frequency shifts and resulting in sound distortion and muddy timbre. Furthermore, when the battery is low, these devices cannot dynamically balance audio transmission bitrate, playback power consumption, and sound quality performance, easily leading to playback interruptions and sudden drops in sound quality, failing to meet users' needs for continuous and stable use.

[0003] Based on the above problems, there is an urgent need for a lollipop Bluetooth audio solution that can adaptively adjust according to the characteristics of the lollipop and the user's usage status. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a lollipop-shaped Bluetooth speaker with NFC and AI control. The speaker includes a handle and a replaceable candy head. The handle has an internal cavity. The lollipop Bluetooth speaker includes: an NFC reading module disposed within the handle for reading NFC tag information embedded in the candy head, the NFC tag information containing acoustic model parameters of the candy head; a pressure sensor deployed in the biting area of ​​the handle for collecting the user's biting pressure value; a tilt sensor deployed within the handle for collecting the tilt angle value of the handle; and an AI inference and control module disposed within the handle and electrically connected to the NFC reading module, the pressure sensor, and the tilt sensor. The inference and control module acquires the acoustic model parameters of the candy head, the biting pressure value, and the tilt angle value. Based on the differences between the acoustic model parameters, the biting pressure value, and the preset resting reference pressure baseline, and the differences between the tilt angle value and the preset reference tilt angle baseline, it calculates the frequency offset of the oral resonance cavity. It then generates a gain adjustment signal based on the frequency offset and a preset compensation strategy. A bone conduction speaker, located inside the handle and close to the inner wall of the cavity, converts electrical signals into vibration signals. An audio processing and power amplifier module, located inside the handle and electrically connected to the AI ​​inference and control module and the bone conduction speaker, receives the audio signal to be played, performs equalization processing on the audio signal based on the gain adjustment signal, amplifies it, and drives the bone conduction speaker to play the audio. A power supply module provides power to all modules.

[0005] Preferably, the acoustic model parameters of the candy head include at least the reference resonant frequency of the candy head. The AI ​​inference and control module calculates the frequency offset of the oral cavity resonance cavity based on the reference resonant frequency, the difference between the biting pressure value and the resting reference pressure baseline, and the difference between the tilt angle value and the reference tilt angle baseline, through the pressure linear influence coefficient, the angle influence coefficient, and the pressure-angle cross-coupling coefficient. The pressure linear influence coefficient, the angle influence coefficient, and the pressure-angle cross-coupling coefficient are all determined by the acoustic model parameters of the candy head.

[0006] More preferably, the audio processing and power amplifier module includes a dynamic equalizer, which has multiple frequency bands with frequency centers. The AI ​​inference and control module calculates a gain adjustment signal based on the frequency offset of the oral cavity resonance chamber, and the gain adjustment signal includes compensation gain values ​​corresponding to each frequency band.

[0007] More preferably, the compensation gain value is calculated from the frequency band baseline gain, the upper limit of the bone conduction speaker frequency response, the compensation intensity coefficient, the frequency offset sign function, the loudness growth index, and the effective bandwidth factor.

[0008] More preferably, the effective bandwidth factor is calculated from the reference bandwidth, bandwidth expansion coefficient, frequency band center, and sugar head reference resonant frequency, where the reference bandwidth and bandwidth expansion coefficient are both preset constants.

[0009] Further preferably, the lollipop Bluetooth speaker also includes a Bluetooth communication module, which is electrically connected to the AI ​​inference and control module and is used to receive audio data streams transmitted from external smart devices; the power module is electrically connected to the AI ​​inference and control module and is used to transmit the current battery voltage value to the AI ​​inference and control module; the AI ​​inference and control module is also used to, when the current battery voltage value is lower than a preset threshold, solve for the target value of the audio transmission bit rate and the target value of the effective value of the bone conduction speaker driving voltage that minimize the overall cost based on the total power consumption of the static power consumption, the power consumption of the Bluetooth communication module, and the power consumption of the audio processing and power amplifier module, under preset constraints.

[0010] More preferably, the overall cost is calculated from the highest transmission bit rate, the current transmission bit rate, the nominal voltage of the bone conduction speaker, the effective value of the current driving voltage, the weighting coefficient, and the estimated total harmonic distortion.

[0011] More preferably, the total harmonic distortion (THD) estimate is calculated from the base distortion coefficient, the drive-rate coupling coefficient, the normalized reference voltage, the current effective value of the drive voltage, the highest transmission rate, and the current transmission rate.

[0012] A control method for a lollipop Bluetooth speaker, the lollipop Bluetooth speaker including a replaceable lollipop head and a handle with a built-in cavity, the control method including: a reading step, acquiring the acoustic model parameters of the lollipop head from the NFC tag built into the lollipop head through an NFC reading module in the handle; a acquisition step, acquiring the user's biting pressure value through a pressure sensor in the biting area of ​​the handle, and acquiring the handle's tilt angle value through a tilt sensor in the handle; a frequency offset calculation step, calculating the oral cavity frequency offset based on the lollipop head acoustic model parameters, the difference between the biting pressure value and a preset resting reference pressure baseline, and the difference between the tilt angle value and a preset reference tilt angle baseline through an AI inference and control module in the handle; a gain adjustment step, generating a gain adjustment signal based on the oral cavity frequency offset and a preset compensation strategy through an AI inference and control module; and an audio adjustment and playback step, receiving the audio signal to be played through an audio processing and power amplifier module in the handle, equalizing the audio signal according to the gain adjustment signal, amplifying it, and driving a bone conduction speaker close to the inner wall of the cavity in the handle to play the audio.

[0013] Further preferably, the control method further includes: a voltage monitoring step, in which the current battery voltage value is transmitted from the power module to the AI ​​inference and control module; a low-power decision step, in which, when the current battery voltage value is lower than a preset threshold, the AI ​​inference and control module solves for the target value of the audio transmission bitrate and the target value of the effective value of the bone conduction speaker driving voltage, which minimize the overall cost, based on the total power consumption of the static power consumption, the power consumption of the Bluetooth communication module, and the power consumption of the audio processing and power amplifier module, under preset constraints; and a degraded playback step, in which the AI ​​inference and control module adjusts the transmission bitrate of the Bluetooth communication module to the target value and limits the effective value of the bone conduction speaker driving voltage to the target value, thereby maintaining continuous audio playback.

[0014] Technical Effects: This invention adapts to different candy head acoustic parameters through an NFC reading module, combines pressure and tilt sensors to collect user usage status, and uses an AI inference and control module to calculate oral resonance frequency shifts in real time and generate compensation signals. This solves the problems of traditional devices being unable to adaptively adjust audio parameters and sound quality being easily affected by user posture. Simultaneously, it dynamically optimizes transmission bitrate and drive voltage under low battery conditions, balancing power consumption and sound quality to ensure playback continuity. The bone conduction speaker's placement against the inner wall of the cavity also improves audio vibration transmission efficiency. Overall, it achieves multimodal perception-based adaptive audio control, significantly enhancing the user experience. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall architecture of the Lollipop Bluetooth speaker system; Figure 2 This is a schematic diagram of the adaptive audio compensation control method. Figure 3 The circuit schematic of the Bluetooth main control module (AB5607F); Figure 4 This is a schematic diagram of a USB interface circuit. Figure 5 Schematic diagram of battery charging management (TP4054); Figure 6 This is the schematic diagram of the Type-C power interface circuit. Figure 7 This is the schematic diagram of the power button control circuit. Figure 8 This is the circuit schematic for the status indicator light. Figure 9 This is a schematic diagram of the overall structure of the Lollipop Bluetooth speaker. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] Existing lollipop Bluetooth speakers cannot adapt to the acoustic characteristics of the lollipop head, and changes in the bite posture cause frequency shift and sound quality degradation.

