Intelligent comb integrating multi-mode biological feature recognition and detection method

Abstract

The invention relates to the technical field of crossing of artificial intelligence and intelligent health equipment, and discloses an intelligent comb integrating multi-mode biological feature recognition and a detection method. According to the method, in the daily combing process, five types of biological feature signals including sebum secretion, keratin metabolism, local inflammation, hair physical characteristics and skin barrier integrity are synchronously collected; and after preprocessing, inputting the data into a multi-modal fusion analysis model, generating a multi-dimensional health score through a cross-modal attention mechanism, and triggering risk early warning. The system comprises a comb body structure conforming to ergonomics, a multi-mode sensing array, a signal processing unit, a local storage, a wireless communication and a power supply module, and supports user identity recognition and environment temperature and humidity compensation. According to the application, non-inductive, high-frequency and comprehensive scalp health quantitative monitoring is realized, and the early risk identification capability and the detection robustness are remarkably improved.

Description

Technical Field

[0001] This application relates to the interdisciplinary field of artificial intelligence and smart health devices, and more specifically, it relates to a smart comb and detection method that integrates multimodal biometric recognition. Background Technology

[0002] As people increasingly value personal health and image management, scalp and hair health has gradually become an important part of daily care. As one of the largest organs in the human body, the health of the scalp directly affects hair growth, appearance, texture, and overall comfort. Common scalp problems such as seborrheic dermatitis, folliculitis, hair loss, dryness, or allergies, if identified and intervened in their early stages, will significantly improve treatment outcomes and user experience. However, current mainstream scalp health assessment methods have significant limitations: on the one hand, while equipment such as dermoscopy and hair microscopy used by professional medical or beauty institutions possesses high-precision detection capabilities, they are expensive, complex to operate, and dependent on professional personnel, making it difficult to meet users' high-frequency and convenient monitoring needs in home settings; on the other hand, ordinary users rely on visual observation of dandruff, oiliness, or subjective feelings (such as itching or tightness) for judgment. This method lacks objective quantitative evidence and is often only noticed when symptoms are obvious, missing the crucial window for early intervention.

[0003] Among them, smart combs have received some attention in recent years as a potential daily health monitoring device. Existing products mostly focus on basic functions, such as roughly judging the dryness or wetness of hair using a single impedance sensor, or recording combing frequency using simple counting logic. Some products even integrate only a vibration motor to provide a massage experience. These solutions generally suffer from a single sensing modality and a lack of information dimensions, failing to comprehensively reflect the complex physiological state of the scalp. For example, relying solely on changes in conductivity is insufficient to distinguish between excessive sebum secretion and the wetness caused by inflammatory exudation; relying solely on visible light reflectance images also cannot accurately identify dandruff types (such as dry scales and sebaceous crusts) and their distribution density. More importantly, scalp health is essentially a complex state involving multiple coupled factors, including sebum metabolism, stratum corneum integrity, local immune response, hair follicle activity, and microenvironmental temperature and humidity. Data from a single sensor is easily affected by external interference and lacks a stable mapping relationship with the actual physiological state, resulting in low reliability and a high misjudgment rate.

[0004] Current technologies therefore face a core challenge: how to simultaneously acquire high-dimensional, highly reliable scalp biometric data during the natural act of daily grooming without increasing the user's operational burden, and to achieve accurate analysis of complex health conditions. Current smart combs lack multimodal sensing capabilities and fusion analysis mechanisms, failing to cover multidimensional indicators of scalp health and struggling to extract effective physiological signals in noisy environments, severely limiting their practical value in home health management scenarios. Therefore, there is an urgent need for a new type of smart comb and its detection method that can seamlessly integrate multiple sensing modalities such as optics, bioelectricity, and mechanics, and achieve non-intrusive, high-precision scalp health risk warnings through collaborative acquisition and intelligent fusion strategies. Summary of the Invention

[0005] This invention provides an intelligent comb and detection method integrating multimodal biometric recognition, aiming to solve the technical problems of inaccurate scalp health assessment and inability to achieve early warning due to single sensing modes and insufficient data dimensions in existing technologies. This invention achieves comprehensive, objective, quantitative, and early risk identification of scalp health by imperceptibly collecting multi-dimensional biometric signals during daily combing and constructing a multimodal fusion analysis model.

[0006] The smart comb includes a comb body structure, a multimodal biometric sensor array, a signal processing unit, a local storage unit, a wireless communication module, and a power supply unit. The comb body structure adopts an ergonomic streamlined design, and its comb teeth are made of medical-grade silicone-coated metal conductive material, with multiple micro sensors embedded inside to ensure natural contact with the scalp and hair during normal combing, completing imperceptible data collection.

