Narrow-frame telephone watch wearing detection system and method

Abstract

The invention relates to the technical field of computers, discloses a narrow-bezel telephone watch wearing detection system and method, and aims to solve the problems of high power consumption, insufficient accuracy and proneness to interference and misjudgment caused by dependence on a single sensor in the prior art. According to the method, initial contact monitoring is carried out through a capacitance sensing unit, a thermosensitive array and an inertial measurement unit are activated in stages after triggering, and body temperature distribution characteristics and dynamic behavior modes are verified in sequence; when fuzzy judgment is carried out, active sound vibration detection based on a linear resonance actuator is started, and vibration attenuation characteristics are analyzed to complete final judgment. The system comprises a data acquisition module, a state judgment engine and an active detection unit, and ultra-low power consumption operation is realized by adopting a layered activation strategy. According to the invention, multi-modal sensing and a physical model are fused, the detection precision and the anti-interference capability are significantly improved, and the endurance time of equipment is prolonged.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers, and particularly relates to a narrow-frame phone watch wearing detection system and method. BACKGROUND

[0002] With the popularity of smart wearable devices, phone watches, as typical terminals integrating communication, positioning, health monitoring and child safety protection, are widely used in low-age user groups. The effectiveness of the core functions of the phone watch is highly dependent on whether the device is correctly worn: for example, heart rate monitoring, fall detection, electronic fence alarm and remote call scenarios all require the watch to be close to the skin. However, the behavior of child users is highly uncertain, and situations such as frequent removal, loose wearing or misplacement in clothing pockets are common, leading to distorted sensor data, positioning drift and even safety function failure. In addition, the frame of the phone watch is getting narrower and narrower, which makes it more likely to cause inaccurate identification. Existing phone watches generally lack accurate discrimination ability for the real wearing state, and only rely on a single accelerometer or photoplethysmography (PPG) signal for simple threshold judgment, which is easily affected by motion artifacts, environmental light interference or poor skin contact, resulting in a large number of false positives or missed detections.

[0003] Among them, the phone watch wearing detection technology aims to accurately distinguish between "effective wearing", "ineffective wearing" and "not wearing" through multi-source sensor information fusion. The basic principle of this technology is to comprehensively utilize inertial measurement unit (IMU), optical sensor, temperature sensor and contact impedance modal data to construct a dynamic perception model of the relative position relationship between the watch and the human body. In an ideal state, the system should maintain high stability in complex daily activities (such as running, washing hands, sleeping), while taking into account the low power consumption and real-time response requirements.

[0004] The prior art has the following defects in achieving the above-mentioned objectives: first, a single sensor scheme cannot overcome the inherent defects in specific scenarios, for example, the signal-to-noise ratio of the PPG signal decreases sharply under intense exercise, and the accelerometer cannot distinguish between the watch being placed on the table and being stably worn on the wrist; second, multi-sensor fusion strategies mostly use fixed weights or rule engines, lacking the ability to adapt to individual differences (such as skin color, wrist thickness, wearing tightness) and environmental changes (such as temperature, humidity, light); third, existing algorithms usually run on application processors, with high computational overhead, making it difficult to meet the limited battery life requirements of phone watches; finally, in the context of child use, the system also needs to cope with non-standard wearing behaviors (such as reverse wearing, wearing outside the sleeve), and current methods have little recognition ability for such edge cases. The above problems collectively result in low wearing detection accuracy, high power consumption and poor user experience, and there is an urgent need for a high-precision, low-power and highly stable phone watch wearing detection system and method. SUMMARY

[0005] The purpose of this invention is to provide a narrow-bezel phone watch wearing detection system and method, which aims to solve the technical problems of existing wearing detection methods that rely on a single sensor or simple logic combination, resulting in high power consumption, insufficient accuracy, and susceptibility to environmental interference leading to misjudgments.

[0006] To achieve the above objectives, this invention provides a method for detecting the wearing status of a narrow-bezel smartwatch. This method utilizes layered acquisition and progressive analysis of multimodal sensor data, combined with active acoustic-vibration feedback detection, to construct a low-power, high-precision composite model for determining the wearing status. The method first uses an ultra-low-power capacitive sensing unit for continuous monitoring of initial contact. After a contact event is triggered, a thermal array sensor and an inertial measurement unit are activated in stages and on demand to sequentially verify the body temperature distribution characteristics and micro-motion behavior patterns of the contact object. In cases of ambiguity, an active detection mechanism based on a micro-vibrator is further activated. By analyzing the attenuation characteristics of the vibration signal in the contact medium, a final accurate determination of the wearing status is achieved.

[0007] According to one aspect of the present invention, a method for detecting the wearing of a narrow-bezel phone watch is provided, the method specifically comprising: The capacitance sensing unit continuously monitors the capacitance change of the back cover of the phone watch at a preset low-frequency sampling period. When the capacitance change exceeds the preset capacitance trigger threshold, an initial contact event signal is generated. In response to the initial contact event signal, a thermal sensor array set on the back cover of the phone watch case is activated. The thermal sensor array collects real-time temperature data from multiple temperature measurement points covering the contact area, forming a two-dimensional temperature distribution map. Analyze the two-dimensional temperature distribution map, calculate its temperature mean, temperature gradient and spatial entropy of temperature distribution. When the temperature mean falls within the preset human body surface temperature range, and the temperature gradient and spatial entropy conform to the preset human wrist thermodynamic model characteristics, an effective body temperature contact signal is generated. Upon receiving a valid body temperature contact signal, an inertial measurement unit is activated. The inertial measurement unit collects triaxial acceleration data and triaxial angular velocity data within a preset time window, forming a time-series dynamic data sequence. Feature extraction is performed on the time-series dynamic data sequence, and multiple dynamic feature parameters, including motion intensity spectral density, posture stability factor and periodic motion frequency, are calculated. The dynamic feature parameters are then matched with a human wearable activity feature database pre-stored in the memory to generate a dynamic feature matching score. Based on the dynamic feature matching score, a preliminary determination of the wearing status is performed; when the dynamic feature matching score is higher than the first preset determination threshold, the phone watch is determined to be in a wearing state; when the dynamic feature matching score is lower than the second preset determination threshold, the phone watch is determined to be in a non-wearing state; wherein, the first preset determination threshold is greater than the second preset determination threshold. When the dynamic feature matching score is between the first preset judgment threshold and the second preset judgment threshold, the active detection process is initiated. The active detection process specifically includes: driving a linear resonant actuator to generate a preset composite frequency vibration sequence through a control signal; while the linear resonant actuator vibrates, using the accelerometer in the inertial measurement unit to collect the shell vibration response data generated by the vibration; performing a fast Fourier transform on the shell vibration response data to extract its resonant peak frequency, bandwidth, and energy decay time constant, forming a vibration decay feature vector; comparing the vibration decay feature vector with a preset human tissue vibration decay model to calculate an acoustic-vibration coupling confidence level. The final determination of the wearing status is based on the acoustic-vibration coupling confidence level. When the acoustic-vibration coupling confidence level is higher than the preset determination threshold, the phone watch is determined to be in the wearing state; otherwise, it is determined to be in the non-wearing state.

