A precise method and system for relieving menstrual cramps based on red and near-infrared light.

Abstract

This invention discloses a method and system for precise relief of dysmenorrhea based on red and near-infrared light, involving optoelectronic medical and physical therapy technologies. The method includes the following steps: acquiring the superficial reflected light intensity characteristics and deep mixed light intensity characteristics of the target area on the human abdomen at the current sampling time; spatially projecting the superficial reflected light intensity characteristics based on the current spatial mapping weights to obtain initial simulated noise characteristics, and calculating the approximation residual between the deep mixed light intensity characteristics and the initial simulated noise characteristics. This invention uses the superficial reflected light intensity as a reference benchmark for macroscopic somatic motion interference and introduces a dynamic optimization gradient based on the approximation residual to construct an adaptive correction closed loop for the spatial mapping weights. This mechanism can accurately extract pure smooth muscle clonus envelope features even under strong interference environments such as patient rolling and curling up, thus effectively solving the problem of motion artifacts masking weak signals in the deep target area.

Description

Technical Field

[0001] This invention relates to optoelectronic medical and physical therapy technologies, specifically a method and system for precisely relieving dysmenorrhea based on red and near-infrared light. Background Technology

[0002] Primary dysmenorrhea is a common gynecological pathological condition in women of reproductive age. Its core mechanism lies in the local tissue ischemia and hypoxia caused by abnormal tetanic contractions of the uterine smooth muscle. In recent years, photobiological modulation technology has been gradually applied to physical interventions in this field due to its non-invasive and non-drug-based advantages. It utilizes optical media of specific wavelengths to penetrate the body surface tissue in order to promote local microvascular dilation and relieve muscle spasms.

[0003] Most existing phototherapy protocols follow the operating logic of continuous irradiation with constant power or pulse output at a fixed preset frequency. In order to improve the targeted nature of the intervention, some advanced devices have begun to introduce optical detection methods to capture physiological feedback in the abdomen, attempting to adjust the light parameters based on these feedback fluctuations.

[0004] Photon energy can only maximize its penetration depth and biological efficacy when it is applied within specific physiological intervals when muscles are not under tension. However, dysmenorrhea attacks are often accompanied by intense discomfort, and patients inevitably exhibit significant physical movements such as rolling, curling up, or rapid breathing under severe pain. This dramatic macroscopic physical displacement directly leads to relative sliding between the detection probe and the epidermal interface, as well as transient distortions within local tissues, resulting in strong low-frequency motion artifacts. Because the signal fluctuations caused by this interference highly overlap with the slow contraction rhythm of the target muscle itself in the frequency band, conventional signal filtering methods, when attempting to remove this strong interference, often erase or severely distort the extremely weak real physiological rhythms deep within the muscle. This results in the device ultimately being unable to accurately determine the true cyclical evolution of the muscle, not only missing the optimal time for energy intervention but also potentially blindly applying high doses of light when the muscle is under extreme tension, causing tissue heat accumulation and exacerbating spasms and discomfort. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for precise relief of menstrual cramps based on red and near-infrared light.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] A precise method for relieving menstrual cramps based on red and near-infrared light includes the following steps:

[0008] Obtain the shallow reflected light intensity characteristics and deep mixed light intensity characteristics of the human abdominal target area at the current sampling time;

[0009] Based on the current spatial mapping weights, the shallow reflected light intensity characteristics are spatially projected to obtain the initial simulated noise characteristics, and the approximation residual between the deep mixed light intensity characteristics and the initial simulated noise characteristics is calculated.

[0010] The optimization gradient is determined based on the shallow reflection intensity characteristics and the approximation residual, and the current spatial mapping weights are corrected based on the optimization gradient.

[0011] The shallow reflected light intensity characteristics are reconstructed and mapped based on the corrected spatial mapping weights, and the shallow noise component representing macroscopic body motion interference is output.

[0012] Differential stripping is performed on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle.

[0013] The smooth muscle clonus envelope features within the preset observation time window are reconstructed using the analytical domain to obtain the smooth muscle phase evolution trajectory;

[0014] Determine whether the sustained convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic reference phase satisfies the preset diastolic window matching condition;

[0015] If so, the power compensation gain is calculated based on the smooth muscle phase evolution trajectory, a targeted intervention luminescence command is generated and output; otherwise, the current intervention state is maintained.