[0007] The multimodal biometric sensor array includes a sebum secretion detection module, a keratin metabolism state detection module, a local inflammation response detection module, a hair physical property detection module, and a skin barrier integrity detection module. The sebum secretion detection module consists of multiple micro-area capacitive oil sensors, each with a sensing area of ​​0.5 to 1.5 square millimeters, used to measure changes in the dielectric constant of local areas of the scalp, thereby quantifying sebum secretion levels. The keratin metabolism state detection module includes a high-resolution optical imaging unit and an image preprocessing unit. The optical imaging unit consists of a blue light-emitting diode with a center wavelength of 470 nanometers and a monochrome image sensor with more than 5 million pixels, used to capture the morphology, size, and distribution density of dandruff on the scalp surface. The image preprocessing unit performs background suppression, edge enhancement, and binarization on the original image, outputting dandruff quantification indicators. The local inflammation response detection module consists of an infrared thermal imaging sensor and a near-infrared spectroscopy analysis unit. The infrared thermal imaging sensor has a spatial resolution of 0.1 degrees Celsius and is used to detect areas of abnormally elevated local scalp temperature. The near-infrared spectroscopy analysis unit emits a continuous spectrum with a wavelength range of 700 nanometers to 1000 nanometers and receives the reflected spectrum. By analyzing changes in the hemoglobin absorption peak, it determines the local microcirculation status and the degree of inflammation activity. The hair physical property detection module consists of a miniature tensile sensor and a vibration frequency analysis unit. The miniature tensile sensor is integrated into the root of the comb teeth and has a range of 0 to 5 millinewtons. It is used to measure the breaking force of a single or multiple hairs when they are pulled apart in real time during combing. The vibration frequency analysis unit excites the comb teeth to generate high-frequency micro-vibrations through a piezoelectric ceramic plate and detects the vibration decay curve to invert the hair elastic modulus and toughness parameters. The skin barrier integrity detection module consists of an AC impedance spectroscopy measurement unit. This unit applies a microampere-level AC excitation signal in the frequency range of 10 Hz to 100 kHz to measure the complex impedance spectrum of the scalp stratum corneum. By fitting a Cole-Cole equivalent circuit model, the hydration degree and barrier function parameters of the stratum corneum are extracted.

[0008] The signal processing unit is electrically connected to the multimodal biometric sensor array and is used to amplify, filter, convert analog-to-digital signals from each sensor, and synchronize their timing. The signal processing unit incorporates a multi-channel synchronous sampling controller with a uniform sampling frequency of 200 Hz to ensure strict alignment of all modal data on the time axis. The local storage unit caches at least thirty days of raw sensor data and intermediate processing results, using a non-volatile flash memory chip with a storage capacity of at least 8 gigabytes. The wireless communication module supports Bluetooth 5.0 protocol for transmitting processed feature data or warning information to user terminal devices. The power supply unit is a rechargeable lithium polymer battery with a rated voltage of 3.7 volts and a capacity of 300 mAh, charged via a magnetic interface.

[0009] The scalp health detection method includes the following steps: First, during daily grooming using a smart comb, the multimodal biometric sensor array simultaneously collects sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data; second, the collected raw data is preprocessed, including region normalization of sebum data, morphological segmentation of dandruff images, non-uniformity correction of infrared thermograms, baseline drift elimination of near-infrared spectra, peak extraction of hair tension signals, and equivalent circuit fitting of impedance spectra; third, the preprocessed modal feature vectors are input into a multimodal fusion analysis model, which consists of a feature encoder and cross-modal attention interaction... The system consists of three parts: a layer, a health status decoder, and a feature encoder. The feature encoder performs independent nonlinear mapping on the features of each modality to generate a high-dimensional semantic representation. The cross-modal attention interaction layer calculates the correlation weights between the representations of different modalities, dynamically weights and fuses key information, and suppresses noise interference. Based on the fused comprehensive representation, the health status decoder outputs a multi-dimensional score of scalp health status, including sebum secretion index, dandruff severity level, inflammation activity probability, hair strength decay rate, and barrier function integrity percentage. Finally, according to a preset health threshold rule library, the system performs risk assessment on each score. When any indicator exceeds the normal range, a corresponding health warning is generated and pushed to the user terminal via a wireless communication module.

