Intelligent earphone integrated with temperature and pressure sensing and wearing detection method thereof

Abstract

The invention discloses an intelligent earphone integrated with temperature and pressure sensing and a wearing detection method thereof, and belongs to the technical field of intelligent wearing equipment. The invention relates to an intelligent earphone integrated with temperature and pressure sensing and a wearing detection method thereof, the intelligent earphone comprises a sound production unit, the sound production unit is fixedly installed on the central axis of a shell, a sensing unit is tightly attached to the contact part of the outer surface of the shell and the ear canal of a human body, and the sensing unit is electrically connected with a signal acquisition and processing module through a flexible circuit. And the judgment module is connected with the signal acquisition and processing module in a wireless mode. The problems that in the prior art, objective wearing detection lacks or only depends on single parameter detection, objective comfort judgment does not exist, dynamic adaptation is poor, and a sensor is not good are solved. According to the invention, the earphone wearing comfort judgment, the falling prevention and the body temperature monitoring are realized, the limitation of only depending on single parameter detection is broken through, and the wearing comfort is improved by using the multifunctional flexible sensor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent wearable devices, in particular to an intelligent earphone integrated with temperature and pressure sensing and a wearing detection method thereof. BACKGROUND

[0002] With the rapid development of intelligent wearable technology, intelligent earphones have gradually evolved from traditional audio playback devices to smart terminals integrating audio interaction, health monitoring, and motion assistance. In various scenarios such as daily use and exercise, users have higher requirements for the wearing stability, comfort, and additional functions of intelligent earphones. However, existing intelligent earphone products still have many technical limitations in wearing detection and function integration, making it difficult to meet the diversified needs of users.

[0003] In the prior art, the wearing detection method of intelligent earphones mainly has the following defects: first, most products rely on a single parameter for wearing state determination, and single parameter detection is easily disturbed by the environment, with poor robustness in dynamic scenarios. Second, the existing technology generally lacks an objective determination mechanism for wearing comfort, and mostly relies on user subjective feedback, which cannot determine uncomfortable states such as tightness, looseness, or skewed compression in real time and quantitatively. Third, the dynamic adaptability of existing wearing detection methods is poor, and in dynamic scenarios such as walking, jogging, running, talking, and coughing, the detection accuracy decreases significantly, the fall detection delay is long, and the anti-falling warning cannot be triggered in time. At the same time, most existing sensors use rigid structure design, have poor fit, poor wearing comfort, and insufficient sweat and waterproof performance, making it difficult to adapt to complex use environments such as exercise; therefore, it does not meet the existing needs, and for this reason, we propose an intelligent earphone integrated with temperature and pressure sensing and a wearing detection method thereof. SUMMARY

[0004] The present application aims to provide an intelligent earphone integrated with temperature and pressure sensing and a wearing detection method thereof, which realizes the determination of earphone wearing comfort, the anti-falling detection of wearing state, and the body temperature monitoring of users, breaks through the limitations of the prior art in lacking objective wearing detection or relying only on single parameter detection, and the limitations of existing earphones in single mode detection, realizes real-time objective determination of wearing comfort, maintains high accuracy and fast response in dynamic exercise scenarios, effectively reduces the misjudgment rate, realizes the dual core functions of earphone temperature abnormality warning and intelligent anti-falling, and while ensuring detection accuracy, considers the use comfort and cost controllability, the sensor uses a flexible structure to improve wearing comfort, and solves the problems raised in the above background technology.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical method: an intelligent earphone integrated with temperature and pressure sensing, comprising a shell, a sound generating unit and a sensing unit, the sound generating unit is fixedly installed on the axis of the shell, the sensing unit is closely attached to the part of the outer surface of the shell which contacts the ear canal of the human body, a signal acquisition and processing module is packaged in the shell, the sensing unit is electrically connected with the signal acquisition and processing module through a flexible circuit, and a determination module is in communication connection with the signal acquisition and processing module through a wireless mode;

[0006] The sensing unit is a multifunctional flexible sensor integrated with temperature and pressure coupling, an independent flexible sensor adopts a three-dimensional porous sensitive structure constructed based on a melamine foam framework, PEDOT:PSS and a carbon nanotube conductive composite material, and is used for comprehensively collecting pressure distribution signals and temperature distribution signals of the part of the earphone which contacts the ear canal.

[0007] The signal acquisition and processing module is used for amplifying and filtering the analog signals output by the sensing unit, converting the processed analog signals into digital signals, pre-processing the digital signals and transmitting the digital signals to the determination module, so as to realize data interaction and instruction transmission with an external terminal device.

[0008] The determination module is used for fusion processing of the pressure signals and temperature signals transmitted by the signal acquisition and processing module, and outputs a wearing state determination result, realizes continuous monitoring of body temperature based on the calibrated temperature signals, and triggers a health warning when the body temperature is abnormal.