[0018] Based on this, please refer to Figures 1-9 The lollipop Bluetooth speaker with NFC and AI control provided in this embodiment includes a handle and a replaceable candy head. The handle has an internal cavity. The device achieves adaptive audio control through multi-module collaboration. Those skilled in the art can implement all functions without creative effort by following the instructions.

[0019] The NFC reading module is located inside the handle, near the connection point between the candy head and the handle. It uses a high-frequency NFC read / write chip with a working frequency of 13.56MHz. The chip's accompanying radio frequency antenna is fabricated using a printed circuit board etching process. The antenna is positioned flush against the inner wall of the handle cavity, maintaining a stable sensing distance of less than 5mm between it and the NFC tag built into the candy head, ensuring stable data reading. The NFC reading module reads the information stored in the NFC tag built into the candy head via radio frequency coupling. The tag information includes the acoustic model parameters of the candy head, stored in binary data format. After reading, the module transmits the acoustic model parameters to the AI ​​inference and control module via an I2C serial bus. Stable parameter reading can be achieved for candies of different materials and structures. The reading process does not require manual calibration by the user; the reading operation is automatically completed after the device is powered on and connected to the candy head. The NFC read / write chip can be selected from RC522, PN532 or FM17550. This embodiment uses the FM17550 chip. The chip's SDA pin is connected to the PB6 pin of the AI ​​inference and control module, the SCL pin is connected to the PB7 pin, the RST pin is connected to the PB8 pin, and the IRQ pin is connected to the PB9 pin to realize bidirectional data communication and interrupt triggering. The chip's power supply voltage is 3.3V, which is provided by the VDDIO pin of the power supply module.

[0020] The pressure sensor is deployed in the biting area of ​​the handle, which is a raised structure on the outer side of the handle for the user to bite on. The sensor is a thin-film piezoresistive sensor, with its sensing surface tightly fitted to the outer surface of the biting area without gaps, ensuring that the biting pressure is completely transmitted to the sensor's sensing surface. The pressure sensor has a measurement range of 0 Pascal to 10000 Pascal. During operation, it outputs an analog voltage signal, which is amplified by an operational amplifier and then converted into a digital signal by a 12-bit analog-to-digital converter. The digital signal is transmitted in real time to the AI ​​inference and control module via an SPI serial peripheral interface. The sensor response time is less than 1 millisecond, accurately capturing instantaneous changes in the user's biting pressure. The resting reference pressure baseline is the reference pressure value collected by the sensor when the user is not biting. This value is automatically collected and stored during the device's power-on initialization phase to eliminate calculation errors caused by sensor zero drift. The pressure sensor can be FSR402, RP-C7.6 or FlexiForceA201. This embodiment uses FSR402 sensor. The output of the sensor is connected to the non-inverting input of the operational amplifier LM358, and the inverting input is connected to a 10K adjustable resistor for zero-point calibration. The amplified signal is connected to the PA0 pin of the AI ​​inference and control module. The PA0 pin has a built-in 12-bit ADC with a sampling frequency of 100Hz to realize real-time acquisition of pressure data.

[0021] The tilt sensor is centrally located inside the handle and employs a six-axis attitude sensor that integrates a micromechanical accelerometer and a gyroscope. It can acquire the handle's tilt angle in three-dimensional space in real time, with a measurement range of 0 to 360 degrees and a resolution of 0.1 degrees. The sensor integrates a Kalman filter algorithm to automatically eliminate interference signals from device vibration and hand movements, outputting a stable digital tilt angle signal. This signal is transmitted to the AI ​​inference and control module via a UART universal asynchronous transceiver. The reference tilt angle baseline is the base angle value when the handle is placed vertically. Baseline calibration is automatically completed during device initialization without manual intervention, ensuring a consistent reference standard for tilt angle calculations. The tilt sensor can be MPU6050, ICM20602 or BMI160. This embodiment uses the MPU6050 sensor. The SDA pin of the sensor is connected to the PB10 pin of the AI ​​inference and control module, the SCL pin is connected to the PB11 pin, and the INT pin is connected to the PB12 pin. The sensor is powered by 3.3V. The internal 16-bit ADC can realize high-precision attitude measurement. The output data format is 16-bit binary two's complement, and the scaling factor for converting it to the actual angle value is 16.4 LSB / degree.

[0022] The AI ​​inference and control module is located on the core circuit board inside the controller, using the AB5607F Bluetooth audio main control chip as the core controller. This chip integrates a 32-bit RISC processor, a floating-point unit, 4Mbit Flash memory, and a Bluetooth 5.0 communication module, enabling simultaneous data processing, Bluetooth communication, and audio processing. Pin 1 (BT_ANT) connects to the PIFA Bluetooth antenna, which is 25mm long and has an impedance of 50 ohms, manufactured using a printed circuit board etching process, enabling the transmission and reception of 2.4GHz Bluetooth signals. Pins 2 (OSCI) and 3 (OSCO) connect to a 26MHz crystal oscillator with a 9pF load capacitance and a frequency accuracy of ±10ppm, providing a stable system clock for the chip. Pins 4 (PB4 / USBDM / SDDAT) and 5 (PB3 / USBDP) connect to the DM and DP pins of the Type-C interface for USB firmware upgrades and data transmission. Pin 8 (PA5 / SDCMD) connects to an LED indicator circuit for device status indication. Pins 9 (PF3 / DACR) and 10 (DACL) connect to the differential input pins of the bone conduction speaker, outputting analog audio signals. Pin 11 (GND) is connected to system ground. Pins 13 and 14 (VDDIO) are connected to a 3.3V power supply, providing power to the chip's I / O ports and peripherals. Pin 15 (VBAT) is connected to the positive terminal of the battery, with a supply voltage range of 3.0V to 4.2V. The chip has a built-in battery voltage detection circuit that can collect the battery voltage value in real time. Pin 16 (PWRKEY) is connected to the power button circuit, realizing the power-on, power-off, and reset functions. The AI ​​inference and control module establishes electrical connections with the NFC reading module, pressure sensor, and tilt sensor, respectively. It first receives the candy head acoustic model parameters transmitted by the NFC reading module, and simultaneously acquires the biting pressure value transmitted by the pressure sensor and the tilt angle value transmitted by the tilt sensor in real time. The module first calculates the difference between the biting pressure value and the preset resting reference pressure baseline, then calculates the difference between the tilt angle value and the preset reference tilt angle baseline. Combining the candy head acoustic model parameters, it completes the calculation of the oral cavity resonance frequency offset. The calculation process is executed in the floating-point unit, and the single-cycle calculation time is less than 2 milliseconds. After the calculation is completed, the module calls the preset compensation strategy stored internally and generates the corresponding gain adjustment signal according to the frequency offset. The signal is output to the audio processing and power amplifier module in digital control form. The module has a built-in watchdog circuit, which can monitor the operating status in real time and avoid equipment failure caused by abnormal data calculation.