[0008] In one embodiment of the present invention, the thermal sensor array consists of four negative temperature coefficient thermistors arranged in a 2x2 matrix. These thermistors are encapsulated on the inner surface of the back cover of the watch case, and a heat conduction path is formed between them and the outer surface of the back cover through a medium with high thermal conductivity. The thermodynamic characteristics of the human wrist are specifically: the temperature in the central region of the two-dimensional temperature distribution map is higher than the temperature in the edge region, and the temperature gradient decreases radially outwards; simultaneously, the spatial entropy of the temperature distribution is less than a preset uniform heat source entropy threshold.

[0009] Furthermore, feature extraction is performed on the time-series dynamic data sequence, specifically including: vector synthesis of the three-axis acceleration data within the time window to obtain the total acceleration time-series signal; short-time Fourier transform of the total acceleration time-series signal to obtain its energy distribution in the frequency range of 0 to 50 Hz, and using it as the motion intensity spectral density; calculation of the variance of the three-axis angular velocity data within the time window, and using the reciprocal of the variance as the attitude stability factor; and autocorrelation analysis of the total acceleration time-series signal to extract the main period with the strongest energy in the signal, and using it as the periodic motion frequency.

[0010] In one embodiment of the present invention, the linear resonant actuator is a piezoelectric ceramic vibrator. The preset composite frequency vibration sequence includes a fundamental sinusoidal signal with a frequency of 150 Hz and a second harmonic signal with a frequency of 300 Hz superimposed on it, with a duration of 200 milliseconds. The human tissue vibration attenuation model is a database that stores the statistical distribution range of vibration attenuation feature vectors excited by the composite frequency vibration sequence when the phone watch is worn on the wrists of users of different ages and body types. The acoustic-vibration coupling confidence is calculated by performing Mahalanobis distance calculation between the real-time extracted vibration attenuation feature vectors and the statistical distribution range stored in the database. The reciprocal of the Mahalanobis distance is normalized and used as the acoustic-vibration coupling confidence.

[0011] Furthermore, the method also includes a confirmation logic for the transition of the wearing state. When determining whether the smartwatch has transitioned from a wearing state to a non-wearing state, the method requires that it be determined to be in a non-wearing state for three consecutive detection cycles before the state transition can be finally confirmed. This is to avoid misjudgments caused by brief, vigorous movements or momentary sensor interference. When determining the transition from a non-wearing state to a wearing state, the method updates the state immediately after completing a full determination process and confirming the wearing state, to ensure real-time response to wearing events.

[0012] According to another aspect of the present invention, a narrow-bezel phone watch wearing detection system is provided, the system being integrated inside a phone watch, and the system specifically includes: A data acquisition module includes a capacitance sensing unit, a thermistor array, and an inertial measurement unit. All three units are electrically connected to a central processing unit. The capacitance sensing unit monitors changes in the capacitance of the housing's back cover. The thermistor array acquires a two-dimensional temperature distribution map of the contact area. The inertial measurement unit acquires time-series dynamic data sequences and housing vibration response data. A state determination engine is embedded in the firmware of the central processing unit. Its configuration is used to receive and process data from the data acquisition module. The state determination engine specifically includes an initial contact analysis unit, a thermodynamic characteristic verification unit, a dynamic behavior matching unit, and a final decision unit. The initial contact analysis unit is used to analyze the data of the capacitance sensing unit. When the change in capacitance exceeds the capacitance trigger threshold, it generates and sends an initial contact event signal to the thermodynamic feature verification unit. The thermodynamic feature verification unit, upon receiving the initial contact event signal, activates and processes the two-dimensional temperature distribution map collected by the thermal sensor array. After confirming that it conforms to the thermodynamic model characteristics of the human wrist, it generates and sends an effective body temperature contact signal to the dynamic behavior matching unit. The dynamic behavior matching unit, upon receiving a valid body temperature contact signal, activates and processes the time-series dynamic data sequence collected by the inertial measurement unit, calculates the dynamic feature matching score, and outputs a wearing, non-wearing, or pending state signal based on the relationship between the score and the first and second preset judgment thresholds. An active detection unit includes a linear resonant actuator and a drive circuit, which is connected to a central processing unit. When the dynamic behavior matching unit outputs a pending state signal, the central processing unit controls the active detection unit to execute the active detection process, and the active detection unit sends the collected shell vibration response data to the final decision unit. The final adjudication unit is used to receive the shell vibration response data, calculate the acoustic-vibration coupling confidence level, and output the final wearing state or non-wearing state determination result based on the comparison result of the confidence level and the adjudication threshold.

[0013] As one embodiment of the present invention, the system further includes a non-volatile storage module connected to the central processing unit. The non-volatile storage module stores a capacitor trigger threshold, a human body surface temperature range, characteristic parameters of a human wrist thermodynamic model, a database of human wearing activity characteristics, a first preset judgment threshold, a second preset judgment threshold, a human tissue vibration attenuation model, and a decision threshold.

[0014] Furthermore, the state determination engine adopts a layered activation power management strategy. When no initial contact event signal is received, the entire system is in sleep mode, with only the capacitive sensing unit operating at a microamp level current. When entering the thermodynamic feature verification stage, the thermal sensor array is activated, and the system power consumption increases to the milliamp level. When entering the dynamic behavior matching stage, the inertial measurement unit is activated, and the system power consumption increases further. Only when the determination is ambiguous is the active detection unit, which has the highest power consumption, activated instantaneously. Thus, while ensuring detection accuracy, the average operating power consumption of the system is reduced to the maximum extent.