[0010] As one embodiment of the present invention, the training process of the multimodal fusion analysis model includes: collecting a large-scale labeled dataset containing comb-sensing data from users of different ages, genders, and regions, and their corresponding clinical diagnostic results; adopting an end-to-end supervised learning strategy, using clinical diagnostic labels as supervision signals, optimizing model parameters to make the output score highly consistent with medical standards; and in the model deployment stage, using model distillation technology to compress the large training model into a lightweight inference model to adapt to the embedded processor inside the comb.

[0011] As one embodiment of the present invention, the comb body structure is provided with an environmental temperature and humidity compensation module. This module is composed of a digital temperature and humidity sensor, which is used to monitor the temperature and humidity of the environment around the comb in real time, and input the compensation parameters to the signal processing unit to eliminate the interference of environmental factors on biometric signals.

[0012] As one embodiment of the present invention, the smart comb also includes a user identification module. This module analyzes the mechanical characteristic sequence of the user when combing their hair, including the time distribution of force on the comb teeth, the force change pattern and the combing trajectory, to construct an individual biomechanical fingerprint, thereby realizing automatic identification of user identity and data isolation, and ensuring data privacy and personalized analysis in multi-user family scenarios.

[0013] As one embodiment of the present invention, the multi-dimensional score output by the health status decoder is further used to generate personalized scalp care suggestions, including shampooing frequency adjustment, hair care product recommendations, diet and rest tips, and medical warning levels. All suggestions are derived based on a preset expert knowledge rule base and user historical data trend analysis.

[0014] Compared with existing technologies, the advantages and positive effects of this invention are as follows: By integrating five biometric sensing modules—sebum, keratin, inflammation, hair, and scalp barrier—into a smart comb, this invention achieves comprehensive, seamless, and high-frequency monitoring of scalp health during daily grooming. The synchronous acquisition and cross-modal fusion analysis mechanism of multimodal data effectively overcomes the shortcomings of single sensors, such as susceptibility to interference and incomplete information, significantly improving the accuracy and robustness of the detection results. The quantitative assessment system constructed by this invention can identify the early risks of common scalp problems such as seborrheic dermatitis, folliculitis, and hair loss, advancing the intervention window to before symptoms appear. Simultaneously, the user identification mechanism based on individual biomechanical fingerprints and the localized data processing architecture ensure user privacy and security. The overall solution integrates professional-grade scalp health detection capabilities into everyday products, resolving the core contradictions of existing technologies such as the inconvenience of professional equipment, inaccurate subjective judgment, and the limited functionality of smart combs, providing a practical and feasible technical path for personal health management. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the steps of an embodiment of this application; Figure 2 This is a schematic diagram of electrical connections according to an embodiment of this application.

[0016] Reference numerals in the attached figures: 1. Multimodal biometric sensor array; 2. Environmental temperature and humidity compensation module; 3. User identification module; 4. Signal processing unit; 5. Power supply unit. Detailed Implementation

[0017] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0018] It should be noted that when a component is referred to as being "fixed to" or "set on" another component, it can be directly or indirectly attached to that other component. When a component is referred to as being "connected to" another component, it can be directly or indirectly connected to that other component.

[0019] It should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] This invention provides a scalp detection method integrating multimodal biometric recognition, aiming to solve the technical problems in the prior art that lead to inaccurate assessment of scalp health status and the inability to achieve early warning due to single sensing modes and insufficient data dimensions.

[0022] See Figure 1 The method collects multi-dimensional biometric signals imperceptibly during daily grooming and constructs a multimodal fusion analysis model to achieve comprehensive, objective, quantitative, and early risk identification of scalp health status. The method includes the following steps: S1, during the user's daily grooming process using a smart comb, sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data are simultaneously collected through a multimodal biometric sensor array; S2, the collected raw data are preprocessed, including region normalization of sebum data, morphological segmentation of dandruff images, non-uniformity correction of infrared thermograms, baseline drift elimination of near-infrared spectra, peak extraction of hair tension signals, and equivalent circuit fitting of impedance spectra; S3, the preprocessed modal feature vectors are input into a multimodal fusion analysis model, which consists of a feature encoder, a cross-modal attention interaction layer, and a health status decoder; S4, based on the comprehensive representation output by the multimodal fusion analysis model, a multi-dimensional score of scalp health status is generated, and risk is determined according to a preset health threshold rule base. When any indicator exceeds the normal range, a corresponding health warning message is generated and pushed to the user terminal through a wireless communication module.