[0016] A precise menstrual pain relief system based on red and near-infrared light includes:

[0017] The optical signal acquisition module is used to acquire the shallow reflected light intensity characteristics and deep mixed light intensity characteristics of the target area on the human abdomen at the current sampling time.

[0018] The active interference suppression module is used to spatially project the shallow reflected light intensity features based on the current spatial mapping weights to obtain the initial simulated noise features, and calculate the approximation residual between the deep mixed light intensity features and the initial simulated noise features; determine the optimization gradient based on the shallow reflected light intensity features and the approximation residual, and correct the current spatial mapping weights based on the optimization gradient; reconstruct the shallow reflected light intensity features based on the corrected spatial mapping weights, and output the shallow noise component representing macroscopic body motion interference;

[0019] The smooth muscle contraction feature extraction module is used to perform differential stripping on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle.

[0020] The phase trajectory analysis module is used to perform analytical domain reconstruction of the smooth muscle clonus envelope features within a preset observation time window to obtain the smooth muscle phase evolution trajectory.

[0021] The targeted intervention decision module is used to determine whether the continuous convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic reference phase meets the preset diastolic window matching condition; if yes, it calculates the power compensation gain based on the smooth muscle phase evolution trajectory, generates a targeted intervention luminescence command and outputs it; otherwise, it maintains the current intervention state.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] This invention uses the intensity of shallow reflected light as a reference benchmark for macroscopic somatic motion interference and introduces a dynamic optimization gradient based on approximation residuals to construct an adaptive correction closed loop for spatial mapping weights. This mechanism can accurately extract pure smooth muscle clonus envelope features even under strong interference environments such as patient rolling and curling up, thus effectively solving the problem of motion artifacts masking weak signals in deep target areas. Addressing the characteristics of large amplitude fluctuations and non-strictly consistent periods in smooth muscle contraction signals, this invention reconstructs the one-dimensional time-domain envelope signal onto a complex plane through analytical domain reconstruction. It directly calculates and extracts the smooth muscle phase evolution trajectory, unaffected by local amplitude abrupt changes. The judgment of smooth muscle state is upgraded from traditional threshold coarse estimation to precise tracking of continuous phases, providing a high-confidence digital foundation for subsequent target window determination. This technology departs from the continuous or blind pulse irradiation methods of existing technologies, strictly limiting the intervention time to the relaxation window of the smooth muscle evolution trajectory. By performing power compensation gain within a small time window when the smooth muscle is in a relaxation trend and the instantaneous phase is stable, it utilizes the physical property of high tissue transmittance during the relaxation period to enable photons to penetrate the target area deeply and maximize the photobiological regulation effect. On the other hand, it maintains the baseline state during the non-relaxation period, effectively avoiding the energy waste caused by ineffective irradiation and the risk of potential accumulated tissue thermal damage. Thus, it achieves intelligent physical intervention with precise targeting and on-demand drug delivery. Attached Figure Description

[0024] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0025] Figure 1 This is a diagram illustrating the method steps of the present invention;

[0026] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0027] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0028] Application Overview:

[0029] In the existing field of non-invasive physical intervention in gynecology, especially in photobiological regulation therapy for primary dysmenorrhea, the precise control of the timing of intervention is regarded as the core indicator for determining the efficacy of phototherapy. This highly efficient phototherapy intervention is essentially a process of strict synchronization of internal and external rhythms at the physiological level. That is, by using red light and near-infrared light to penetrate the body surface tissue, it precisely captures and matches the mechanical contraction cycle of uterine smooth muscle, and injects photon energy in a targeted manner during a specific window period when the smooth muscle is in a relaxation trend, thereby maximizing the local microvascular dilation effect and spasm relief rate.

[0030] However, existing technologies lack an effective decoupling mechanism for the coupling relationship between macroscopic motion interference on the body surface and weak physiological signals in the deep layers. This makes it impossible to accurately remove the complex somatic motion artifacts caused by the severe pain during a dysmenorrhea attack. Macroscopic somatic motion manifests as the relative displacement of the detection interface and the violent fluctuations in light intensity caused by the patient rolling, curling up, or deep breathing. In contrast, weak physiological signals in the deep layers are deeply masked or superimposed by this low-frequency and high-amplitude external interference when the smooth muscle slowly spasms. As a result, a strict mapping relationship cannot be established between the optical image captured by the detector and the actual mechanical contraction state of the deep smooth muscle. This leads to serious misjudgment of the smooth muscle phase or signal destruction by the system, which in turn affects the accurate determination of the timing of targeted illumination and the pertinence of power compensation.