[0023] The bone conduction speaker is housed inside the handle, with its vibrating surface tightly fitted to the inner wall of the handle cavity. Utilizing a piezoelectric bone conduction vibration unit, it efficiently converts the input electrical signal into a mechanical vibration signal. This vibration signal is then transmitted directly to the user's oral cavity through the handle cavity and the candy head, achieving audio playback via bone conduction. The bone conduction speaker has a frequency response range of 20 Hz to 20000 Hz, a rated power of 1 watt, and an impedance of 8 ohms. With no air conduction loss, it effectively avoids interference from external environmental noise and is well-suited to the audio propagation characteristics within the oral cavity. The speaker is rigidly fixed to prevent displacement during vibration, ensuring efficient signal transmission. The two input pins of the bone conduction speaker are connected to pin 9 (PF3 / DACR) and pin 10 (DACL) of the AB5607F chip, respectively, to receive differential analog audio signals. Differential signal transmission effectively suppresses common-mode interference and improves audio signal quality.

[0024] The audio processing and power amplifier module is integrated within the AB5607F chip, including a digital dynamic equalizer, a digital-to-analog converter (DAC), and a Class-D power amplifier. It can receive both PCM digital and analog audio signals. The module first receives the gain adjustment signal output from the AI ​​inference and control module. Based on the signal parameters, it performs equalization processing on the audio signal, precisely adjusting the audio amplitude in different frequency bands. The equalized digital audio signal is then input to a 16-bit DAC to convert it into an analog audio signal. The analog audio signal is then input to the Class-D power amplifier, which has an energy conversion efficiency greater than 90%, amplifying the audio signal to the standard drive level of the bone conduction speaker. The amplified signal directly drives the bone conduction speaker to complete audio playback. The module incorporates a multi-stage digital filtering circuit, including a low-pass filter, a high-pass filter, and a notch filter, to eliminate power supply noise, signal transmission interference, and power frequency noise, ensuring pure, noise-free playback sound quality. The digital dynamic equalizer divides the full frequency response range into 10 independent frequency bands. The frequency center and bandwidth of each band can be configured by software. The default frequency band division follows the loudness characteristics of human hearing, with a low frequency bandwidth of 100 Hz, a mid frequency bandwidth of 200 Hz, and a high frequency bandwidth of 500 Hz.

[0025] The power module provides a stable power supply to all electronic modules of the device, including a rechargeable lithium-ion battery, a TP4054 linear charging management chip, a Type-C interface, and a power management circuit. The rechargeable battery has a rated voltage of 3.7 volts and a capacity of 300 mAh. The positive terminal of the battery is connected to pin 15 (VBAT) of the AB5607F chip and pin 3 (BAT) of the TP4054 chip, while the negative terminal is connected to system ground. Pin 4 (VIN) of the TP4054 chip is connected to the +5V pin of the Type-C interface, pin 1 (CHRG) is connected to the red charging indicator light, pin 5 (PROG) is connected to a 2K ohm resistor to ground, setting the charging current to 500 mA, and pin 2 (GND) is connected to system ground. When a 5V power supply is plugged into the Type-C interface, the TP4054 chip automatically enters charging mode, and the red indicator light illuminates. Once the battery is fully charged, the red indicator light turns off, and the chip automatically enters standby mode. The Type-C interface uses a female connector design. Pins A1, A12, B1, and B12 are connected to system ground, pins A4, A9, B4, and B9 are connected to +5V power, and pins A5 and B5 are each connected to ground via 5.1K ohms resistors, serving as CC pins for USB host identification. Pins A6, A7, B6, and B7 are connected to DM and DP signals respectively for USB data transmission. The power module converts the battery voltage to 3.3VVDDIO voltage through the LDO regulator inside the AB5607F chip, powering the NFC reader module, pressure sensor, tilt sensor, and chip I / O ports. The module has a built-in voltage detection circuit that uses the AB5607F chip's internal 12-bit ADC to acquire the battery output voltage value in real time, with a voltage detection accuracy of 0.01 volts. The detection results are transmitted to the AI ​​inference and control module in real time. The power module's static power consumption is less than 10 microamps, maximizing device battery life.

[0026] The power button circuit includes a tactile button S1. One end of the button is connected to pin 16 (PWRKEY) of the AB5607F chip, and the other end is connected to system ground. The chip has a built-in 90K ohm pull-up resistor to pull the PWRKEY pin high. When the button is pressed, the pin level is pulled low, and the chip executes the corresponding operation upon detecting the low-level signal. A 3-second press and hold of the button powers on the device, illuminating the blue indicator light. A 10-second press and hold of the button resets the system, re-initializing all modules. A 3-second press and hold of the button powers off the device, extinguishing all indicator lights. The LED indicator circuit includes a blue LED and a red LED connected in series. The common terminal is connected to pin 8 (PA5 / SDCMD) of the AB5607F chip. The anode of the blue LED is connected to 3.3V VDDIO, and the cathode of the red LED is connected to system ground. The chip integrates a 300-ohm current-limiting resistor to drive the LED, eliminating the need for external resistors. When the device is running normally, the blue LED is constantly lit; when charging, the red LED is constantly lit and the blue LED is off; when the battery is low, the red LED flashes at a frequency of 1Hz to remind the user to charge; when the firmware is being upgraded, the blue LED and the red LED flash alternately.

[0027] Furthermore, the acoustic model parameters of the candy head include at least the reference resonant frequency. The AI ​​inference and control module calculates the oral cavity resonance frequency offset based on the reference resonant frequency, the difference between the biting pressure value and the resting reference pressure baseline, and the difference between the tilt angle value and the reference tilt angle baseline, using pressure linear influence coefficient, angle influence coefficient, and pressure-angle cross-coupling coefficient. These coefficients are all determined by the acoustic model parameters of the candy head. The reference resonant frequency of the candy head is measured in Hertz and is uniquely determined by the physical characteristics of the candy head material, such as density, geometric dimensions, and cavity structure. Different candies have different reference resonant frequencies. This value is pre-written into the candy head's built-in NFC tag, which is directly retrieved by the device after reading it. The pressure linear influence coefficient is a dimensionless constant, characterizing the linear influence of relative changes in occlusal pressure on frequency shift; the angle influence coefficient is a dimensionless constant, characterizing the influence of individual changes in handle tilt angle on frequency shift; the pressure-angle cross-coupling coefficient is a dimensionless constant, correcting the nonlinear coupling error caused by the combined effect of pressure and angle. All three coefficients are pre-calibrated based on the acoustic characteristics of the candy head and stored in the NFC tag, requiring no real-time calibration during calculation. The AI ​​inference and control module, based on oral acoustic resonance theory and the principle of multi-physics coupling, converts pressure changes into relative rates of change to eliminate dimensional differences, and converts tilt angle changes into trigonometric function values ​​to adapt to periodic changes in spatial posture, accurately quantifying the influence of the synergistic effect of multiple variables on the resonant frequency and improving the accuracy of frequency shift calculation.