[0015] In summary, this application includes at least one of the following beneficial technical effects: (1): By adopting four sensing modes with completely different physical principles, namely capacitance, thermal array, inertial measurement unit and active acoustic vibration detection, and performing layered and progressive fusion judgment, the accuracy and anti-interference ability of wear detection are greatly improved, and the misjudgment problem caused by single sensor due to environmental changes, object disguise and other factors is effectively overcome.

[0016] (2): By constructing a state-progressive power management model that ranges from ultra-low power sleep monitoring to high-precision active detection, the high-power sensors and processing units are woken up step by step only when necessary, so that the system operates at a standby power level of microamps most of the time, which significantly extends the battery life of the phone watch.

[0017] (3): An active acoustic-vibration coupling detection mechanism based on linear resonant actuator and accelerometer is introduced. By analyzing the attenuation characteristics of vibration in the contact medium, human tissue and non-biological body are distinguished. This method provides a physical criterion that is difficult to forge and has high confidence for wear detection, and fundamentally solves the technical bottleneck of traditional passive sensing methods that cannot accurately determine in static or ambiguous scenarios.

[0018] (4): The judgment logic of this application is based on in-depth analysis of physical models and data characteristics, rather than simple threshold judgment. Its built-in human thermodynamic model, activity feature database and vibration attenuation model have better universality and stability, and can adapt to different users and changing usage scenarios. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention. Detailed Implementation

[0020] This invention provides a narrow-bezel phone watch wearing detection system and method, aiming to solve the problems of high power consumption, insufficient accuracy, and susceptibility to environmental interference leading to misjudgments caused by existing technologies that rely on a single sensor or simple logic combination for wearing status determination. Through hierarchical acquisition and progressive state analysis of multimodal sensor data, combined with an active acoustic vibration feedback detection mechanism, a low-power, high-precision composite wearing status determination model is constructed. The following is in conjunction with the appendix... Figure 1 The paper will provide a detailed description of the specific implementation steps of the method and disclose the necessary system structure that supports the operation of the wear detection system method of this application.

[0021] In order to more clearly disclose the technical solution of this application, the following description is presented in a step-by-step manner.

[0022] The first step, S1, is the narrow-bezel phone watch wearing detection method. It first uses an ultra-low power capacitive sensing unit to continuously monitor the back cover of the phone watch casing.

[0023] S101: Configure and start the capacitance monitoring unit. After the system powers on, the central processing unit initializes and configures the capacitance sensing unit, putting it into continuous monitoring mode. Specifically, this includes: The capacitive sensing unit employs a mutual capacitance sensing scheme, with a low-power capacitive digital-to-digital converter chip at its core. This chip connects to the central processing unit via an I²C interface. The sensing electrode is a copper foil pattern printed on a flexible circuit board, which is attached to the inner surface of the back cover of the phone watch casing.

[0024] The sampling rate register of the capacitor-to-digital converter chip is configured to 1Hz via the I²C bus, and its operating mode is set to low-power single-conversion mode. Simultaneously, its internal data-ready interrupt function is enabled.

[0025] After configuration, a start measurement command is sent to the capacitance-to-digital converter chip. The chip then automatically performs capacitance measurements at a frequency of 1Hz, and after each measurement, an interrupt signal is sent to the central processing unit to read the capacitance value. The average current consumption in this operating mode can be maintained below 10 microamps.

[0026] S102: Acquire Reference Capacitance Value. The system provides a calibration mechanism to acquire and store the reference capacitance value C_ref. Typical triggering times include: automatic execution during factory testing, or manual initiation of the calibration process by the user through the device settings menu when the watch is clearly not being worn. When the calibration process is triggered, the capacitance sensing unit performs one or more measurements, uses the obtained stable capacitance value as the reference capacitance value C_ref, and permanently stores it in the device's non-volatile storage module. This value serves as a static reference for subsequent judgment of contact events.

[0027] S103: Continuous monitoring and capacitance value comparison. When the system is operating in low-power monitoring mode, the capacitance sensing unit periodically samples the current capacitance value C_current at a frequency of 1Hz. After each sampling, the system calculates the absolute difference between the current capacitance value and the reference capacitance value: ΔC = |C_current - C_ref|. This difference ΔC will serve as the key input for determining whether an initial contact event is triggered.

[0028] The steps S101 to S103 described above constitute a complete, low-power background monitoring loop. This system continuously detects whether an object is approaching or touching the watch back cover by sampling at extremely low frequencies and comparing the data with a pre-stored reference value, providing the initial trigger condition for the entire wear detection process.

[0029] S104: Determine and trigger the initial contact event. Compare the calculated capacitance change ΔC with a preset capacitance trigger threshold C_th. The threshold C_th is a relative value determined based on experimental data, typically set to 5% to 10% of the reference capacitance value C_ref. If ΔC ≥ C_th, a valid initial contact event is determined, and the system immediately generates an initial contact event signal. This signal will wake up the subsequent processing unit, marking the official start of the wear detection process. The specific percentage of the threshold can be fine-tuned based on the shell material, structure, and environmental factors to optimize anti-interference capabilities.

[0030] Step S2, namely, in response to the initial contact event signal, activates the thermal sensor array located on the inner surface of the back cover of the phone watch case, specifically: S201: Response triggers and activates the sensor. Upon receiving the initial contact event signal generated in step S104, the central processing unit sends a high-level enable signal to the power management module via its general purpose input / output pins. The power management module then powers the thermal sensor array located on the inner surface of the back cover of the phone watch case, activating it from sleep mode.

[0031] S202: The specific configuration of the thermistor array is as follows: four negative temperature coefficient thermistors are arranged in a 2x2 matrix on a printed circuit board. The spacing between the resistors is designed according to the size of the housing back cover, typically 5 to 15 mm, to achieve effective spatial sampling of the temperature distribution in the contact area. The sensing surface of each thermistor element is tightly bonded to the inner surface of the housing back cover using high thermal conductivity silicone grease. This silicone grease fills the tiny gaps between the element and the housing, thereby establishing a low thermal resistance heat conduction path between the sensing surface of the thermistor and the outer surface of the housing back cover. This design ensures that temperature changes of externally contacted objects can be quickly and accurately transmitted to the sensor.