[0031] For example, in real-world phototherapy scenarios for dysmenorrhea, when patients unconsciously curl up and roll over due to paroxysmal colic, traditional photoelectric monitoring systems can only capture the dramatic changes in mixed light intensity caused by epidermal displacement, but cannot adaptively distinguish and remove superficial noise interference in the spatial domain. Furthermore, when the frequency band of this motion artifact highly overlaps with the frequency band of the slow contraction of smooth muscle itself, the system's conventional linear filtering can only filter out high-frequency electrical noise, but will also distort or even erase the deep, real smooth muscle envelope characteristics. Specifically, the system is very likely to misjudge the dramatic fluctuations in body displacement as the tetanic contraction phase of smooth muscle and give incorrect intervention and restraint, or misclassify the troughs of artifacts as the smooth muscle relaxation window and blindly apply high-power irradiation. As a result, the system's intervention instructions are seriously out of sync with the real pathophysiological rhythm, and it is impossible to form a closed-loop targeted feedback that conforms to the principles of photophysical transmission.

[0032] If the above problems are not addressed, the phototherapy intervention system will continue to lose its objective ability to detect the true evolutionary state of uterine smooth muscle. Specifically, the failure to effectively eliminate macroscopic somatic interference will subject the system to noisy signals from appearances, causing the timing of photon energy injection to deviate from the principle of efficient absorption during the targeted relaxation phase, thus significantly weakening the core efficacy of phototherapy in promoting blood circulation. Simultaneously, the failure to thoroughly correct the misjudgment of phase evolution trajectory will result in energy waste during ineffective irradiation periods and abnormal accumulation of heat in local tissues, causing phototherapy intervention to lose its proper dynamic adaptive characteristics. Ultimately, it may even exacerbate patient discomfort due to the superposition of heat and pain. Therefore, the inaccuracy of closed-loop feedback will systematically hinder dysmenorrhea patients from obtaining precise physical analgesia support, severely restricting the achievement of clinical intervention goals for advanced phototherapy equipment.

[0033] like Figure 1 and Figure 2 As shown, a method for precisely relieving menstrual cramps based on red and near-infrared light includes the following steps:

[0034] Step 1: Obtain the superficial reflected light intensity characteristics and deep mixed light intensity characteristics of the human abdominal target area at the current sampling time; simultaneously acquire the optical physiological signals of the abdominal target area using optical detection arrays with different detection depths. The superficial reflected light intensity characteristics mainly characterize the fluctuations of the epidermal optical interface and the interference signals generated by macroscopic displacement of the body; the deep mixed light intensity characteristics include physiological signals that penetrate to the uterine smooth muscle layer and carry its mechanical contraction state, while also being superimposed with superficial tissue interference along the photon propagation path. Through this step, spatial layering acquisition of interference sources and mixed signal sources is achieved at the physical acquisition front end.

[0035] Step 2: Based on the current spatial mapping weights, spatial projection is performed on the shallow reflected light intensity characteristics to obtain the initial simulated noise characteristics, and the approximation residual between the deep mixed light intensity characteristics and the initial simulated noise characteristics is calculated. Since there is nonlinear physical attenuation and time delay in the propagation of shallow interference to deep tissue, the current spatial mapping weights are extracted as the propagation model, and spatial projection operation is performed on the collected shallow reflected light intensity characteristics to estimate the initial simulated noise characteristics currently existing in the deep tissue. Subsequently, the difference between the deep mixed light intensity characteristics and the initial simulated noise characteristics is compared to obtain the approximation residual between the two. This residual objectively quantifies the estimation deviation between the current spatial mapping model and the actual physical tissue state.

[0036] Step 3: Determine the optimization gradient based on the shallow reflection intensity characteristics and the approximation residual, and correct the current spatial mapping weights based on the optimization gradient; To address the dynamic time-varying problem of tissue optical parameters caused by patient rolling, breathing, etc., a closed-loop adaptive feedback mechanism is established. By extracting the correlation characteristics between the shallow reflection intensity characteristics and the approximation residual, an optimization gradient aimed at minimizing the approximation residual is calculated. The current spatial mapping weights are iteratively corrected using this optimization gradient, giving the system the ability to track and adapt to the physical deformation of the detection interface caused by body movements in real time.