[0023] In the aforementioned scalp health detection method, step S1 involves simultaneously collecting sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data through a multimodal biometric sensor array during the user's daily combing process. It should be understood that scalp health is a complex physiological state, with changes involving multiple dimensions such as sebaceous gland secretion activity, keratinocyte metabolic rate, local microcirculation, hair follicle structural integrity, and skin barrier function. Single-modal sensor data is insufficient to comprehensively reflect these interrelated physiological processes and is easily affected by environmental interference or individual differences. Therefore, this invention integrates five types of high-precision micro-sensors within the comb structure to form a multimodal biometric sensor array, ensuring seamless, synchronous, and multidimensional data acquisition during the user's natural combing action. Specifically, the sebum secretion detection module consists of multiple micro-area capacitive oil sensors, each with a sensing area of ​​0.5 to 1.5 square millimeters, distributed on the tips and sidewalls of the comb teeth to cover different scalp areas. When the comb teeth contact the scalp, sebum acts as a dielectric medium, filling the space between the sensor electrodes and causing a change in capacitance. This capacitance is positively correlated with sebum thickness and dielectric constant; by measuring the capacitance change, the local sebum secretion level can be quantified. The keratin metabolism state detection module includes a high-resolution optical imaging unit and an image preprocessing unit. The optical imaging unit consists of a blue light-emitting diode with a center wavelength of 470 nanometers and a monochrome image sensor with more than five million pixels. The blue light wavelength can effectively excite the fluorescence properties of dandruff on the scalp surface, enhancing the contrast between dandruff and the background. The image sensor continuously captures images of the scalp surface at a rate of 30 frames per second, with each frame covering an area of ​​approximately ten square centimeters. The local inflammation response detection module consists of an infrared thermal imaging sensor and a near-infrared spectroscopy analysis unit. The infrared thermal imaging sensor has a spatial resolution of 0.1 degrees Celsius and can detect areas of abnormally elevated local scalp temperature, which are usually associated with increased blood flow due to inflammation. The near-infrared spectroscopy analysis unit emits a continuous spectrum with wavelengths ranging from 700 nm to 1000 nm and receives the reflectance spectrum. By analyzing the absorption peak change of hemoglobin at 805 nm, it determines the local microcirculation status and the degree of inflammation activity. The hair physical property detection module consists of a miniature tensile sensor and a vibration frequency analysis unit. The miniature tensile sensor is integrated into the root of the comb teeth, with a range of 0 to 5 millinewtons, and is used to measure the breaking force of a single or multiple hairs when they are pulled apart in real time during combing. The vibration frequency analysis unit excites the comb teeth to generate high-frequency micro-vibrations with a frequency range of 1 kHz to 5 kHz through a piezoelectric ceramic plate, and detects the vibration decay curve to invert the hair elastic modulus and toughness parameters. The skin barrier integrity detection module consists of an AC impedance spectroscopy measurement unit, which applies a microampere-level AC excitation signal in the frequency range of 10 Hz to 100 kHz to measure the complex impedance spectrum of the scalp stratum corneum.All sensors are driven by the same clock source and are uniformly controlled by the multi-channel synchronous sampling controller built into the signal processing unit. The sampling frequency is 200 Hz, ensuring that all modal data are strictly aligned on the time axis, forming a synchronous multi-dimensional time-series data stream.