[0028] The formula for calculating the frequency offset of the oral resonator is: The formula derivation is based on the Helmholtz resonator theory. The acoustic system composed of the oral cavity and the candy head can be approximated as a variable Helmholtz resonator, whose natural resonant frequency is inversely proportional to the square root of the cavity volume and directly proportional to the square root of the opening area. When the user applies biting pressure, the candy head undergoes elastic deformation, the cavity volume decreases, and the resonant frequency increases. The change in frequency is linearly related to the relative rate of change of pressure; therefore, a pressure linear influence coefficient is introduced. When the handle tilt angle changes, the incident angle and reflection path of the sound wave in the oral cavity change, the equivalent opening area and neck length change, and the resonant frequency changes accordingly. Since the tilt angle change is periodic, a sine function is used to characterize its effect on frequency shift, and an angle influence coefficient is introduced. When the biting pressure and tilt angle change simultaneously, a nonlinear coupling effect occurs between them. The pressure change alters the acoustic impedance of the candy head, which in turn affects the frequency response to the tilt angle change. Therefore, a cross-coupling term and a pressure-angle cross-coupling coefficient are introduced. To correct the nonlinear error of multivariate synergistic effects.

[0029] Each parameter in the formula has a clear physical meaning, dimension, and basis for its value. This is the frequency offset of the oral cavity resonance chamber, measured in Hertz. It is the core basis for audio signal equalization compensation. A positive value indicates that the resonance frequency shifts to higher frequencies, while a negative value indicates that the resonance frequency shifts to lower frequencies. The magnitude of the offset directly determines the audio compensation amplitude. The value range has been determined to be -500 Hertz to 500 Hertz through a limited number of experiments. The reference resonant frequency of the candy head, measured in Hertz, is the inherent resonant frequency of the candy head under standard posture and without external force. It serves as the reference parameter for the entire calculation process. The value range varies depending on the material of the candy head: 400 Hz to 800 Hz for silicone soft candy heads, 1000 Hz to 1500 Hz for hard candy heads, and 600 Hz to 1000 Hz for chocolate candy heads. This value is obtained by measuring the acoustic resonant frequency of the candy head using the impedance tube method. The pressure linear influence coefficient is dimensionless, and its value range was determined to be 0.05 to 0.2 through a finite number of experiments. The larger the coefficient, the more significant the effect of pressure changes on frequency deviation. (The last sentence appears to be unrelated and possibly a fragment from another context: "The candy head has greater elasticity.") The value is too large, resulting in less elasticity in the hard candy tip. The value is too small. This is a real-time biting pressure value in Pascals, which dynamically changes with the user's biting force. The value ranges from 0 Pascals to 10,000 Pascals and is collected in real time by a pressure sensor. This is the resting reference pressure baseline, measured in Pascals. It is used to eliminate calculation errors caused by sensor zero drift and ranges from 50 Pascals to 150 Pascals. It is obtained by automatically collecting 100 pressure data points and averaging them during device startup initialization. The angular influence coefficient is dimensionless, and its value range was determined to be 0.03 to 0.15 through a finite number of experiments. It is directly related to the acoustic propagation characteristics of the candy head. (Spherical candy head) The value is too large, stick-shaped candy. The value is too small. This is the real-time handle tilt angle value, in degrees, reflecting the change in the angle at which the user holds the device. The value ranges from 0 to 360 degrees and is collected in real time by the tilt sensor. The reference tilt baseline, in degrees, serves as a standard for unified tilt angle calculation. The value is 90 degrees, which is the angle when the handle is placed vertically downwards. This is obtained through automatic calibration during device initialization. The pressure angle cross-coupling coefficient is dimensionless, and its value range was determined to be 0.02 to 0.1 through a finite number of experiments. It corrects the nonlinear error of multivariate synergistic effects. The softer the material, the better. The larger the value, the better.

[0030] A silicone gummy candy tip was selected as an example for verification. The reference resonant frequency of the candy tip was 600 Hz, the pressure linearity influence coefficient was 0.15, the angle influence coefficient was 0.1, the pressure-angle cross-coupling coefficient was 0.06, the resting reference pressure baseline was 100 Pascal, the reference tilt angle baseline was 90 degrees, and when the user's real-time biting pressure was 300 Pascal and the real-time handle tilt angle was 60 degrees, the frequency offset was calculated using the formula as follows: Hertz, the actual measured oral resonance frequency offset using an acoustic testing system was 85 Hz, with a calculation error of 3.1%, verifying the accuracy of the formula calculation. A hard candy head example was selected for verification. The reference resonance frequency of this candy head was 1200 Hz, the pressure linearity influence coefficient was 0.08, the angle influence coefficient was 0.05, the pressure-angle cross-coupling coefficient was 0.03, the resting reference pressure baseline was 100 Pascals, the reference tilt angle baseline was 90 degrees, and when the user's real-time biting pressure was 700 Pascals and the real-time handle tilt angle was 150 degrees, the frequency offset calculated using the formula was... The actual measured frequency offset was 510 Hz, and the calculation error was 1.9%, further verifying the reliability of the formula.

[0031] Specifically, the audio processing and amplifier module includes a dynamic equalizer. The dynamic equalizer has multiple frequency bands with frequency centers. The AI ​​inference and control module calculates a gain adjustment signal based on the frequency offset of the oral cavity resonance chamber. This gain adjustment signal contains the compensation gain value for each frequency band. The dynamic equalizer divides the full frequency response range of the bone conduction speaker into multiple independent audio bands, each with a fixed frequency center measured in Hertz. The band division follows the equal loudness characteristics of human hearing, with the width of the low-frequency, mid-frequency, and high-frequency bands adjusted progressively to ensure that the compensation effect of each band conforms to auditory habits. The AI ​​inference and control module calculates a dedicated compensation gain value for each frequency band based on the magnitude and direction of the frequency offset. This compensation gain value is measured in decibels. The gain adjustment signal integrates the compensation gain values ​​of each frequency band into a digital control command and transmits it to the dynamic equalizer control terminal. Upon receiving the signal, the dynamic equalizer adjusts the amplitude of the audio signal in each frequency band individually, avoiding timbre imbalance caused by uniform compensation across the entire frequency band. This achieves segmented precise equalization, specifically correcting frequency band distortion caused by frequency offset and improving audio playback fidelity. The dynamic equalizer is implemented using IIR digital filters, with one second-order peak filter corresponding to each frequency band. The center frequency, gain, and Q value of the filter can be configured in real time via software with a configuration delay of less than 1 millisecond, ensuring the real-time performance of the compensation.