[0009] Preferably, the conductive dispersion liquid is prepared by mixing a PEDOT:PSS water dispersion, dimethyl sulfoxide, a surfactant and a carbon nanotube solution.

[0010] Preferably, the construction step of the conductive sensitive network is specifically as follows:

[0011] After the components are mixed according to the above-mentioned mass percentage, they are placed in an ultrasonic cleaner for ultrasonic dispersion for 30 min to 60 min, and the ultrasonic power is 150 W to 200 W, so as to obtain a uniform conductive dispersion liquid;

[0012] The modified flexible porous framework is placed in a vacuum impregnation tank, and the conductive dispersion liquid is added, so that the conductive dispersion liquid completely immerses the foam framework;

[0013] The vacuum impregnation tank is closed, and vacuum is extracted to a pressure of-0.08 MPa to-0.1 MPa, and maintained for 30 min to 60 min, so that the conductive dispersion liquid is fully penetrated into the three-dimensional porous network of the foam framework;

[0014] The vacuum is released, the foam framework is taken out, the excess conductive dispersion liquid on the surface is removed, and then it is placed in an oven at 100 DEG C to 120 DEG C for drying for 1 h to 2 h, to complete a cycle of impregnation to drying.

[0015] According to the required conductivity and sensitivity, repeat the above-mentioned immersion to dry cycle 2-4 times, and finally form a composite conductive foam.

[0016] Preferably, the preparation process of the independent flexible sensor comprises:

[0017] Preparation and modification of flexible porous framework: the flexible porous framework takes commercial melamine foam as the base, is cut into shape, and then is immersed in dopamine-Tris buffer solution for surface modification;

[0018] Construction of conductive sensitive network: the modified flexible porous framework is placed in a conductive dispersion liquid, vacuum impregnated, heat treated, dried and solidified to form a conductive sensitive network;

[0019] Packaging and integration: the packaging process adopts a polydimethylsiloxane protective packaging layer to coat the conductive sensitive network.

[0020] Preferably, the protective packaging layer adopts polydimethylsiloxane material, and the mass ratio of polydimethylsiloxane prepolymer to curing agent is 10:1. The preparation steps of the protective packaging layer are specifically as follows:

[0021] Mix the polydimethylsiloxane prepolymer and the curing agent uniformly, and vacuum degas for 15-30 min;

[0022] Use a spin coating process to coat the mixed liquid on the surface of the composite conductive foam, the spin coating speed is 2000-3000 r / min, and the spin coating time is 30-60 s;

[0023] After coating, place it in an oven at 80-100°C for 2-3 h to form a protective packaging layer with a thickness of 50-100 μm.

[0024] A wearing detection method applied in the intelligent earphone integrated with temperature and pressure sensing, comprising the following steps:

[0025] S1: the user wears the intelligent earphone and keeps still for more than 30 seconds, collects the pressure signal and temperature signal of the ear canal contact part, and pre-processes the pressure signal, and statistically analyzes the pre-processed pressure signal;

[0026] S2: collect the real-time pressure data P and real-time temperature data T output by the sensing unit, and the determination module calls the built-in threshold determination algorithm to compare the real-time pressure data P with the pressure reference P0 and the preset tight threshold and loose threshold , and preliminarily determine the wearing comfort degree combined with the stability of the temperature signal;

[0027] S3: When the initial judgment is that the state is suspected to be detached, the judgment module continuously collects 10 sets of real-time temperature data T to verify whether the real-time temperature T drops rapidly to close to the ambient temperature T ring.

[0028] S4: If the real-time pressure data satisfies P<0.1P0 and the duration is >100ms, and the real-time temperature data satisfies |TTring|<0.5℃ and the duration is >100ms, then the detachment status is finally confirmed, and an anti-detachment warning message is sent to the external terminal device;

[0029] S5: If only the pressure signal meets the suspected detachment condition, but the temperature signal does not, it is determined to be a pressure signal fluctuation, and the process returns to S2 to continue real-time monitoring.

[0030] S6: If only the temperature signal meets the condition of being close to the ambient temperature, while the pressure signal does not meet the suspected detachment condition, it is determined to be temperature signal interference, and the process returns to S2 to continue real-time monitoring.

[0031] S7: When wearing smart headphones in dynamic scenarios, the judgment module calls a multimodal deep learning fusion algorithm to further process the preprocessed pressure and temperature time series signals.

[0032] Preferably, S1 specifically includes:

[0033] The sensing unit begins to collect pressure and temperature signals at the ear canal contact point, and the collection time is 30 seconds.

[0034] The signal acquisition and processing module preprocesses the acquired raw signal, performs statistical analysis on the preprocessed pressure signal, removes the maximum and minimum values ​​within 30 seconds, and takes the average value of the remaining pressure signal as the wearing pressure reference P0.