[0024] In the aforementioned scalp health detection method, step S2 involves preprocessing the collected raw data, including region normalization of sebum data, morphological segmentation of dandruff images, non-uniformity correction of infrared thermograms, baseline drift elimination of near-infrared spectra, peak extraction of hair tension signals, and equivalent circuit fitting of impedance spectra. It should be understood that the raw sensor data contains noise, environmental interference, and inherent equipment biases, and requires standardization before use in subsequent analysis. Specifically, for sebum secretion data, due to the natural differences in sebaceous gland density across different scalp regions, such as the forehead, crown, and occipital region, directly comparing absolute capacitance values ​​can lead to misjudgment. Therefore, a region normalization strategy is adopted: the median of multiple measurements of the same region during each combing process is taken as the benchmark, and the remaining measurements are normalized relative to this benchmark to obtain a dimensionless sebum secretion index. For keratin metabolism image data, background suppression is first performed by smoothing the image with Gaussian filtering and subtracting local means to eliminate the effects of uneven illumination. Then, edge enhancement is performed using the Laplacian operator to highlight dandruff boundaries. Finally, binarization is performed, converting the image into a black-and-white binary image with an adaptive threshold, where white pixels represent dandruff. Based on this, morphological opening operations are applied to remove isolated noise points, and connected component analysis is used to calculate the number, average area, and total area density of dandruff particles, outputting quantifiable dandruff indicators. For infrared thermal imaging data, due to slight differences in the response of each pixel in the sensor array, non-uniformity correction is required. The correction method is as follows: when the comb is activated, the response matrix under a uniform temperature field is obtained through a built-in blackbody reference source, establishing pixel-level gain and bias correction coefficients. These correction coefficients are applied in real-time during subsequent measurements to eliminate fixed-mode noise. For near-infrared spectral data, the spectral baseline will slowly drift due to fluctuations in light source intensity and detector drift. Asymmetric least squares is used for baseline correction, setting smoothing parameters and penalty weights, iteratively fitting the baseline curve and subtracting it from the original spectrum to retain the true absorption characteristics. For hair tension signals, which are transient pulses, their peak values ​​need to be extracted as a breaking force indicator. A sliding window peak detection algorithm is used, with a window width of ten milliseconds. When the signal reaches a local maximum within the window and exceeds a preset threshold (e.g., 0.5 millinewtons), the peak value and its timestamp are recorded. For skin barrier impedance spectrum data, a Cole-Cole equivalent circuit model is used for fitting. This model consists of a resistor R0, a constant-phase element CPE, and a resistor R∞ connected in series. The measured impedance spectrum is fitted using a nonlinear least squares method, and the exponent n and time constant τ in the CPE parameters are extracted to characterize the hydration of the stratum corneum and the integrity of the barrier function, respectively.

[0025] In the aforementioned scalp health detection method, step S3 involves inputting the preprocessed feature vectors of each modality into a multimodal fusion analysis model. This model consists of three parts: a feature encoder, a cross-modal attention interaction layer, and a health status decoder. It should be understood that while multimodal data is abundant, semantic gaps and information redundancy exist between different modalities, necessitating effective fusion through a deep learning architecture. The feature encoder performs independent nonlinear mapping on the features of each modality, generating high-dimensional semantic representations. Specifically, sebum secretion index, dandruff quantification index, inflammation activity probability, hair strength decay rate, and barrier function integrity percentage are input into five independent fully connected neural networks. Each network contains three hidden layers, with 128, 64, and 32 nodes per layer, and the activation function is a modified linear unit. The output is a five-dimensional high-dimensional semantic vector with a dimension of 64. The cross-modal attention interaction layer calculates the correlation weights between different modal representations, dynamically weights and fuses key information, and suppresses noise interference. This layer employs a multi-head self-attention mechanism, concatenating the semantic vectors of the five modalities into an input sequence, and calculating attention weights through query, key, and value projection matrices. The attention weight formula is expressed as:

[0026] Where Q, K, and V are the query, key, and value matrices, respectively, and dk is the dimension of the key vector. Through this mechanism, the model can automatically identify which modal combinations are more discriminative in the current scenario. For example, when both the sebum secretion index and dandruff quantification index are high, the model will enhance the interaction weight between the two to determine whether it is seborrheic dermatitis. The health status decoder outputs a multi-dimensional score of scalp health status based on the fused comprehensive representation. The decoder is a fully connected network, with the input being the fusion vector output from the attention layer, and the output being five scores: sebum secretion index (range 0 to 10), dandruff severity level (level 1 to 5), inflammation activity probability (0 to 1), hair strength decay rate (0 to 1), and barrier function integrity percentage (0 to 100). The training process of the model includes collecting a large-scale labeled dataset containing comb-sensing data and corresponding clinical diagnostic results from users of different ages, genders, and regions; adopting an end-to-end supervised learning strategy, using clinical diagnostic labels as supervision signals to optimize model parameters; and in the model deployment stage, using model distillation technology to compress the large training model into a lightweight inference model to adapt to the embedded processor inside the comb.