[0032] The steps S201 and S202 above together complete the hardware preparation for the second layer of detection. S201 is a logical trigger, while S202 specifies the specific structure and key installation process of the physical sensor for this layer of detection. The latter is the material basis for ensuring the accuracy of temperature measurement and response speed.

[0033] S203: Synchronous temperature data acquisition. After the thermistor array is activated, its built-in signal conditioning and analog-to-digital conversion circuits begin to work. The array synchronously acquires real-time analog temperature signals from four temperature measurement points at a sampling frequency of 20Hz (20 times per second) and converts them into digital temperature values ​​T_i (where i=1,2,3,4). This sampling frequency can capture sufficient time-series information on temperature changes in dynamic contact scenarios.

[0034] S204: Organizing Temperature Distribution Data. The system organizes the four temperature values ​​[T1, T2, T3, T4] acquired at each sampling time according to their corresponding physical locations (2×2 matrix), forming a discrete two-dimensional temperature distribution map. This data map is the direct input for subsequent thermodynamic feature analysis, and its essence is a temperature dataset containing spatial location information.

[0035] The above steps S201 to S204 are the second-level verification after capacitor triggering. This stage activates a thermal sensor array with a specific spatial layout and efficient thermal path to quickly acquire the raw temperature field data of the contact area and structure it into a two-dimensional distribution map that can be analyzed.

[0036] After completing the synchronous acquisition and spatial distribution mapping of temperature data as described in step S2, the system obtains a discrete but structured two-dimensional temperature distribution map. This map carries the original temperature field information of the contact area, but it is not sufficient on its own to directly prove that the contact object conforms to human biological characteristics. Therefore, step S3—thermodynamic characteristic analysis and effective contact verification—is performed. This step aims to determine whether the previous contact event originated from human tissue with vital body temperature characteristics by calculating a set of carefully designed statistical quantities and spatial characteristic parameters and comparing them with a preset thermodynamic model of the human wrist. The specific implementation method of this analysis and verification process will be described in detail below.

[0037] Step S3: Thermodynamic Characteristic Analysis and Effective Contact Verification. Based on the two-dimensional temperature distribution map generated in Step S2, the system performs thermodynamic characteristic analysis to verify whether the contact object conforms to human biometrics. This analysis is achieved by calculating and evaluating a set of key characteristic parameters, and the specific steps are as follows: S301: Calculate the mean temperature distribution. First, calculate the mean temperature T_avg of the two-dimensional temperature distribution map. For a 2×2 matrix consisting of four temperature measurement points, the mean temperature is its arithmetic mean: T_avg=(T1+T2+T3+T4) / 4, which reflects the overall temperature level of the contact area.

[0038] S302: Calculate the spatial temperature gradient and verify the radial trend. To quantify the temperature change from the center to the edge and verify its radial decrease pattern, calculate the gradient characteristics according to the following steps: Define center and edge: The average temperature of two temperature measurement points on the diagonal of the 2×2 array is approximated as the center temperature T_center (e.g., (T1+T4) / 2 or (T2+T3) / 2), and the average temperature of the other pair of points on the diagonal is approximated as the edge temperature T_edge.

[0039] Calculate the radial temperature difference: ΔT_radial = T_center - T_edge.

[0040] Calculate the normalized gradient: Divide the radial temperature difference by the effective radial distance d_eff of the sensor array (i.e., the sensor spacing value set in step S202, for example, 10 mm) to obtain the normalized temperature gradient value G: G=ΔT_radial / d_eff.

[0041] Verify the trend: At the same time, check whether ΔT_radial>0 holds true, and use this as the criterion for "center temperature is higher than edge temperature", that is, radial decreasing trend.

[0042] S303: Calculate the spatial entropy of the temperature distribution. Further, to assess the uniformity (or concentration) of the temperature distribution, calculate the spatial entropy H of the temperature distribution. Treat the four temperature values ​​as a discrete system, with the proportion of temperature values ​​at each point forming a probability distribution. Calculate using the Shannon entropy formula: H = -Σ(p_i*log2(p_i)), where i = 1 to 4, p_i = T_i / ΣT_i (normalizing the temperature value into a probability).

[0043] The smaller the entropy value H, the more concentrated the temperature distribution (e.g., heat is concentrated in the center); the larger the entropy value, the more uniform the distribution.

[0044] S304: Perform feature matching and generate valid signals. Compare the three feature parameters (T_avg, ΔT_radial, H) calculated in steps S301 to S303 with the preset human wrist thermodynamic model judgment conditions: Temperature range condition: The average temperature T_avg must fall within the preset human body surface temperature range, which is typically 32℃ to 37℃.

[0045] Gradient trend condition: The radial temperature difference ΔT_radial (calculated in step S302) must be greater than zero. This directly and uniquely verifies the radially decreasing spatial distribution pattern of "temperature in the central region being higher than that in the peripheral region".

[0046] Spatial homogeneity condition: The spatial entropy H must be less than a preset uniform heat source entropy threshold, typically 0.8. This threshold is used to distinguish between human skin (with relatively concentrated heat distribution) and objects with uniform ambient temperature (such as metals and plastics).

[0047] The system determines that the object being contacted possesses the thermodynamic characteristics of a human wrist, and then generates a valid body temperature contact signal, provided that all three conditions are met simultaneously. This signal indicates that preliminary biometric verification has been passed, triggering the next stage of detection.

[0048] After analyzing the thermodynamic characteristics and generating an effective body temperature contact signal in step S3, the system has preliminarily confirmed that the contact object possesses the biothermal characteristics of human skin. However, static temperature distribution alone cannot distinguish whether the watch is correctly worn on an active wrist or merely statically attached to the skin or other limbs (e.g., held in the hand or placed on the arm but not securely). To further verify the authenticity and effectiveness of the wearing status, the system needs to incorporate analysis of typical dynamic wrist behaviors of the user.