[0035] Before the user wears the headphones, the ambient temperature T-ring is collected and stored. When collecting the ambient temperature, the sensing unit is exposed to the air for 10 seconds, and the average value is taken as the final ambient temperature T-ring.

[0036] Preferably, the multimodal deep learning fusion algorithm is implemented based on a deep learning model, which includes a pressure signal processing branch, a temperature signal processing branch, and a feature fusion layer.

[0037] Preferably, the training process of the deep learning model specifically includes:

[0038] We collected time-series pressure and temperature signals under different scenarios, manually labeled the collected samples, and used three methods—time stretching, additive noise, and signal shifting—to enhance the dataset.

[0039] The enhanced dataset is divided into a training set, a validation set, and a test set in the ratio of 7:2:1. The training set is used to train the deep learning model, and the validation set and the test set are used to validate and test the deep learning model during the training process;

[0040] The hyperparameter combination that maximizes the accuracy of the validation set is found through grid search and used as the hyperparameters of the final model to obtain the final deep learning model.

[0041] Preferably, the specific logic for the preliminary determination of wearing comfort is as follows:

[0042] If the real-time pressure data P > , and the duration of this state exceeds 5 minutes, and at the same time the real-time temperature data T is stably higher than the ambient temperature T_env, it is determined as an over-tight state;

[0043] If the real-time pressure data P < , and the duration of this state exceeds 1 minute, it is determined as an over-loose state;

[0044] If the single-point pressure value collected by any one of the independent flexible sensors > 2P0, and the pressure values collected by the remaining sensors are lower than 50% of P0, it is determined as a skewed compression state;

[0045] If the real-time pressure data P suddenly drops to near 0 or equals the ambient pressure, it is determined as a suspected detachment state;

[0046] If the real-time pressure data P is within , and the fluctuation amplitude of the pressure values collected by 8 independent flexible sensors is all < 20% of P0, and at the same time the real-time temperature data T is stably higher than the ambient temperature T_env, it is determined as a moderate state.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This invention uses a multifunctional flexible sensor to collect signals and integrates threshold judgment and a deep learning model to determine the comfort of wearing headphones, detect the wearing status to prevent them from falling off, and monitor the user's body temperature. It overcomes the limitations of existing technologies that lack objective wearing detection or rely solely on single-parameter detection, as well as the limitations of existing headphone single-modal detection. It achieves real-time objective judgment of wearing comfort. By judging pressure distribution, it avoids relying solely on subjective perception and can maintain high accuracy and fast response in dynamic motion scenarios, effectively reducing the false judgment rate. It constructs an active protection system through dual-modal collaborative detection of temperature and pressure, integrates body temperature monitoring function, and can achieve health monitoring without additional equipment. It realizes the dual core functions of abnormal headphone temperature warning and intelligent anti-fall-off, and while ensuring detection accuracy, it also takes into account the comfort of use and cost controllability. It can be adapted to various application scenarios such as daily use and sports. The multifunctional flexible sensor adopts a flexible structure, which improves wearing comfort and has sweat and water resistance. Attached Figure Description

[0049] Figure 1 This is a top view of the overall structure of the earphone and sensor of the present invention;

[0050] Figure 2 This is a side view of the overall structure of the earphone and sensor of the present invention;

[0051] Figure 3 This is a performance curve of the sensor foam compression assembly of the present invention;

[0052] Figure 4 This is a piezoresistive performance curve of the sensor of the present invention;

[0053] Figure 5 This is a graph showing the thermoelectric performance of the sensor of the present invention;

[0054] Figure 6 This is a timing diagram of the dynamic pressure response of the upper ear canal according to the present invention;

[0055] Figure 7 This is a timing diagram of the dynamic pressure response in the lower ear canal according to the present invention;

[0056] Figure 8 This is a schematic diagram of the wear detection method of the present invention.

[0057] In the diagram: 1. Sensing unit; 2. Housing; 3. Sound generating unit. Detailed Implementation

[0058] The technical methods of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0059] To address the issues of existing technologies lacking objective wear detection or relying solely on a single parameter, resulting in no objective comfort assessment, poor dynamic adaptation, and subpar sensors, please refer to [link to relevant documentation]. Figures 1-8 This embodiment provides the following technical methods:

[0060] A smart earphone integrating temperature and pressure sensing includes a shell 2, a sound-generating unit 3, and a sensing unit 1. The sound-generating unit 3 is fixedly installed on the central axis of the shell 2. The sensing unit 1 is closely fitted to the outer surface of the shell 2 at the part that contacts the human ear canal. A signal acquisition and processing module is encapsulated inside the shell 2. The sensing unit 1 is electrically connected to the signal acquisition and processing module through a flexible circuit. The judgment module and the signal acquisition and processing module establish a communication connection wirelessly.