[0027] In the aforementioned scalp health detection method, step S4 generates a multi-dimensional score of scalp health status based on the comprehensive representation output by the multimodal fusion analysis model. Risk assessment is then performed according to a preset health threshold rule base. When any indicator exceeds the normal range, a corresponding health warning is generated and pushed to the user terminal via a wireless communication module. It should be understood that the quantitative score needs to be translated into health advice and risk warnings that the user can understand. The preset health threshold rule base is stored in a local storage unit and includes the normal range, warning thresholds, and intervention suggestions for each indicator. For example, when the sebum secretion index is greater than seven, it is judged as "excessive sebum secretion," and it is recommended to "reduce high-sugar and high-fat diets and increase shampooing frequency"; when the dandruff severity level reaches level four, it is judged as "severe dandruff," and it is recommended to "use medicated shampoo containing ketoconazole and seek medical attention"; when the inflammation activity probability is greater than 0.8, it is judged as "high inflammation risk," and it is recommended to "avoid scratching and discontinue the use of irritating hair care products." All warning information is transmitted to the user terminal device, such as a smartphone or smartwatch, via Bluetooth 5.0 protocol. Furthermore, the multi-dimensional score output by the health status decoder is used to generate personalized scalp care suggestions, which are derived based on a preset expert knowledge rule base and trend analysis of user historical data. For example, if a user's sebum secretion index shows an upward trend for seven consecutive days, even if it does not reach the warning threshold, the system will still push a prompt such as "Recent sebum secretion has increased; please pay attention to your diet and rest."

[0028] Based on scalp health detection methods, the system structure of the smart comb is constructed, see [link / reference]. Figure 2The system structure includes a comb body, a multimodal biometric sensor array 1, a signal processing unit 4, a local storage unit, a wireless communication module, and a power supply unit 5. The comb body adopts an ergonomic streamlined design, with its comb teeth made of medical-grade silicone-coated metal conductive material. Multiple micro-sensors are embedded internally to ensure natural contact with the scalp and hair during normal combing, enabling non-invasive data acquisition. The multimodal biometric sensor array 1 includes a sebum secretion detection module, a keratin metabolism state detection module, a local inflammation response detection module, a hair physical property detection module, and a skin barrier integrity detection module. The specific structure and parameters of each module have been detailed in the preceding method steps. The signal processing unit 4 is electrically connected to the multimodal biometric sensor array 1 and is used to amplify, filter, convert analog-to-digital signals from each sensor, and synchronize their timing. The signal processing unit 4 uses a low-power embedded microcontroller with a main frequency of 200 MHz and a built-in 12-bit analog-to-digital converter and digital signal processing coprocessor. The local storage unit caches at least thirty days of raw sensor data and intermediate processing results, employing a non-volatile flash memory chip with a storage capacity of no less than eight gigabytes. The wireless communication module supports Bluetooth 5.0 protocol, featuring low-power broadcast and secure pairing capabilities to ensure data transmission reliability and privacy. The power supply unit 5 is a rechargeable lithium polymer battery with a rated voltage of 3.7 volts and a capacity of 300 mAh, charged via a magnetic interface, supporting continuous use for thirty days on a single charge. Furthermore, the comb body structure incorporates an environmental temperature and humidity compensation module 2, composed of a digital temperature and humidity sensor, used to monitor the temperature and humidity of the environment surrounding the comb in real time and input compensation parameters to the signal processing unit 4 to eliminate interference from environmental factors on biometric signals. The smart comb also includes a user identification module 3, which analyzes the mechanical characteristic sequence of the user's combing, including the temporal distribution of force on the comb teeth, force change patterns, and combing trajectory, to construct an individual biomechanical fingerprint, achieving automatic user identification and data isolation, ensuring data privacy and personalized analysis in multi-user home scenarios. The workflow of the user identification module 3 is as follows: at the start of each hair combing session, the mechanical signal sequence of the first five seconds is collected, and time-domain features such as root mean square, kurtosis, and skewness are extracted, along with frequency-domain features such as the main frequency energy ratio. These features are then input into a lightweight support vector machine classifier to match the pre-stored user template. After successful identification, the corresponding historical data and personalized threshold rules are loaded.

[0029] In summary, this invention integrates five biometric sensing modules across five dimensions into a smart comb, enabling comprehensive, seamless, and high-frequency monitoring of scalp health during daily grooming. The synchronous acquisition and cross-modal fusion analysis mechanism of multimodal data effectively overcomes the limitations of single sensors, such as susceptibility to interference and incomplete information, significantly improving the accuracy and robustness of the detection results. The quantitative assessment system constructed in this invention can identify the early risks of common scalp problems such as seborrheic dermatitis, folliculitis, and hair loss, advancing the intervention window to before symptoms appear. Simultaneously, the user identification mechanism based on individual biomechanical fingerprints and the localized data processing architecture ensure user privacy and security. The overall solution integrates professional-grade scalp health detection capabilities into an everyday product, resolving the core contradictions of existing technologies such as the inconvenience of professional equipment, inaccurate subjective judgment, and the limited functionality of smart combs, providing a practical and feasible technical path for personal health management.