[0049] Based on this, the present invention includes step S4—inertial data acquisition and dynamic behavior analysis. This step aims to activate the inertial measurement unit to collect and analyze the user's micro-movements or activity patterns during the initial suspected wearing phase, thereby determining whether the current contact is accompanied by motion characteristics consistent with human wearing characteristics.

[0050] Step S4: After receiving a valid body temperature contact signal, the system immediately enters the dynamic behavior feature acquisition stage, the core of which is to acquire sensor data reflecting the user's limb micro-movements during the initial wearing phase. This process follows these steps: S401: In response to the aforementioned effective body temperature contact signal, the central processing unit immediately switches the inertial measurement unit from low-power sleep mode to working mode.

[0051] S402: The system sets a preset data acquisition time window (e.g., 2 seconds) and aligns the start time of this window with the generation time of the effective body temperature contact signal. Subsequently, the system controls the inertial measurement unit to begin continuous sampling.

[0052] S403: Within the stated time window, the inertial measurement unit synchronously acquires triaxial acceleration data output from the triaxial accelerometer and triaxial angular velocity data output from the triaxial gyroscope. These multi-axis data, arranged in chronological order, are combined to form a time-series dynamic data sequence for subsequent analysis.

[0053] Step S5 then proceeds: After obtaining the time-series dynamic data sequence, the system enters the dynamic feature extraction stage. This stage aims to quantify key behavioral features from the raw inertial data and assess the likelihood of wearing the device by comparing it with known models. Specifically, it is executed as follows: S501: Acceleration Vector Synthesis and Signal Generation. Vector synthesis is performed on the triaxial acceleration data (Ax, Ay, Az) within a time window. Specifically, at each sampling time t, the magnitude of the synthesized acceleration is calculated according to the formula A_total(t)=sqrt(Ax(t)²+Ay(t)²+Az(t)²). The magnitudes at all sampling times are arranged in chronological order to obtain a total acceleration time-series signal representing the overall motion intensity.

[0054] S502: Perform a short-time Fourier transform on the total acceleration time-series signal obtained in step S501 (e.g., using a Hanning window with a window length of 256 points and an overlap rate of 50%) to obtain its energy distribution spectrum (i.e., power spectral density) in the frequency range of 0 to 50 Hz. Calculate the integral value of the energy spectrum within this frequency band (i.e., sum the energy values ​​at all frequency points within this band) and define it as the motion intensity spectral density. This parameter reflects the overall intensity of the user's activity within this time window.

[0055] S503: Calculate the sample variance of the three-axis angular velocity data (Wx, Wy, Wz) within a preset time window. Calculate the arithmetic mean of these three variances, and define the reciprocal of this mean as the attitude stability factor. A larger factor value indicates smaller fluctuations in the wrist's attitude angular velocity during the measurement period, meaning greater attitude stability. This characteristic is a typical indicator of the device being stably worn on the wrist as part of the body, compared to its placement on a table or being shaken by hand.

[0056] The steps S501 to S503 above complete the feature quantification of motion intensity and posture stability. Among them, the motion intensity spectral density measures the amount of activity in the frequency domain, while the posture stability factor characterizes the smoothness of the movement in the time domain. The combination of the two can effectively distinguish between scenarios such as active wearing and scenarios where the device is moved at will.

[0057] S504: Calculation of Periodic Motion Frequency. Autocorrelation analysis is performed on the total acceleration time-series signal obtained in step S501. Specifically, the normalized autocorrelation function can be calculated using the Fast Fourier Transform (FFT) method, a common technique in this field, to improve computational efficiency and facilitate peak comparison. Subsequently, after excluding zero-delay points, the first significant peak of the autocorrelation function is found within a reasonable physiological motion cycle delay range (e.g., corresponding to frequencies of 0.5Hz to 3Hz). This peak is identified by using local maximum detection or the first peak whose amplitude exceeds a preset threshold. Based on the signal sampling frequency, the number of delayed sampling points corresponding to this peak is converted into an actual delay time in seconds. The reciprocal of this delay time is calculated to obtain the periodic motion frequency in Hertz (Hz). This parameter is a key feature for identifying regular whole-body movements such as walking and running.

[0058] S505: Combine the three parameters extracted in steps S502, S503, and S504 respectively—motion intensity spectral density, attitude stability factor, and periodic motion frequency—in a preset order to construct a three-dimensional feature vector.

[0059] Subsequently, the feature vector is input into a human wearable activity feature database pre-stored in a non-volatile storage module, ready for similarity matching. This database stores statistical model data of wearable state feature vectors learned from a large number of samples in various typical scenarios (such as walking, sitting, running, and sleeping). The statistical model data includes at least typical feature vectors (mean vectors) of various states and weight information (such as variance or covariance matrices) reflecting the dispersion of data in each dimension, providing a basis for subsequent quantitative similarity calculations.

[0060] S506: Dynamic Feature Matching Score Calculation. A weighted Euclidean distance algorithm is used to calculate the difference between the current feature vector and the reference feature vectors (mean vectors) of various states in the database (e.g., typical wearing, typical non-wearing). Specifically, the weighting coefficients are determined based on the variance dynamics of each dimension of the features in the database (e.g., taking the inverse of the variance of each dimension and normalizing it) to balance the impact of different feature dimensions and dispersion on the total distance. The calculated weighted Euclidean distance is denoted as D. Subsequently, the distance D is mapped to a normalized dynamic feature matching score using the formula score = 1 / (1+D). This score directly characterizes the similarity between the current dynamic feature and the target state (e.g., typical wearing behavior); the higher the score, the higher the similarity.

[0061] Step S5 ultimately generates a quantified dynamic feature matching score. Based on this, the system proceeds to step S6, where a preliminary determination and decision regarding the wearing status will be made based on this score.

[0062] Step S6: Preliminary Judgment and Active Detection Trigger. Based on the dynamic feature matching score calculated in Step S5, the system performs a preliminary judgment on the wearing status and triggers a high-confidence active detection process when the judgment is ambiguous. This process is executed according to the following steps: S601: Preliminary state determination based on threshold.

[0063] The system compares the dynamic feature matching score with two preset judgment thresholds: If the score is higher than the first preset judgment threshold (e.g., 0.85), it is directly determined that the phone watch is being worn.

[0064] If the score is lower than the second preset judgment threshold (e.g., 0.4), it is directly judged as not being worn.