[0032] Preferably, the compensation gain value is calculated from the frequency band baseline gain, the upper limit of the bone conduction speaker frequency response, the compensation intensity coefficient, the frequency offset sign function, the loudness growth index, and the effective bandwidth factor. The frequency band baseline gain is the default gain value for each frequency band without frequency offset, measured in decibels (dB). It is preset according to the frequency response characteristics of the bone conduction speaker and represents the basic output amplitude of the audio signal. The upper limit of the bone conduction speaker frequency response is the maximum output frequency of the hardware, measured in Hertz (Hz). It is an inherent performance parameter of the speaker and determines the upper limit of the audio compensation frequency. The compensation intensity coefficient is a dimensionless constant that controls the overall audio compensation amplitude, and its value range is determined through a limited number of experiments. The frequency offset sign function is a dimensionless function, with a positive value of 1, a negative value of -1, and a zero value of 0, guiding the increase or decrease of the compensation gain. The loudness growth index is a dimensionless constant that matches the human ear's loudness perception characteristics, ensuring that the change in audio loudness after compensation conforms to auditory laws. The effective bandwidth factor is the effective operating width of each frequency band, measured in Hertz, characterizing the effective propagation range of the audio signal within the frequency band and dynamically changing with the frequency center. Multi-parameter collaborative calculation enables audio compensation to simultaneously match hardware performance, frequency offset status, and human auditory characteristics, avoiding overcompensation or undercompensation.

[0033] The formula for calculating the compensation gain value is as follows: The formula derivation is based on audio equalization compensation theory and the Weber-Fechner law. Human hearing perceives sound loudness according to a logarithmic law; subjective loudness is approximately proportional to the 1 / 3 power of sound intensity. Therefore, a 1 / 3 power term is introduced to represent the ratio of the frequency center to the upper limit of the frequency response, achieving a natural attenuation of the high-frequency compensation amplitude, consistent with the characteristics of the human ear's equal-loudness curve. The ratio of the relative value of the frequency offset to the effective bandwidth factor characterizes the proportion of the offset relative to the frequency band width. A larger proportion requires a larger compensation gain. A power function is used to represent the relationship between the compensation gain and the offset ratio, matching the power-law characteristic of human ear loudness growth. The sign function determines the compensation direction. When the resonant frequency shifts to higher frequencies, the high-frequency gain needs to be reduced and the low-frequency gain increased, and vice versa. Therefore, the sign function value is opposite to the direction of change of the compensation gain, and a minus sign is used in the formula to achieve this logic.

[0034] Each parameter in the formula has a clear physical meaning, dimension, and basis for its value. The real-time compensation gain value, in decibels, is the final value of the dynamic equalizer adjusting the gain of the corresponding frequency band. It changes dynamically with time and frequency center, and its range has been determined to be -12 dB to 12 dB through a limited number of experiments. This is the baseline gain of the frequency band, measured in decibels. It is the default gain in the uncompensated state of the frequency band corresponding to the frequency center. It is calibrated according to the frequency response curve of the bone conduction speaker and has a value range of -6 dB to 6 dB. It is used to correct the frequency response unevenness of the speaker itself. To compensate for the intensity coefficient, a dimensionless coefficient with a value range of 1 to 5 determined through a limited number of experiments, the overall compensation level can be adjusted to suit different users' listening preferences. Users sensitive to sound quality can choose a smaller value. For users seeking high fidelity, a larger value is recommended. value. The frequency center of the frequency band, measured in Hertz, is the center frequency at which the dynamic equalizer divides the frequency band. It determines the target frequency band for compensation operations and ranges from 20 Hertz to 20,000 Hertz. This represents the upper limit of the bone conduction speaker's frequency response, measured in Hertz, limiting the compensation operation frequency range to avoid ineffective compensation exceeding the hardware's capabilities. In this embodiment... The value is 20000 Hz. This is a sign function, dimensionless, whose value is determined by the sign of the frequency offset. The value is 1. The value is -1. When the value is 0, the direction of compensation gain enhancement or attenuation is precisely controlled. This represents the frequency offset of the oral cavity resonance chamber, measured in Hertz, and is the core basis for compensation calculations in the early stages. The effective bandwidth factor, measured in Hertz, affects the compensation amplitude attenuation rate and ranges from 50 Hertz to 1000 Hertz. The loudness growth index is dimensionless, and its value range was determined to be 0.6 to 1.0 through a finite number of experiments. It matches the logarithmic loudness perception characteristics of the human ear. In this embodiment... The value is 0.8, which aligns with the hearing habits of most people.

[0035] An example was conducted using the 1000 Hz frequency band as the center of the frequency. In this band, the baseline gain was 0 dB, the upper limit of the bone conduction speaker's frequency response was 20000 Hz, the compensation intensity coefficient was 2.5, the loudness growth index was 0.8, the effective bandwidth factor was 200 Hz, and the frequency offset was 42.8 Hz. Substituting these values ​​into the formula, the real-time compensation gain was calculated to be... The compensation value of 1 dB effectively corrects frequency shift distortion in the frequency band. Using an audio analyzer to test the frequency response of this band, the peak frequency shift before compensation was 42.8 Hz, and the gain deviation was 1.5 dB. After compensation, the peak frequency recovered to 1000 Hz, and the gain deviation was 0.2 dB. Sound quality tests showed that the frequency band distortion decreased from 8% to 2%, and the timbre reproduction was significantly improved. An example was taken at the 5000 Hz frequency center. The baseline gain for this band was 2 dB, the upper limit of the bone conduction speaker frequency response was 20000 Hz, the compensation intensity coefficient was 2.5, the loudness growth index was 0.8, the effective bandwidth factor was 400 Hz, and the frequency shift was 87.6 Hz. Substituting these values ​​into the formula, the real-time compensation gain value was calculated to be... In actual testing, the gain deviation in this frequency band was 0.3 dB after compensation, and the distortion was reduced from 10% to 2.5%, verifying the effectiveness of the formula compensation.

[0036] Furthermore, the effective bandwidth factor is calculated from the reference bandwidth, bandwidth expansion coefficient, frequency center of the band, and reference resonant frequency of the bone conduction speaker. Both the reference bandwidth and bandwidth expansion coefficient are preset constants. The reference bandwidth is the default effective bandwidth for the low-frequency band, measured in Hertz, and is set based on the low-frequency response characteristics of the bone conduction speaker; it forms the fundamental width for audio frequency band division. The bandwidth expansion coefficient is a dimensionless constant that controls the rate of expansion of the effective bandwidth with respect to the frequency center. Its value range is determined through a limited number of experiments to ensure smooth and continuous bandwidth changes. The effective bandwidth factor calculation combines the frequency center of the band with the reference resonant frequency of the bone conduction speaker, ensuring that the bandwidth value adapts to the acoustic characteristics of different bone conduction speakers and the propagation patterns of different frequency bands, avoiding compensation errors caused by a fixed bandwidth.

[0037] The formula for calculating the effective bandwidth factor is as follows: The formula derivation is based on the human ear's critical bandwidth theory. The human ear's critical bandwidth increases logarithmically with frequency; the higher the frequency, the wider the critical bandwidth. Therefore, a logarithmic function is used to characterize the expansion of the effective bandwidth with respect to the frequency center. A reference resonant frequency for the earphone is introduced as a normalization parameter to adapt the effective bandwidth factor to the acoustic characteristics of different earphones. Earphones with lower reference resonant frequencies have wider effective bandwidths corresponding to the same frequency center, consistent with the acoustic characteristics of strong diffraction and wide propagation range of low-frequency sound waves.