[0061] Among them, sensing unit 1 is an integrated temperature and pressure coupling multifunctional flexible sensor. The independent flexible sensor adopts a three-dimensional porous sensitive structure based on melamine foam framework, PEDOT:PSS and carbon nanotube conductive composite material. The multifunctional flexible sensor consists of 8 independent flexible sensors with the same structure. The 8 independent flexible sensors are evenly distributed along the circumference of the outer shell 2 cylinder wall. The central angle between two adjacent independent flexible sensors is 45°, and the detection surface of the 8 independent flexible sensors all face outwards, which is used to comprehensively collect the pressure distribution signal and temperature distribution signal of the contact part between the earphone and the ear canal.

[0062] The signal acquisition and processing module includes a signal conditioning circuit, an analog-to-digital converter, a microprocessor, and a wireless communication module, which are connected in sequence. The signal conditioning circuit amplifies and filters the analog signal output from sensor unit 1. The analog-to-digital converter converts the processed analog signal into a digital signal. The microprocessor preprocesses the digital signal and transmits it to the judgment module. The wireless communication module enables data interaction and command transmission with external terminal devices. The signal conditioning circuit includes a pressure signal conditioning branch and a temperature signal conditioning branch. Both branches have the same structure, consisting of an instrumentation amplifier, a low-pass filter, and a voltage follower. The instrumentation amplifier uses an AD8237 with an adjustable gain range of 1 to 1000 times, and is used to amplify the weak signal output from sensor unit 1. The low-pass filter is an active RC low-pass filter with a cutoff frequency of 10Hz to 20Hz, used to remove high-frequency noise; the voltage follower is an OPA333, used to improve the signal's load-carrying capacity; the analog-to-digital converter is an ADS1256, a 24-bit high-precision analog-to-digital converter with an adjustable sampling rate from 10SPS to 38400SPS. The sampling rate is set to 100Hz, and the resolution is not less than 16 bits; the analog-to-digital converter and the microprocessor transmit data via an SPI interface. The microprocessor is a low-power ARM Cortex-M4 series chip, model STM32L476RG, with 128KB Flash and 40KB RAM, supporting low-power mode, and power consumption in sleep mode does not exceed 10μA;

[0063] The wireless communication module uses the nRF52840 chip, supports Bluetooth 5.2 protocol, and is also compatible with BLE (Bluetooth Low Energy mode). The maximum transmission rate is 2Mbps and the communication distance is no less than 10m. The wireless communication module is connected to the microprocessor through the UART interface to transmit the wearing status determination result, body temperature monitoring data and early warning information to the external terminal device, and at the same time receive control commands sent by the external terminal device.

[0064] The determination module is used to fuse and process the pressure and temperature signals transmitted by the signal acquisition and processing module, and output the wearing status determination result. The wearing status includes four types: too tight, too loose, moderate, and fallen off. At the same time, the determination module is also used to continuously monitor body temperature based on the calibrated temperature signal and trigger a health warning when the body temperature is abnormal.

[0065] The conductive dispersion is prepared by mixing PEDOT:PSS aqueous dispersion, dimethyl sulfoxide, surfactant and carbon nanotube solution. The mass percentages of each component are as follows: PEDOT:PSS aqueous dispersion 40%–60%, dimethyl sulfoxide 10%–20%, surfactant 0.5%–2%, and carbon nanotube solution 20%–40%. Among them, the solid content of PEDOT:PSS aqueous dispersion is 1%–3%, sodium dodecylbenzenesulfonate is selected as surfactant, and the concentration of carbon nanotubes in carbon nanotube solution is 0.5 mg / mL–2 mg / mL. The carbon nanotubes are single-walled carbon nanotubes with a diameter of 1 nm–2 nm and a length of 5 μm–10 μm.

[0066] The specific steps for constructing a conductive sensing network are as follows:

[0067] After mixing the components according to the above mass percentages, place them in an ultrasonic cleaner and ultrasonically disperse for 30 min to 60 min at an ultrasonic power of 150 W to 200 W to obtain a uniform conductive dispersion.

[0068] The modified flexible porous skeleton was placed in a vacuum impregnation tank, and a conductive dispersion was added to completely submerge the foam skeleton.

[0069] Close the vacuum impregnation tank, evacuate to a pressure of -0.08MPa to -0.1MPa, and maintain for 30 to 60 minutes to allow the conductive dispersion to fully penetrate into the three-dimensional porous network of the foam skeleton.

[0070] Release the vacuum, remove the foam skeleton, remove excess conductive dispersion from the surface, and then place it in an oven at 100℃~120℃ to dry for 1h~2h, completing one cycle of impregnation to drying;

[0071] Depending on the required conductivity and sensitivity, the above impregnation and drying cycle is repeated 2 to 4 times to finally form a composite conductive foam, which serves as the sensitive core of an independent flexible sensor.

[0072] The fabrication process of the independent flexible sensor includes:

[0073] Preparation and modification of flexible porous framework: The flexible porous framework was prepared by cutting and shaping commercial melamine foam and then impregnating it with dopamine-Tris buffer solution for surface modification.