[0030] The above embodiments are merely explanations of this application and are not intended to limit it. After reading this specification, those skilled in the art can make modifications to these embodiments without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

Claims

1. A scalp detection method integrating multimodal biometric recognition, characterized in that, include: During the user's daily grooming process using the smart comb, data such as sebum secretion, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical properties data, and skin barrier impedance data are collected simultaneously through a multimodal biometric sensor array. The sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data are preprocessed to obtain sebum secretion index, dandruff quantitative index, inflammation activity probability, hair strength attenuation rate, and barrier function integrity percentage. The sebum secretion index, dandruff quantification index, inflammation activity probability, hair strength attenuation rate, and barrier function integrity percentage are respectively input into the feature encoder to obtain high-dimensional semantic representation vectors for each modality. Cross-modal attention interaction parsing is performed on the high-dimensional semantic representation vectors of each modality to obtain the scalp health status fusion representation vector; Based on the scalp health status fusion representation vector, a multi-dimensional score of scalp health status is output through the health status decoder, and risk is determined according to the preset health threshold rule library. When any indicator exceeds the normal range, corresponding health warning information is generated and pushed to the user terminal through the wireless communication module.

2. The scalp detection method integrating multimodal biometric recognition according to claim 1, characterized in that, The sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data are preprocessed to obtain the sebum secretion index, dandruff quantitative index, inflammation activity probability, hair strength attenuation rate, and barrier function integrity percentage, including: The sebum secretion data are subjected to region normalization to obtain the sebum secretion index; After performing background suppression, edge enhancement, and binarization on the keratin metabolism image data, the number of dandruff particles, average area, and total area density are calculated through morphological segmentation and connected component analysis to obtain the dandruff quantification index. The non-uniformity correction of the local inflammation thermal imaging data is performed, and the change of hemoglobin absorption peak in the near-infrared spectral data is analyzed to obtain the probability of inflammation activity. Peak values ​​are extracted from the tensile signals in the physical property data of the hair, and the elastic modulus and toughness parameters of the hair are inverted by combining the vibration decay curve to obtain the hair strength decay rate. The skin barrier impedance data are fitted with a Cole-Cole equivalent circuit in the frequency range of 10 Hz to 100 kHz to extract the stratum corneum hydration and barrier function parameters, thereby obtaining the percentage of barrier function integrity.

3. The scalp detection method integrating multimodal biometric recognition according to claim 2, characterized in that, The sebum secretion index, dandruff quantification index, inflammation activity probability, hair strength decay rate, and barrier function integrity percentage are respectively input into the feature encoder to obtain high-dimensional semantic representation vectors for each modality, including: The sebum secretion index, dandruff quantification index, inflammation activity probability, hair strength decay rate, and barrier function integrity percentage are respectively input into five independent fully connected neural networks. Each fully connected neural network contains three hidden layers with 128, 64, and 32 nodes respectively. The activation function is a modified linear unit, and the output dimension is a high-dimensional semantic representation vector of 64 as the high-dimensional semantic representation vector of each modality.

4. The scalp detection method integrating multimodal biometric recognition according to claim 3, characterized in that, Cross-modal attention interaction parsing is performed on the high-dimensional semantic representation vectors of each modality to obtain a fusion representation vector of scalp health status, including: After concatenating the high-dimensional semantic representation vectors of each modality into an input sequence, the relevance weights between the query, key, and value matrices are calculated through a multi-head self-attention mechanism. The high-dimensional semantic representation vectors of each modality are dynamically weighted and aggregated based on the correlation weights to obtain the scalp health status fusion representation vector.

5. The scalp detection method integrating multimodal biometric recognition according to claim 4, characterized in that, Based on the scalp health status fusion representation vector, a multi-dimensional score of scalp health status is output through a health status decoder, and risk is determined according to a preset health threshold rule base. When any indicator exceeds the normal range, a corresponding health warning is generated and pushed to the user terminal via a wireless communication module, including: The scalp health status fusion representation vector is input into a fully connected decoding network to output five scores, which include sebum secretion index, dandruff severity level, inflammation activity probability, hair strength decay rate, and barrier function integrity percentage. The five scores are compared with the corresponding thresholds in the preset health threshold rule library. When any score exceeds the normal range, the corresponding intervention suggestion is retrieved and the health warning information is generated. The health warning information is transmitted to the user's terminal device via Bluetooth 5.0 protocol.