[0038] Each parameter in the formula has a clear physical meaning and dimension. The effective bandwidth factor, measured in Hertz, is the effective operating width of the corresponding frequency band of the dynamic equalizer. It directly affects the calculation result of the compensation gain value and ranges from 50 Hertz to 1000 Hertz. The reference bandwidth, measured in Hertz, is the fundamental bandwidth when the frequency center approaches zero. It is preset based on the low-frequency propagation characteristics of audio signals, and its value range was determined to be 50 Hertz to 150 Hertz through a limited number of experiments. In this embodiment... The value is 100 Hz. The bandwidth expansion factor is dimensionless, and its value range was determined to be 0.2 to 0.5 through a finite number of experiments. It controls the rate at which the effective bandwidth expands with increasing frequency. In this embodiment... The value is 0.3. It represents the frequency center of the frequency band, measured in Hertz, and reflects the position of the frequency band in the entire frequency domain. The reference resonant frequency of the sugar head, in Hertz, is used to adapt the effective bandwidth factor to the inherent acoustic characteristics of different sugar heads, thereby improving the versatility of the calculation.

[0039] An example was selected with a reference bandwidth of 100 Hz, a bandwidth expansion factor of 0.3, a sugar head reference resonant frequency of 800 Hz, and a frequency center of 1000 Hz for verification. Substituting these values ​​into the formula, the effective bandwidth factor was calculated as follows: Hertz, the effective bandwidth of this frequency band was actually tested using an audio analyzer and found to be 120 Hz, with a calculation error of 3.6%, ensuring the accuracy of the compensation gain calculation. An example was selected with a sugar head reference resonant frequency of 600 Hz and a frequency band center of 1000 Hz for verification. Substituting these values ​​into the formula, the effective bandwidth factor was calculated to be... Hertz, the actual effective bandwidth is 125 Hz, the calculation error is 2%, verifying the adaptability of the formula to different sugar heads.

[0040] Specifically, the Lollipop Bluetooth speaker also includes a Bluetooth communication module integrated within the AB5607F chip. This module is electrically connected to the AI ​​inference and control module and is used to receive audio data streams transmitted from external smart devices. The power module is also electrically connected to the AI ​​inference and control module and is used to transmit the current battery voltage value to it. The Bluetooth communication module uses the Bluetooth 5.0 protocol and supports A2DP, AVRCP, HFP, and HSP audio transmission standards. It operates at a frequency of 2.402GHz to 2.480GHz, with a transmit power of 0dBm and a receive sensitivity of -90dBm. It can stably receive audio data streams transmitted from external smart devices such as mobile phones and tablets within a 10-meter range, and the transmission bitrate supports multi-level adjustment from 64kbps to 320kbps. After the data stream is received by the PIFA Bluetooth antenna, it is demodulated into a digital audio signal by the chip's internal Bluetooth RF front-end and transmitted to the audio processing and power amplifier module. The power module voltage detection circuit uses the 12-bit ADC inside the AB5607F chip to collect the current battery voltage value in real time, in volts. When the voltage value is lower than a preset threshold, a low-power decision process is triggered. The preset threshold is set to 3.2 volts based on the battery discharge characteristics to ensure stable operation of the device under low voltage conditions. The AI ​​inference and control module collects the device's static power consumption, the real-time power consumption of the Bluetooth communication module, and the real-time power consumption of the audio processing and power amplifier module. The total power consumption is the sum of these three. The preset constraint is that the total power consumption does not exceed the power budget corresponding to the current battery voltage. The power budget is directly determined by the battery voltage value; the lower the battery voltage, the smaller the power budget. Under the constraint, the module aims to minimize the overall cost and uses a gradient descent numerical iterative algorithm to solve for the optimal audio transmission bitrate and the effective value of the bone conduction speaker drive voltage, balancing device power consumption and audio playback quality.

[0041] Preferably, the overall cost is calculated from the maximum transmission bit rate, the current transmission bit rate, the nominal voltage of the bone conduction speaker, the effective value of the current driving voltage, a weighting coefficient, and the estimated total harmonic distortion (THD). The maximum transmission bit rate is the maximum audio transmission bit rate supported by the Bluetooth communication module, measured in kilobits per second, and is an inherent performance parameter of the module. The current transmission bit rate is the real-time operating bit rate of the Bluetooth module, measured in kilobits per second, and can be adjusted between 0 and the maximum transmission bit rate. The nominal voltage of the bone conduction speaker is the rated driving voltage of the speaker, measured in volts, and is the standard voltage for normal speaker operation. The effective value of the current driving voltage is the voltage value used to drive the speaker in real time, measured in volts, and directly affects the speaker's output power and sound quality. The weighting coefficient is a dimensionless constant that assigns weights to the transmission bit rate, driving voltage, and distortion on the overall cost, preset according to the user's battery life and sound quality requirements. The estimated total harmonic distortion (THD) is an estimated value of the audio playback distortion level, dimensionless, reflecting the degree of sound quality degradation. The overall cost is calculated by weighting multiple parameters. The smaller the value, the better the balance between power consumption and sound quality, providing a clear quantitative target for low power consumption optimization.

[0042] The formula for calculating the overall cost is as follows: The formula derivation is based on multi-objective optimization theory. In low-battery conditions, three objectives need to be optimized simultaneously: maximizing the audio transmission bitrate to ensure sound quality, maximizing the driving voltage to ensure loudness, and minimizing total harmonic distortion (THD) to ensure sound quality. These three objectives conflict with each other and cannot be simultaneously optimized. Therefore, the three objectives are transformed into a unified cost function. By using weighted summation, the multi-objective optimization problem is transformed into a single-objective optimization problem, and the Pareto optimal solution is found. The cost of reducing the bitrate is proportional to the relative reduction in bitrate, the cost of reducing the driving voltage is proportional to the relative reduction in voltage, and the cost of increasing distortion is proportional to the degree of distortion. Different optimization preferences can be achieved by adjusting the weighting coefficients: increasing the weight of bitrate and voltage when prioritizing battery life, and increasing the weight of distortion when prioritizing sound quality.

[0043] Each parameter in the formula has a clear physical meaning, dimension, and optimization logic. The overall cost is dimensionless and is the core objective function for low-power optimization. The smaller the value, the better the optimization effect. The value range is from 0 to 1. This is the bitrate weighting coefficient, dimensionless. The voltage weighting coefficient is dimensionless. The distortion weighting coefficients are dimensionless, and the sum of the three coefficients is 1. They are preset based on battery life and sound quality priorities, with battery life priority mode applied in this mode. , , Balanced mode , , In audio quality priority mode , , . The maximum transmission rate, measured in kilobits per second, represents the maximum transmission capacity of the Bluetooth communication module. In this embodiment... The value is 320 kilobits per second. The current transmission bit rate is expressed in kilobits per second. This is a variable parameter to be optimized, with a value range of 64 kilobits per second to 320 kilobits per second. The nominal voltage of the bone conduction speaker is expressed in volts. It is the standard operating voltage of the speaker, as used in this embodiment. The value is 3.7 volts. The current effective value of the driving voltage is in volts. It is a parameter to be optimized, with a value range of 1.0 volts to 3.7 volts. This is a dimensionless estimate of total harmonic distortion, reflecting the degree of audio quality distortion at the current bitrate and voltage, with a value ranging from 0 to 1.