[0074] Construction of conductive sensing network: The conductive sensing network is formed by placing the modified flexible porous framework in a conductive dispersion, followed by vacuum impregnation, heat treatment, drying and curing.

[0075] Packaging and Integration: The packaging process uses a polydimethylsiloxane protective encapsulation layer to cover the conductive sensitive network.

[0076] The protective encapsulation layer uses polydimethylsiloxane, with a mass ratio of polydimethylsiloxane prepolymer to curing agent of 10:1. The specific preparation steps of the protective encapsulation layer are as follows:

[0077] Mix the polydimethylsiloxane prepolymer and curing agent evenly, and then degas under vacuum for 15 to 30 minutes.

[0078] The mixture was coated onto the surface of the composite conductive foam using a spin coating process. The spin coating speed was 2000 r / min to 3000 r / min, and the spin coating time was 30 s to 60 s.

[0079] After coating, it is placed in an oven at 80℃~100℃ for 2h~3h to cure, forming a protective encapsulation layer with a thickness of 50μm~100μm; the Shore hardness of the protective encapsulation layer is A10~A20, and it has good flexibility, sweat resistance and water resistance, with a waterproof rating of not less than IPX5.

[0080] A wear detection method is applied to a smart earphone that integrates temperature and pressure sensing, including the following steps:

[0081] S1: The user wears the smart earphone and remains still for more than 30 seconds. The pressure and temperature signals at the ear canal contact point are collected and preprocessed. The preprocessed pressure signals are then statistically analyzed.

[0082] S2: Acquire real-time pressure data P and real-time temperature data T output by sensor unit 1, and transmit a set of processed pressure and temperature data to the judgment module every 500ms; the judgment module calls the built-in threshold judgment algorithm to compare the real-time pressure data P with the pressure reference P0 and the preset over-tightness threshold. Loose threshold By comparing the results and considering the stability of the temperature signal, a preliminary assessment of wearing comfort can be made.

[0083] Among them, the overly tight threshold With loose threshold The method for determining the comfortable wearing pressure range is as follows: A large sample experiment was conducted to collect the comfortable wearing pressure range of users with different ear canal sizes. The sample size of the experiment was no less than 100 people, covering different age groups and genders from 18 to 60 years old, and ear canal sizes were divided into small, medium, and large types. The collected comfortable wearing pressure ranges were statistically analyzed, and 1.2 times the upper limit of the comfortable pressure range was taken as the excessively tight threshold. Take 0.8 times the lower limit of the comfort pressure range as the over-loose threshold. The system has built-in corresponding features for different ear canal sizes. and Users can select the matching parameter type according to their own ear canal size, or the system can automatically identify and match it;

[0084] S3: When the initial judgment is that the condition is suspected to be detached, the judgment module continuously collects 10 sets of real-time temperature data T (the collection interval is 10ms) to verify whether the real-time temperature T drops rapidly to close to the ambient temperature T ring (|TT ring|<0.5℃).

[0085] S4: If the real-time pressure data satisfies P<0.1P0 (or equals the ambient pressure) and the duration is >100ms, and the real-time temperature data satisfies |TT ring|<0.5℃ and the duration is >100ms, then the detachment status is finally confirmed, and an anti-detachment warning message is sent to the external terminal device. The warning message includes text prompts and sound prompts. The external terminal device will trigger the reminder immediately after receiving the message.

[0086] S5: If only the pressure signal meets the suspected detachment condition, but the temperature signal does not (|TT ring|≥2℃), then it is determined to be a pressure signal fluctuation, and return to S2 to continue real-time monitoring;

[0087] S6: If only the temperature signal meets the condition of being close to ambient temperature, while the pressure signal does not meet the suspected detachment condition, then it is determined to be temperature signal interference, and the process returns to S2 to continue real-time monitoring;

[0088] S7: When wearing smart headphones in dynamic scenarios, the judgment module calls a multimodal deep learning fusion algorithm to further process the preprocessed pressure and temperature time series signals.

[0089] S1 specifically includes:

[0090] Sensing unit 1 begins to collect pressure and temperature signals at the ear canal contact point for 30 seconds.

[0091] The signal acquisition and processing module preprocesses the acquired raw signal. The preprocessing process includes: first, using mean filtering to process the raw signal with a filtering window size of 5 to 10 sampling points to remove high-frequency noise from the signal; then, using wavelet transform, selecting the db4 wavelet as the base wavelet, to perform three-level decomposition and reconstruction of the signal to eliminate baseline drift.

[0092] Statistical analysis is performed on the preprocessed pressure signal, the maximum and minimum values ​​within 30 seconds are removed, the average value of the remaining pressure signal is taken as the wearing pressure reference P0, and stored in the Flash memory of the microprocessor;

[0093] Before the user wears the headphones, the ambient temperature T-ring is collected and stored. When collecting the ambient temperature, the sensing unit 1 is exposed to the air for 10 seconds. The average value is taken as the final ambient temperature T-ring, which is used for subsequent temperature calibration and status judgment.