6. The scalp detection method integrating multimodal biometric recognition according to claim 1, characterized in that, During daily grooming with the smart comb, data on sebum secretion, keratin metabolism, local inflammation thermal imaging and spectral data, hair physical properties, and skin barrier impedance are simultaneously collected via a multimodal biometric sensor array, including: The sebum secretion data is collected by micro-area capacitive oil sensors distributed on the tips and sidewalls of the comb teeth. The sensing area of ​​each sensor is 0.5 square millimeters to 1.5 square millimeters. The keratin metabolism image data is acquired by an optical imaging unit consisting of a blue light-emitting diode with a center wavelength of 470 nanometers and a monochrome image sensor with more than 5 million pixels. The local inflammation thermal imaging and spectral data were acquired using an infrared thermal imaging sensor with a spatial resolution of 0.1 degrees Celsius and a near-infrared spectral analysis unit that emits a continuous spectrum from 700 nanometers to 1,000 nanometers. The physical properties of the hair are collected by a miniature tension sensor with a range of 0 to 5 millinewtons integrated into the root of the comb teeth and a piezoelectric ceramic vibration unit with an excitation frequency of 1 kHz to 5 kHz. The skin barrier impedance data were acquired by an AC impedance spectroscopy measurement unit that applied a microampere-level AC excitation signal in the frequency range of 10 Hz to 100 kHz. All sensors are driven by the same clock source, and the sampling frequency is uniformly set to 200 Hz to ensure that the data of each mode are strictly aligned on the time axis.

7. The scalp detection method integrating multimodal biometric recognition according to claim 1, characterized in that, Also includes: At the start of each combing session, data on the force distribution, force variation pattern, and combing trajectory of the comb teeth were collected for the first five seconds. After extracting the time-domain and frequency-domain features of the data, they are input into a lightweight support vector machine classifier to match pre-stored user templates. Once the identification is successful, the corresponding historical data and personalized threshold rules are loaded to achieve automatic user identification and data isolation.

8. The scalp detection method integrating multimodal biometric recognition according to claim 1, characterized in that, Also includes: The built-in digital temperature and humidity sensor monitors the temperature and humidity of the environment around the comb in real time. The temperature and humidity are input as compensation parameters to the signal processing unit to correct environmental interference components in the original data of each mode.

9. A smart comb integrating multimodal biometric recognition, characterized in that, include: A multimodal biometric sensor array is used to simultaneously collect sebum secretion data, keratin metabolism image data, local inflammation thermal imaging and spectral data, hair physical property data, and skin barrier impedance data during the user's daily grooming process using a smart comb; The signal processing unit is electrically connected to the multimodal biometric sensor array and is used to preprocess the collected raw data to obtain sebum secretion index, dandruff quantification index, inflammatory activity probability, hair strength attenuation rate and barrier function integrity percentage. A feature encoder is used to map the sebum secretion index, dandruff quantification index, inflammation activity probability, hair strength attenuation rate and barrier function integrity percentage into high-dimensional semantic representation vectors for each modality. A cross-modal attention interaction parsing unit is used to perform cross-modal attention interaction parsing on the high-dimensional semantic representation vectors of each modality to obtain a scalp health status fusion representation vector. A health status decoder is used to output a multi-dimensional score of scalp health status based on the scalp health status fusion representation vector. The early warning information generation module is used to determine the risk of the multi-dimensional score according to the preset health threshold rule library, and generate corresponding health early warning information when any indicator exceeds the normal range. The wireless communication module is used to push the health warning information to the user terminal.

10. The smart comb integrating multimodal biometric recognition according to claim 9, characterized in that, The multimodal biometric sensing array includes: The sebum secretion detection module consists of multiple micro-area capacitive oil sensors with a sensing area of ​​0.5 square millimeters to 1.5 square millimeters, distributed on the top and sidewalls of the comb teeth; The keratin metabolism status detection module includes a blue light-emitting diode with a center wavelength of 470 nanometers and a monochrome image sensor with more than 5 million pixels; The local inflammatory response detection module consists of an infrared thermal imaging sensor with a spatial resolution of 0.1 degrees Celsius and a near-infrared spectral analysis unit with an emission wavelength range of 700 nanometers to 1,000 nanometers. The hair physical property detection module consists of a miniature tensile sensor with a range of zero to five millinewtons and a piezoelectric ceramic vibration unit with an excitation frequency of one kilohertz to five kilohertz. The skin barrier integrity detection module consists of an AC impedance spectroscopy measurement unit that applies a microampere-level AC excitation signal in the frequency range of 10 Hz to 100 kHz. All sensors are driven by the same clock source, and the sampling frequency is uniformly set to 200 Hz.

Images