[0044] The maximum transmission bit rate was selected as 320 kilobits per second, the nominal voltage of the bone conduction speaker was 3.7 volts, and the weighting coefficient for balanced mode was selected. , , Example verification: With a current battery voltage of 3.0 volts, a power consumption budget of 100 milliwatts, and a static power consumption of 10 milliwatts, iterative solutions using the gradient descent algorithm were used. When the transmission bit rate was 128 kilobits per second and the effective driving voltage was 2.2 volts, the Bluetooth communication module power consumption was 30 milliwatts, the audio processing and power amplifier module power consumption was 55 milliwatts, and the total power consumption was... Milliwatts, satisfying the power budget constraint. Substituting into the formula, the overall cost is calculated as follows: This is the optimal solution under the current constraints. Actual testing showed that the device's continuous playback time under these parameters was 8 hours, a 45% increase compared to the normal mode's 5.5 hours. The total harmonic distortion was 12%, which is acceptable to users, verifying the effectiveness of the formula optimization.

[0045] Furthermore, the total harmonic distortion (THD) prediction is calculated from the base distortion coefficient, the drive bit rate coupling coefficient, the normalized reference voltage, the effective value of the current drive voltage, the highest transmission bit rate, and the current transmission bit rate. The base distortion coefficient is a dimensionless constant, a quantified value of the inherent distortion of the hardware circuitry and speaker, determined by the hardware characteristics of the device. The drive bit rate coupling coefficient is also a dimensionless constant, characterizing the synergistic effect of drive voltage and transmission bit rate on distortion; its value range is determined through a limited number of experiments. The normalized reference voltage, measured in volts, serves as a standard for voltage parameter normalization, eliminating dimensional differences. The THD prediction calculation, combining the coupling relationship between drive voltage and transmission bit rate, accurately predicts the degree of sound quality distortion under different operating parameters, providing reliable data for comprehensive cost calculation.

[0046] The formula for calculating the total harmonic distortion (THD) prediction is as follows: The formula derivation is based on the physical laws of audio distortion. Crossover distortion in Class D power amplifiers and nonlinear distortion in bone conduction speakers increase exponentially with increasing drive voltage. Decreasing the audio bitrate leads to increased quantization noise, and the distortion degree increases exponentially with decreasing bitrate. Therefore, an exponential function is used to fit the coupling relationship between distortion degree and drive voltage and transmission bitrate. The ratio of drive voltage to normalized reference voltage characterizes the relative change in voltage, and the ratio of transmission bitrate to the maximum transmission bitrate characterizes the relative change in bitrate. The ratio reflects the coupling strength between drive and bitrate; a larger ratio indicates higher distortion.

[0047] Each parameter in the formula has a clear physical meaning and dimension. This is a dimensionless estimate of total harmonic distortion. The larger the value, the more severe the audio distortion. The value ranges from 0 to 1. The basic distortion coefficient is dimensionless and represents the inherent distortion value under standard operating conditions of the equipment. It is determined by hardware testing and calibration. In this embodiment... The value is 0.02. The driving bit rate coupling coefficient is dimensionless, and its value range was determined to be 1.0 to 1.5 through a finite number of experiments. It characterizes the coupling effect of driving voltage and transmission bit rate on distortion. In this embodiment... The value is 1.2. This is the effective value of the current driving voltage, in volts. To normalize the reference voltage, the unit is volts. The driving voltage is normalized to unify the parameter dimensions. In this embodiment... The value is 3.7 volts, which is consistent with the nominal voltage of the bone conduction speaker. This represents the current transmission rate, expressed in kilobits per second. This is the highest transmission rate, measured in kilobits per second.

[0048] Using a baseline distortion coefficient of 0.02, a drive rate coupling coefficient of 1.2, a normalized reference voltage of 3.7 volts, a current effective drive voltage of 2.2 volts, a current transmission rate of 128 kilobits per second, and a maximum transmission rate of 320 kilobits per second as examples for verification, the total harmonic distortion (THD) estimate is calculated using the formula. The total harmonic distortion (THD) was measured to be 0.125 using an audio analyzer, with a calculation error of 4.8%, verifying the accuracy of the formula's prediction. An example with a driving voltage of 3.7 volts and a transmission rate of 320 kilobits per second was selected for verification. Substituting these values ​​into the formula, the predicted THD was obtained as follows: The actual test distortion was 0.07, and the calculation error was 5.7%, further verifying the reliability of the formula.

[0049] The control method for the Lollipop Bluetooth speaker includes a replaceable candy head and a handle with a built-in cavity. The control method comprises reading, acquisition, frequency offset calculation, gain adjustment, and audio adjustment and playback steps. The reading step uses an NFC reading module within the handle to obtain the acoustic model parameters of the candy head from the NFC tag built into the candy head. The reading operation is automatically performed after the device connects to the candy head, with a reading time of less than 100 milliseconds. After the parameters are read, they are stored in the Flash cache unit of the AI ​​inference and control module. The acquisition step uses a pressure sensor in the bite area of ​​the handle to collect the user's biting pressure value and a tilt sensor within the handle to collect the handle's tilt angle value. Both types of data are acquired at a frequency of 100 Hz to ensure real-time and continuous data. The frequency offset calculation step uses the AI ​​inference and control module within the handle to calculate the oral resonance cavity frequency offset based on the candy head acoustic model parameters, the difference between the biting pressure value and a preset resting reference pressure baseline, and the difference between the tilt angle value and a preset reference tilt angle baseline. The calculation process is performed in real-time with no data delay. The gain adjustment step utilizes an AI inference and control module to generate a gain adjustment signal based on the oral cavity resonance frequency offset and a preset compensation strategy. The signal generation cycle is synchronized with the data acquisition cycle to ensure real-time compensation. The audio adjustment and playback step receives the audio signal to be played through the audio processing and amplification module within the handle. After equalization processing based on the gain adjustment signal, the audio signal is amplified and driven by a bone conduction speaker positioned close to the cavity wall within the handle. Audio signal processing and playback are uninterrupted, ensuring a smooth user experience.

[0050] Preferably, the control method further includes a voltage monitoring step, a low-power decision step, and a degraded playback step. The voltage monitoring step transmits the current battery voltage value to the AI ​​inference and control module via the power module. The voltage monitoring frequency is 1 Hz, tracking changes in battery power in real time. The low-power decision step is initiated when the current battery voltage value is lower than a preset threshold. Based on the total power consumption of the static power consumption, Bluetooth communication module power consumption, and audio processing and power amplifier module power consumption, the AI ​​inference and control module solves for the target value of the audio transmission bitrate and the target value of the effective value of the bone conduction speaker driving voltage, minimizing the overall cost, under preset constraints. The decision-making process takes less than 50 milliseconds and does not affect playback continuity. The degraded playback step adjusts the Bluetooth communication module transmission bitrate to the target value and limits the effective value of the bone conduction speaker driving voltage to the target value via the AI ​​inference and control module. The device continues to operate in a reduced power consumption state, audio playback is uninterrupted, and sound quality loss is controlled within an acceptable range until the battery voltage recovers or the power is depleted.