[0094] The multimodal deep learning fusion algorithm is implemented based on a deep learning model, which includes a pressure signal processing branch, a temperature signal processing branch, and a feature fusion layer. The pressure and temperature signal processing branches use the same network structure, a combination of convolutional neural networks and long short-term memory networks. The convolutional neural network part includes 3 convolutional layers, 3 batch normalization layers, and 3 pooling layers. The first convolutional layer has 32 kernels, a kernel size of 3×1, and a stride of 1. The second convolutional layer has... The first convolutional layer has 64 kernels, a kernel size of 3×1, and a stride of 1. The third convolutional layer has 128 kernels, a kernel size of 3×1, and a stride of 1. Each convolutional layer is followed by a normalization layer using the ReLU activation function. The pooling layer uses max pooling with a kernel size of 2×1 and a stride of 2. The output of the convolutional neural network serves as the input to the long short-term memory network. The long short-term memory network consists of two hidden layers, each with 64 nodes, using the tanh activation function, and has an output dimension of 64.

[0095] The feature fusion layer employs an attention mechanism to achieve weighted fusion of features from two branches. Specifically, the process is as follows: First, the 64-dimensional feature vectors output from the pressure signal processing branch and the temperature signal processing branch are concatenated to obtain a 128-dimensional concatenated feature vector. Then, a fully connected layer maps the concatenated feature vector to a 64-dimensional attention weight vector, where each element corresponds to a weight of an input feature. Finally, the concatenated feature vector and the attention weight vector are multiplied to obtain a weighted fused 64-dimensional feature vector. The fused feature vector is then input to the fully connected layer, passes through a Softmax activation function, and outputs the final wearing status classification result. The classification results include four types: too tight, too loose, moderate, and dislodged. The output is a probability distribution, and the category with the highest probability is taken as the final judgment.

[0096] The training process of a deep learning model specifically includes:

[0097] The system collects time-series signals of pressure and temperature under different scenarios, including static wearing, walking, jogging, running, talking, coughing, rapid falling off, and slow falling off. At least 1000 samples are collected for each scenario, with each sample lasting 5 seconds and a sampling frequency of 100Hz. Each sample contains 500 pressure data points and 500 temperature data points. The collected samples are manually labeled, including the wearing status category (too tight, too loose, moderate, fallen off) and the scenario type.

[0098] Three methods were used to augment the dataset: time stretching, additive noise, and signal shifting. Time stretching involves stretching or compressing the original signal's duration to 0.8–1.2 times its original length. Additive noise involves adding Gaussian white noise with an amplitude of 5%–10% of the original signal's maximum amplitude to the original signal. Signal shifting involves shifting the original signal ±50 sampling points along the time axis. Through data augmentation, the dataset size was increased to 3 times its original size.

[0099] The augmented dataset was divided into training, validation, and test sets in a 7:2:1 ratio; the initial learning rate of the deep learning model was set to 0.001, and the Adam optimizer was used. The parameter is 0.9. The parameters are 0.999 and the weight decay coefficient is 0.0001. The cross-entropy loss function is used as the optimization objective of the deep learning model. The number of training rounds is 50. After each round of training, the performance of the deep learning model is verified using a validation set. Training is stopped when the accuracy of the validation set is stable above 95% for 5 consecutive rounds.

[0100] The hyperparameters of the deep learning model were adjusted using a grid search method. The hyperparameters to be optimized included the number of convolutional kernels in the convolutional neural network, the kernel size, the number of hidden nodes in the long short-term memory network, and the attention mechanism weight coefficients. The parameter range for the grid search was: 32–256 convolutional kernels, 3×1–7×1 kernel size, 32–128 hidden nodes in the long short-term memory network, and 0.1–0.9 attention weight coefficients. The grid search was used to find the hyperparameter combination that achieved the highest accuracy on the validation set, which was then used as the hyperparameters of the final deep learning model.

[0101] The specific logic for the initial assessment of wearing comfort is as follows:

[0102] If the real-time pressure data P> If this state lasts for more than 5 minutes, and the real-time temperature data T is consistently higher than the ambient temperature Tring (|TTring|≥3℃), then it is determined to be an over-tight state.

[0103] If the real-time pressure data P < If this state lasts for more than 1 minute, it is judged as an overly loose state;

[0104] If the single-point pressure value collected by any one of the eight independent flexible sensors is greater than 2P0, and the pressure values ​​collected by at least the other six sensors are less than 50% of P0, then it is determined to be a state of skewed compression.