[0065] If the score falls between the first and second preset judgment thresholds (i.e., falls into the fuzzy range), the judgment result is pending, and the subsequent active detection process is triggered.

[0066] S602: When the determination result is "pending", the central processing unit sends a control command to the drive circuit of the active detection unit. The drive circuit generates a preset composite frequency vibration sequence according to the command. This sequence is used to drive a linear resonant actuator (e.g., a piezoelectric ceramic vibrator) to generate mechanical vibration. The vibration sequence is an analog electrical signal with a duration of 200 milliseconds. Its waveform consists of a fundamental frequency sine wave with a frequency of 150 Hz superimposed with a second harmonic sine wave with an amplitude of 30% of the fundamental frequency and a frequency of 300 Hz.

[0067] S603: The drive circuit amplifies the power of the composite frequency electrical signal generated in step S602, and then drives the linear resonant actuator to work. The actuator converts the electrical signal into mechanical vibration of the same frequency, thereby exciting the casing of the phone watch to produce forced vibration.

[0068] S604: Simultaneously with the start of vibration of the linear resonant actuator, the triaxial accelerometer in the system control inertial measurement unit (IMU) operates synchronously at a sampling rate of no less than 1 kHz to acquire acceleration data transmitted from the shell vibration. The acquisition process needs to cover the entire excitation phase (200 milliseconds) and the free decay phase after vibration stops, together forming a complete vibration timing signal.

[0069] S605: Perform a Fast Fourier Transform on the vibration time-series signal acquired in step S604 to obtain its spectrum, and then perform the following feature extraction: In the frequency spectrum, find the peak point with the highest amplitude and record its corresponding frequency as the resonant peak frequency.

[0070] Using the aforementioned peak frequency as the center, calculate the difference between the two frequency points corresponding to the point where the resonant peak drops to half the peak value (-3dB), and use this half-width value as the bandwidth.

[0071] Calculate the energy decay time constant: Extract the time-domain signal during the free decay phase after vibration excitation ends. Calculate the envelope of this signal segment (e.g., through Hilbert transform or by low-pass filtering after taking the absolute value). Define the envelope amplitude at the decay start time (t=0) as A0. Find the time t_10% corresponding to when the envelope decays to 0.1*A0. Define t_10% as the energy decay time constant.

[0072] S606: Combine the three characteristic parameters extracted in step S605—resonant peak frequency, bandwidth, and energy decay time constant—into a three-dimensional vibration decay feature vector. This vector comprehensively characterizes the response and decay characteristics of the current contact medium to a specific vibration excitation.

[0073] Thus far, through the preliminary judgment and active detection process in step S6, the system has completed the following tasks: First, based on the dynamic behavior matching score, a clear judgment of "wearing" or "not wearing" is made, or a "pending" state that needs further verification is identified; Second, for the "pending" state, the system performs a complete active acoustic and vibration detection, including generating specific vibration excitation, synchronously acquiring response signals, and extracting key vibration attenuation feature vectors from them.

[0074] This feature vector comprehensively reflects the physical interaction between the shell and the contact medium under vibration excitation, but it cannot be directly converted into a conclusion about the wearing state. In order to transform this physical feature into a final state determination, the system needs to perform a final quantification comparison and decision.

[0075] Therefore, in step S7: based on the vibration attenuation feature vector output in step S6, the system enters the final decision-making stage. This stage involves quantitatively comparing this vector with a pre-built human tissue physical model to calculate the criteria for high-confidence judgment, thereby completing the final confirmation of the wearing status. The specific steps are as follows: S701: The final decision unit receives the vibration attenuation feature vector constructed in step S606, denoted as x. Simultaneously, it retrieves a pre-set human tissue vibration attenuation model from the system's non-volatile storage module. This model is a multi-category statistical database based on massive amounts of data collected from active vibration detection of different user groups (covering different age groups such as children, adolescents, adults, and the elderly, as well as different body types such as thin, standard, and robust) while they are wearing the device.

[0076] S702: The system first selects the most matching reference group model from the database based on the current user's basic information (such as preset age group and body type) or through preliminary screening. This model is characterized by the mean vector μ and covariance matrix Σ of the sample population.

[0077] Subsequently, the Mahalanobis distance between the real-time feature vector x and the population model is calculated. The calculation formula is as follows: Where x represents the vibration attenuation feature vector obtained in real time through the active detection process. This vector is a three-dimensional data point, which is composed of three feature parameters extracted in step S605: resonant peak frequency, bandwidth, and energy attenuation time constant, and is used to quantitatively characterize the physical response characteristics of the current phone watch casing and contact medium under specific vibration excitation.

[0078] μ represents the reference mean vector selected from the pre-stored "Human Tissue Vibration Attenuation Model" database, matching the current user group. This vector statistically represents the average value of the vibration attenuation characteristic vector of the target user group under correct wearing conditions, and represents the central position of the typical characteristics of this group.

[0079] Σ represents the covariance matrix corresponding to the mean vector μ. This matrix describes the dispersion (variance) of each dimension of the vibration attenuation feature vector of the target user group under wearing conditions, as well as the linear correlation between different dimensions, which together define the distribution pattern of the feature data around the center μ.

[0080] This represents the inverse of the covariance matrix Σ. In calculating Mahalanobis distance, its mathematical role is to decorrelate and standardize the dimensions of the eigenvectors, thereby eliminating the influence of differences in the dimensions of different features and their correlations on the distance metric. This allows the calculated distance to more accurately reflect the degree of deviation of real-time data from the reference statistical distribution.

[0081] T represents the transpose operator for vectors. It is used to transpose column vectors. Convert it to a row vector to satisfy the rules of matrix multiplication.

[0082] This represents the calculated Mahalanobis distance. This distance is a dimensionless scalar value. The smaller the value, the closer the real-time feature vector x is statistically to the center of the target reference distribution (completely defined by μ and Σ), meaning a higher probability that x belongs to this wearing state reference group. This distance is subsequently calculated using a mapping function (…). The direct input for calculating the acoustic-vibration coupling confidence C.