[0051] The entire device and control method operate fully automatically, requiring no manual user setup. Multimodal data acquisition and adaptive adjustment work in tandem to adapt to different head sizes, user preferences, and usage postures. It automatically optimizes power consumption in low-battery conditions, resolving issues such as susceptibility to sound quality interference, insufficient battery life, and playback interruptions common in traditional devices. This results in stable, high-quality, and adaptive audio playback. Actual testing shows that under normal use, the device significantly reduces total harmonic distortion (THD), resulting in a substantial improvement in sound quality compared to traditional devices. Playback time is significantly extended in low-battery conditions. Frequency offset compensation responds rapidly, adjusting audio parameters in real-time to follow changes in the user's bite and posture, fully meeting user needs.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A lollipop Bluetooth speaker with NFC and AI control, comprising a handle and a replaceable sugar head, wherein a cavity is arranged inside the handle, characterized in that, The lollipop Bluetooth speaker includes: an NFC reading module, located within the handle, for reading NFC tag information embedded in the lollipop head, the NFC tag information containing acoustic model parameters of the lollipop head; a pressure sensor, deployed in the biting area of ​​the handle, for collecting the user's biting pressure value; a tilt sensor, deployed within the handle, for collecting the tilt angle value of the handle; and an AI inference and control module, located within the handle and electrically connected to the NFC reading module, the pressure sensor, and the tilt sensor, the AI ​​inference and control module being used to acquire the acoustic model parameters of the lollipop head, the biting pressure value, and the tilt angle value, based on the lollipop head's acoustic... The system calculates the oral cavity frequency offset based on the model parameters, the difference between the biting pressure value and the preset resting reference pressure baseline, and the difference between the tilt angle value and the preset reference tilt angle baseline. A gain adjustment signal is generated based on the frequency offset and a preset compensation strategy. A bone conduction speaker, located inside the handle and close to the inner wall of the cavity, converts electrical signals into vibration signals. An audio processing and amplification module, located inside the handle and electrically connected to the AI ​​inference and control module and the bone conduction speaker, receives the audio signal to be played, performs equalization processing on the audio signal based on the gain adjustment signal, amplifies it, and drives the bone conduction speaker to play the audio. A power supply module provides power to all modules.

2. The lollipop Bluetooth speaker with NFC and AI control according to claim 1, wherein, The acoustic model parameters of the candy head include at least the reference resonant frequency of the candy head. The AI ​​inference and control module calculates the frequency offset of the oral cavity resonance cavity based on the reference resonant frequency, the difference between the biting pressure value and the resting reference pressure baseline, and the difference between the tilt angle value and the reference tilt angle baseline, through the pressure linear influence coefficient, the angle influence coefficient, and the pressure-angle cross-coupling coefficient. The pressure linear influence coefficient, the angle influence coefficient, and the pressure-angle cross-coupling coefficient are all determined by the acoustic model parameters of the candy head.

3. The lollipop Bluetooth speaker with NFC and AI control according to claim 1, wherein, The audio processing and power amplifier module includes a dynamic equalizer, which has multiple frequency bands with frequency centers. The AI ​​inference and control module calculates a gain adjustment signal based on the frequency offset of the oral cavity resonance chamber. The gain adjustment signal includes compensation gain values ​​corresponding to each frequency band.

4. The lollipop Bluetooth speaker with NFC and AI control according to claim 3, characterized in that, The compensation gain value is calculated from the frequency band baseline gain, the upper limit of the bone conduction speaker frequency response, the compensation intensity coefficient, the frequency offset sign function, the loudness growth index, and the effective bandwidth factor.

5. The lollipop Bluetooth speaker of claim 4, wherein, The effective bandwidth factor is calculated from the reference bandwidth, bandwidth expansion coefficient, frequency band center, and sugar head reference resonant frequency. The reference bandwidth and bandwidth expansion coefficient are both preset constants.

6. The lollipop Bluetooth speaker with NFC and AI control according to claim 1, wherein, The lollipop Bluetooth speaker also includes a Bluetooth communication module, which is electrically connected to the AI ​​inference and control module and is used to receive audio data streams transmitted from external smart devices. The power module is electrically connected to the AI ​​inference and control module and is used to transmit the current battery voltage value to the AI ​​inference and control module. The AI ​​inference and control module is also used to solve for the target value of audio transmission bit rate and the target value of effective value of bone conduction speaker driving voltage that minimize the overall cost when the current battery voltage value is lower than a preset threshold, based on the total power consumption of static power consumption, Bluetooth communication module power consumption, and audio processing and power amplifier module power consumption, under preset constraints.

7. The lollipop Bluetooth speaker with NFC and AI control according to claim 6, characterized in that, The overall cost is calculated from the highest transmission bit rate, the current transmission bit rate, the nominal voltage of the bone conduction speaker, the effective value of the current driving voltage, the weighting coefficient, and the estimated total harmonic distortion.

8. The lollipop Bluetooth speaker with NFC and AI control according to claim 7, characterized in that, The total harmonic distortion (THD) estimate is calculated from the base distortion coefficient, drive-rate coupling coefficient, normalized reference voltage, current effective value of drive voltage, maximum transmission rate, and current transmission rate.

9. A control method for a lollipop Bluetooth speaker, the lollipop Bluetooth speaker comprising a replaceable candy head and a handle with a built-in cavity, characterized in that, The control method includes: The reading process involves obtaining the acoustic model parameters of the candy head from the NFC tag built into the candy head via the NFC reading module inside the handle. The data acquisition process involves collecting the user's biting pressure value through a pressure sensor in the biting area of ​​the handle, and collecting the handle tilt value through a tilt sensor inside the handle. The frequency offset calculation step involves using the AI ​​inference and control module within the handle to calculate the oral cavity frequency offset based on the parameters of the candy head acoustic model, the difference between the biting pressure value and the preset resting reference pressure baseline, and the difference between the tilt angle value and the preset reference tilt angle baseline. In the gain adjustment step, the AI ​​inference and control module generates a gain adjustment signal based on the oral resonant cavity frequency offset and a preset compensation strategy. The audio adjustment and playback process involves receiving the audio signal to be played through the audio processing and amplifier module inside the handle, equalizing the audio signal according to the gain adjustment signal, amplifying it, and driving the bone conduction speaker inside the handle, which is close to the inner wall of the cavity, to play the audio.

10. The control method of claim 9, wherein, The control method further includes: a voltage monitoring step, in which the current battery voltage value is transmitted from the power module to the AI ​​inference and control module; a low-power decision step, in which, when the current battery voltage value is lower than a preset threshold, the AI ​​inference and control module calculates the target value of the audio transmission bitrate and the target value of the effective value of the bone conduction speaker driving voltage to minimize the overall cost based on the total power consumption of the static power consumption, the power consumption of the Bluetooth communication module, and the power consumption of the audio processing and power amplifier module, under preset constraints; and a degraded playback step, in which the AI ​​inference and control module adjusts the transmission bitrate of the Bluetooth communication module to the target value and limits the effective value of the bone conduction speaker driving voltage to the target value to maintain continuous audio playback.

Images