[0105] If the real-time pressure data P suddenly drops to near 0 (P<0.1P0) or equals the difference between the ambient pressure (and the pressure value collected when the headphones are not worn) and ≤0.05P0, it is judged as a suspected detachment state;

[0106] If the real-time pressure data P is within the range, and the fluctuation amplitude of the pressure values collected by 8 independent flexible sensors is all < 20% of P0, and at the same time the real-time temperature data T is stably higher than the ambient temperature T_env (|T - T_env| ≥ 3°C), it is determined to be in a moderate state.

[0107] In terms of body temperature monitoring, the determination module uses a temperature drift compensation algorithm to calibrate the real-time temperature signal to improve the accuracy of body temperature monitoring; the compensation coefficient of the temperature drift compensation algorithm is obtained through experimental calibration: at different ambient temperatures (5°C to 40°C), a standard body temperature simulator is selected to simulate different body temperature values (36°C to 39°C), the difference between the temperature detection value of the sensor and the standard body temperature value is collected, and a mapping table of ambient temperature and compensation coefficient is established; during actual use, the system queries the mapping table according to the collected ambient temperature T_env to obtain the corresponding compensation coefficient k; the correction formula for temperature drift compensation is: T_corrected = T + k × (T_env - T_std), where T_std is the standard ambient temperature (25°C); the corrected body temperature value T_corrected is used as the final body temperature monitoring value, and the body temperature data is recorded once every 1 minute and stored in the Flash memory of the microprocessor, and the body temperature data of the last 7 days can be stored; when T_corrected continuously exceeds 37.3°C and the duration exceeds 10 minutes, the determination module triggers a fever trend warning and sends a warning message to an external terminal device through a wireless communication module to remind the user to pay attention to their health status.

[0108] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0109] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made in these embodiments without departing from the principles and spirit of the present invention.

Claims

1. A smart earphone integrating temperature and pressure sensing, comprising a shell (2), a sound-generating unit (3), and a sensing unit (1), characterized in that, The sound-generating unit (3) is fixedly installed on the central axis of the outer shell (2). The sensing unit (1) is closely attached to the outer surface of the outer shell (2) where it contacts the human ear canal. The outer shell (2) is encapsulated with a signal acquisition and processing module. The sensing unit (1) is electrically connected to the signal acquisition and processing module through a flexible circuit. The judgment module and the signal acquisition and processing module establish a communication connection wirelessly. Among them, the sensing unit (1) is an integrated temperature and pressure coupling multifunctional flexible sensor. The independent flexible sensor adopts a three-dimensional porous sensitive structure based on melamine foam framework, PEDOT:PSS and carbon nanotube conductive composite material, which is used to comprehensively collect the pressure distribution signal and temperature distribution signal of the contact part between the earphone and the ear canal. The signal acquisition and processing module is used to amplify and filter the analog signal output by the sensing unit (1), convert the processed analog signal into a digital signal, preprocess the digital signal and transmit it to the judgment module, so as to realize data interaction and instruction transmission with external terminal equipment. The determination module is used to fuse the pressure signal and temperature signal transmitted by the signal acquisition and processing module, output the wearing status determination result, realize continuous body temperature monitoring based on the calibrated temperature signal, and trigger a health warning when the body temperature is abnormal.

2. The smart earphone integrating temperature and pressure sensing according to claim 1, characterized in that, The conductive dispersion is prepared by mixing PEDOT:PSS aqueous dispersion, dimethyl sulfoxide, surfactant and carbon nanotube solution.

3. The intelligent earphone integrating temperature and pressure sensing according to claim 2, characterized in that, The specific steps for constructing the conductive sensing network are as follows: After mixing the components according to the above mass percentages, place them in an ultrasonic cleaner and ultrasonically disperse for 30 min to 60 min at an ultrasonic power of 150 W to 200 W to obtain a uniform conductive dispersion. The modified flexible porous skeleton was placed in a vacuum impregnation tank, and a conductive dispersion was added to completely submerge the foam skeleton. Close the vacuum impregnation tank, evacuate to a pressure of -0.08MPa to -0.1MPa, and maintain for 30 to 60 minutes to allow the conductive dispersion to fully penetrate into the three-dimensional porous network of the foam skeleton. Release the vacuum, remove the foam skeleton, remove excess conductive dispersion from the surface, and then place it in an oven at 100℃~120℃ to dry for 1h~2h, completing one cycle of impregnation to drying; Depending on the required conductivity and sensitivity, repeat the above impregnation and drying cycle 2 to 4 times to finally form a composite conductive foam.

4. The intelligent earphone integrating temperature and pressure sensing according to claim 3, characterized in that, The fabrication process of the independent flexible sensor includes: Preparation and modification of flexible porous framework: The flexible porous framework is based on commercial melamine foam, which is cut and shaped and then impregnated with dopamine-Tris buffer solution for surface modification. Construction of conductive sensitive network: The modified flexible porous framework is placed in a conductive dispersion, and then vacuum impregnated, heat-treated and dried to form a conductive sensitive network; Packaging and Integration: The packaging process uses a polydimethylsiloxane protective encapsulation layer to cover the conductive sensitive network.