[0083] S703: Calculate the Mahalanobis distance This is mapped to an intuitive and normalized confidence score. The acoustic-vibration coupling confidence score C is calculated using the following formula: This formula uses Mahalanobis distance (Values ​​range from 0 to positive infinity) are mapped to confidence level C (values ​​range from 0 to 1). As C approaches 0, it approaches 1, indicating a very high confidence level. When C increases, C decreases, indicating a decrease in confidence level.

[0084] S704: Compare the calculated acoustic-vibration coupling confidence C with a preset decision threshold (typically 0.7): If C is higher than the decision threshold, the phone watch is finally determined to be in the wearing state.

[0085] If C is lower than or equal to the adjudication threshold, it will be ultimately determined to be in a non-wearing state.

[0086] Step S7 described above constitutes the final decision-making loop of the entire wearing detection process. It utilizes the difficult-to-forge physical features (vibration attenuation vector) actively detected in step S6, and achieves high-precision state determination through rigorous statistical model (Mahathano distance) comparison and normalized confidence calculation. This step fully leverages the fundamental differences in vibration energy attenuation characteristics between human soft tissue and rigid objects, fundamentally solving the problem of insufficient judgment capability of passive sensing methods in static, ambiguous, or interference scenarios, ensuring that the final wearing state result output by the entire system has high reliability and anti-interference capabilities.

[0087] Furthermore, the detection method disclosed in this application also includes confirmation logic for the transition of the wearing state. When the system determines that the phone watch has transitioned from a wearing state to a non-wearing state, it requires outputting a non-wearing state determination result for three consecutive detection cycles before the state transition can be finally confirmed. The interval between each detection cycle is 5 seconds. This design can effectively avoid misjudgments caused by the user briefly removing the watch to check information, momentary sensor failure due to vigorous movement, or accidental obstruction by external objects. When determining the transition from a non-wearing state to a wearing state, after completing a complete four-stage determination process (capacitive trigger → thermodynamic verification → dynamic matching → active detection when necessary) and confirming the wearing state, the system immediately updates the internal status flag to ensure real-time response to the wearing event and meet the user's need for immediate function activation.

[0088] On the other hand, the implementation of the method described in this application relies on a highly integrated narrow-bezel phone watch wearing detection system. This system is integrated inside the phone watch and mainly includes a data acquisition module, a status determination engine, an active detection unit, and a non-volatile storage module.

[0089] The data acquisition module includes a capacitance sensing unit, a thermistor array, and an inertial measurement unit. The capacitance sensing unit employs a mutual capacitance structure, with its sensing electrodes embedded inside the back cover of the housing. It detects changes in capacitance at the microfarad level using a dedicated low-power analog front-end chip. The thermistor array consists of four negative temperature coefficient thermistors packaged on a flexible printed circuit board, tightly attached to the inner surface of the back cover, and secured with thermally conductive adhesive. The inertial measurement unit is a commercially available MEMS device, integrating a three-axis accelerometer and a three-axis gyroscope. It supports various range and bandwidth configurations, and its data output interface is an I2C bus, connecting to the central processing unit.

[0090] The state determination engine is embedded in the firmware of the central processing unit. It is implemented as an embedded program code, logically divided into four functional units: the initial contact analysis unit processes capacitance data and generates trigger signals; the thermodynamic feature verification unit performs temperature distribution analysis and model matching; the dynamic behavior matching unit extracts time-series data features and compares them with the database; and the final decision unit is dedicated to handling vibration data analysis and confidence calculation during the active detection phase. These units are scheduled through a state machine mechanism to ensure that subsequent modules are activated only when preconditions are met.

[0091] The active detection unit consists of a linear resonant actuator and its drive circuit. The linear resonant actuator is a monolithic piezoelectric ceramic element with a factory-calibrated resonant frequency of 150 Hz, and is installed close to the center of the housing to maximize vibration transmission efficiency. The drive circuit includes an H-bridge power amplifier and a waveform generator, capable of accurately reproducing a preset composite frequency vibration sequence.

[0092] The non-volatile storage module uses flash memory chips to store all preset thresholds, model parameters, and feature databases. These include capacitive trigger thresholds, upper and lower limits of human body surface temperature ranges, feature parameters of the human wrist thermodynamic model (such as entropy thresholds and gradient direction constraints), a database of human wearing activity features (stored in the form of K-nearest neighbors or Gaussian mixture models), first and second preset judgment thresholds, a human tissue vibration attenuation model (containing multiple sets of mean vectors and covariance matrices), and decision thresholds. All parameters can be remotely updated via firmware upgrades to adapt to differences in different product batches or user groups.

[0093] The entire system employs a layered activation power management strategy. In the contactless state, the system is in deep sleep mode, with only the capacitive sensing unit continuously operating at a current of approximately five microamps. Once an initial contact event is triggered, the thermal sensor array is awakened, and the system power consumption rises to approximately two hundred microamps. If thermodynamic verification is passed, the inertial measurement unit is activated, increasing power consumption to approximately one and a half milliamps. Only when the dynamic matching result is ambiguous is the active detection unit momentarily activated, at which point the peak power consumption can reach ten milliamps, but the duration is only two hundred milliseconds. Through this on-demand, step-by-step wake-up mechanism, the system's average power consumption is controlled below fifty microamps over twenty-four hours, significantly extending the battery life of the smartwatch.

[0094] In summary, this invention constructs a wear detection scheme that combines high precision, strong stability, and ultra-low power consumption by hierarchically fusing four heterogeneous sensing modes: capacitance, thermal array, inertial measurement, and active acoustic vibration. Its judgment logic is based on deep modeling of human physiological characteristics and physical interaction mechanisms, rather than simple threshold comparison, thus effectively overcoming the limitations of existing technologies in complex usage scenarios.

[0095] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.