5. The intelligent earphone integrating temperature and pressure sensing according to claim 1, characterized in that, The protective encapsulation layer is made of polydimethylsiloxane material, and the mass ratio of polydimethylsiloxane prepolymer to curing agent is 10:

1. The specific preparation steps of the protective encapsulation layer are as follows: Mix the polydimethylsiloxane prepolymer and curing agent evenly, and then degas under vacuum for 15 to 30 minutes. The mixture was coated onto the surface of the composite conductive foam using a spin coating process. The spin coating speed was 2000 r / min to 3000 r / min, and the spin coating time was 30 s to 60 s. After coating, place it in an oven at 80℃~100℃ for 2h~3h to cure, forming a protective encapsulation layer with a thickness of 50μm~100μm.

6. A wear detection method applied to a smart earphone integrating temperature and pressure sensing as described in claim 5, characterized in that, Includes the following steps: S1: The user wears the smart earphone and remains still for more than 30 seconds. The pressure and temperature signals at the ear canal contact point are collected and preprocessed. The preprocessed pressure signals are then statistically analyzed. S2: The real-time pressure data P and real-time temperature data T output by the sensing unit (1) are collected. The judgment module calls the built-in threshold judgment algorithm to compare the real-time pressure data P with the pressure reference P0 and the preset over-tightness threshold. and loose threshold By comparing the results and considering the stability of the temperature signal, a preliminary assessment of wearing comfort can be made. S3: When the initial judgment is that the state is suspected to be detached, the judgment module continuously collects 10 sets of real-time temperature data T to verify whether the real-time temperature T drops rapidly to close to the ambient temperature T ring. S4: If the real-time pressure data satisfies P<0.1P0 and the duration is >100ms, and the real-time temperature data satisfies |TTring|<0.5℃ and the duration is >100ms, then the detachment status is finally confirmed, and an anti-detachment warning message is sent to the external terminal device; S5: If only the pressure signal meets the suspected detachment condition, but the temperature signal does not, it is determined to be a pressure signal fluctuation, and the process returns to S2 to continue real-time monitoring. S6: If only the temperature signal meets the condition of being close to the ambient temperature, while the pressure signal does not meet the suspected detachment condition, it is determined to be temperature signal interference, and the process returns to S2 to continue real-time monitoring. S7: When wearing smart headphones in dynamic scenarios, the judgment module calls a multimodal deep learning fusion algorithm to further process the preprocessed pressure and temperature time series signals.

7. The wearing detection method according to claim 6, characterized in that, S1 specifically includes: The sensing unit (1) begins to collect pressure and temperature signals at the ear canal contact point for 30 seconds. The signal acquisition and processing module preprocesses the acquired raw signal, performs statistical analysis on the preprocessed pressure signal, removes the maximum and minimum values ​​within 30 seconds, and takes the average value of the remaining pressure signal as the wearing pressure reference P0. Before the user wears the headphones, the ambient temperature T ring is collected and stored. When collecting the ambient temperature, the sensing unit (1) is exposed to the air. The collection time is 10 seconds, and the average value is taken as the final ambient temperature T ring.

8. The wearing detection method according to claim 6, characterized in that, The multimodal deep learning fusion algorithm is implemented based on a deep learning model, which includes a pressure signal processing branch, a temperature signal processing branch, and a feature fusion layer.

9. The wearing detection method according to claim 6, characterized in that, The training process of the deep learning model specifically includes: We collected time-series pressure and temperature signals under different scenarios, manually labeled the collected samples, and used three methods—time stretching, additive noise, and signal shifting—to enhance the dataset. The enhanced dataset was divided into training, validation, and test sets in a ratio of 7:2:

1. The deep learning model was trained using the training set, and the validation and test sets were used to validate and test the deep learning model during the training process. The hyperparameter combination that achieves the highest accuracy on the validation set is found through grid search and used as the hyperparameters of the final model, thus obtaining the final deep learning model.

10. The wearing detection method according to claim 6, characterized in that, The specific logic for the preliminary determination of wearing comfort is as follows: If the real-time pressure data P> If this state lasts for more than 5 minutes, and the real-time temperature data T is consistently higher than the ambient temperature Tring, then it is determined to be an over-tight state. If the real-time pressure data P < If this state lasts for more than 1 minute, it is judged as an overly loose state; If the single-point pressure value collected by any one of the independent flexible sensors is greater than 2P0, and the pressure values ​​collected by the other sensors are less than 50% of P0, then it is determined to be a state of skewed compression. If the real-time pressure data P suddenly drops to near 0 or equal to the ambient pressure, it is judged as a suspected detachment state; If the real-time pressure data P is within and the fluctuation range of the pressure values collected by 8 independent flexible sensors is all < 20% of P0, and at the same time the real-time temperature data T is stably higher than the ambient temperature Tenv, it is determined to be in a moderate state.

Images