[0096] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A narrow bezel phone watch wearing detection method, characterized in that, The application comprises the following steps: Continuously monitoring the capacitance value change of the phone watch shell back cover by the capacitance sensing unit at a preset low frequency sampling period, and generating an initial contact event signal when the capacitance value change exceeds a preset capacitance trigger threshold; In response to the initial contact event signal, activating the thermal sensor array arranged on the phone watch shell back cover to collect real-time temperature data of the back cover covering contact area, forming a two-dimensional temperature distribution map; Analyzing the two-dimensional temperature distribution map, calculating its temperature mean, temperature gradient and spatial entropy of temperature distribution to generate an effective body temperature contact signal; After receiving the effective body temperature contact signal, activating the inertial measurement unit, which collects three-axis acceleration data and three-axis angular velocity data within a preset time window to form a time series of dynamic data; Extracting features from the time series of dynamic data, calculating dynamic feature parameters including motion intensity spectral density, posture stability factor and periodic motion frequency, and matching the dynamic feature parameters with the human body wearing activity feature database pre-stored in the memory to generate a dynamic feature matching degree score; According to the dynamic feature matching degree score, performing preliminary judgment of the wearing state; when the dynamic feature matching degree score is higher than a first preset judgment threshold, it is determined that the phone watch is in a wearing state; when the dynamic feature matching degree score is lower than a second preset judgment threshold, it is determined that the phone watch is in a non-wearing state; wherein the first preset judgment threshold is greater than the second preset judgment threshold; When the dynamic feature matching degree score is between the first preset judgment threshold and the second preset judgment threshold, start the active detection process and calculate the acoustic vibration coupling confidence; Based on the acoustic vibration coupling confidence, make a final decision on the wearing state; when the acoustic vibration coupling confidence is higher than a preset decision threshold, it is determined that the phone watch is in a wearing state, otherwise it is determined to be in a non-wearing state. 2.The narrow-bezel phone watch wearing detection method of claim 1, wherein The thermal sensor array is composed of negative temperature coefficient thermistors, which are packaged on the inner surface of the phone watch shell back cover and form a heat conduction path with the outer surface of the shell back cover through a high thermal conductivity medium; The human wrist thermodynamic model features are that the temperature of the central region of the two-dimensional temperature distribution map is higher than that of the edge region, and the temperature gradient decreases outward along the radial direction, and the spatial entropy of the temperature distribution is less than a preset uniform heat source entropy threshold. 3.The narrow-bezel phone watch wearing detection method of claim 1, wherein, The feature extraction of the time series of dynamic data specifically includes: vector synthesis of the three-axis acceleration data within the time window to obtain the total acceleration time series signal; Performing short-time Fourier transform on the total acceleration time series signal to obtain its energy distribution in the frequency range of 0 to 50 hz, and taking it as the motion intensity spectral density; Calculating the variance of the three-axis angular velocity data within the time window, and taking the reciprocal of the variance as the posture stability factor; Performing autocorrelation analysis on the total acceleration time series signal to extract the main period with the strongest energy in the signal, and taking it as the periodic motion frequency. 4.The narrow-bezel phone watch wearing detection method of claim 1, wherein, The active detection process specifically includes: driving the linear resonant actuator to generate a preset composite frequency vibration sequence through a control signal; While the linear resonant actuator is vibrating, the accelerometer in the inertial measurement unit is used to collect the shell vibration response data caused by the vibration; The shell vibration response data is subjected to fast Fourier transform to extract the resonance peak frequency, frequency band width and energy decay time constant, thereby forming a vibration decay characteristic vector; The vibration decay characteristic vector is compared with a preset human tissue vibration decay model to calculate an acoustic vibration coupling confidence. 5.The narrow-bezel phone watch wearing detection method of claim 1, wherein, The confirmation logic of the wearing state transition is also included, when determining that the phone watch is in the wearing state to the non-wearing state, it is required to determine the non-wearing state in the continuous three detection periods, and the state transition can be finally confirmed only when the non-wearing state is determined, when determining that the non-wearing state is changed to the wearing state, the state is updated immediately after completing a complete determination process and confirming the wearing state. 6.The narrow-bezel phone watch wearing detection method of claim 1, wherein, The capacitance trigger threshold is between 5% and 10% of the reference capacitance value, and the reference capacitance value is a reference value pre-calibrated when the phone watch is in the non-contact state and stored in the non-volatile storage module.

7. A narrow bezel phone watch wearing detection system, characterized in that, It comprises: The data acquisition module includes a capacitance sensing unit, a thermal sensor array and an inertial measurement unit, and the capacitance sensing unit, the thermal sensor array and the inertial measurement unit are electrically connected with a central processing unit; the capacitance sensing unit is used to monitor the capacitance value change of the shell back cover; the thermal sensor array is used to collect the two-dimensional temperature distribution map of the contact area; the inertial measurement unit is used to collect the time sequence dynamic data sequence and the shell vibration response data; The state determination engine is solidified in the firmware of the central processing unit, and is configured to receive and process data from the data acquisition module; the state determination engine specifically includes an initial contact analysis unit, a thermodynamic characteristic verification unit, a dynamic behavior matching unit and a final decision unit; The initial contact analysis unit is used to analyze the data of the capacitance sensing unit, and generates and sends an initial contact event signal to the thermodynamic characteristic verification unit when the capacitance value change exceeds the capacitance trigger threshold; The thermodynamic characteristic verification unit activates and processes the two-dimensional temperature distribution map collected by the thermal sensor array after receiving the initial contact event signal, and generates and sends an effective body temperature contact signal to the dynamic behavior matching unit after confirming that it conforms to the human wrist thermodynamic model characteristics; The dynamic behavior matching unit activates and processes the time sequence dynamic data sequence collected by the inertial measurement unit after receiving the effective body temperature contact signal, calculates a dynamic characteristic matching degree score, and outputs a wearing, non-wearing or pending state signal according to the relationship between the score and the first and second preset determination thresholds; The active detection unit includes a linear resonant actuator and a driving circuit, and the driving circuit is connected with the central processing unit; when the dynamic behavior matching unit outputs the pending state signal, the central processing unit controls the active detection unit to execute the active detection process, and the active detection unit sends the collected shell vibration response data to the final decision unit; The final decision unit is used to receive the shell vibration response data, calculate the acoustic vibration coupling confidence, and output the final wearing state or non-wearing state determination result based on the comparison result of the confidence and the decision threshold.

8. The narrow bezel watch phone wearing detection system according to claim 7, characterized in that, The system further comprises a non-volatile storage module connected with the central processing unit, and the non-volatile storage module internally stores the capacitance trigger threshold, the human body surface temperature interval, the human body wrist thermodynamic model characteristic parameter, the human body wearing activity characteristic database, the first preset determination threshold, the second preset determination threshold, the human body tissue vibration attenuation model and the arbitration threshold.

Images