[0037] Step 4: Reconstruct the shallow reflected light intensity characteristics based on the corrected spatial mapping weights, and output the shallow noise component representing macroscopic bodily motion interference. Discard the outdated weights that produce inherent delays, and substitute the corrected spatial mapping weights with the latest physical state recognition accuracy back into the calculation link to reconstruct the shallow reflected light intensity characteristics. This process eliminates the phase lag caused by adaptive iteration and outputs the shallow noise component that is highly aligned with the actual deep interference state in both amplitude and phase, providing a high-confidence dynamic reference benchmark for subsequent signal stripping.

[0038] Step 5: Perform differential stripping on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle; subtract the high-precision shallow noise components from the deep mixed light intensity features to remove macroscopic body motion artifacts and extract the pure smooth muscle optical signal. Further extract and smooth the energy profile of the pure signal to filter out high-frequency optical noise and fluctuations in non-target physiological frequency bands. Finally, output the smooth muscle clonus envelope features that exhibit low-frequency fluctuations and objectively characterize the periodic contraction and relaxation mechanical tension changes of uterine smooth muscle.

[0039] Step Six: Reconstruct the smooth muscle clonus envelope features within the preset observation time window using the analytical domain to obtain the smooth muscle phase evolution trajectory. To overcome the influence of signal absolute amplitude drift on state determination, the envelope feature sequence within the preset time window is extracted and mapped from the one-dimensional time domain to a two-dimensional complex analytical domain containing orthogonal features. By extracting the instantaneous phase angle of the analytical signal and performing continuous processing, the amplitude information representing muscle tension is transformed into phase information representing the periodic evolution state, generating a monotonically continuous smooth muscle phase evolution trajectory.

[0040] Step 7: Determine whether the continuous convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic baseline phase meets the preset diastolic window matching condition; by comprehensively monitoring the current position and rate of change of the phase evolution trajectory, only when the instantaneous phase falls within the preset range representing the mechanical relaxation of the smooth muscle, and the evolution rate of the phase is in a preset stable state, and can maintain this dual characteristic in a continuous observation period, is it determined that the current physiological state accurately meets the preset diastolic window matching condition, thereby effectively avoiding false triggering caused by transient spikes.

[0041] Step 8: If yes, calculate the power compensation gain based on the smooth muscle phase evolution trajectory, generate a targeted intervention luminescence command, and output it; otherwise, maintain the current intervention state. When the diastolic window matching condition is met, dynamically calculate the corresponding power compensation gain based on the degree to which the current phase trajectory approaches the optimal diastolic depth (i.e., the preset diastolic baseline phase), and generate a targeted intervention luminescence command to drive the hardware to increase the optical radiation flux, achieving highly efficient phototherapy with maximum microvascular dilation. If the diastolic window matching condition is not met, indicating that the smooth muscle is in a tetanic contraction or unstable transition period, maintain the baseline low-power intervention state to avoid the risk of tissue heat accumulation caused by ineffective irradiation, and achieve intelligent physical intervention with on-demand drug delivery.

[0042] Example 1: In a monitoring scenario where a patient with dysmenorrhea experiences body rolling and disturbance due to paroxysmal pain while wearing a phototherapy device with dual-layer detection channels, the sampling frequency is set to... .

[0043] Assuming the current time The patient is experiencing slow clonic contractions of the uterine smooth muscle, accompanied by rolling of the body. The preset baseline detection light power is... The maximum optical power compensation limit is .

[0044] Step 1: The device's underlying hardware drives the near-end and far-end photodetectors to complete analog-to-digital conversion. At a certain moment, the proximal channel collected high-amplitude voltage fluctuations caused by abdominal skin compression, and the shallow reflected light intensity characteristics were obtained after normalization. The voltage acquired by the remote channel, which is mixed with superficial interference and deep smooth muscle contraction signals, was normalized to obtain the deep mixed light intensity characteristics. .

[0045] Step 2: The microprocessor extracts the first-order spatial mapping weights stored in the current register:

[0046] And retrieve the current and the previous two shallow features to form a vector:

[0047] Perform inner product projection calculations to estimate the initial simulation noise characteristics:

[0048] Then, the approximation residuals are calculated:

[0049] The negative residual indicates that the current model estimates the noise slightly higher than the actual amplitude of the deep mixed signal.

[0050] Step 3: Set the convergence step size using the approximation residual correction propagation model. Calculate the optimization gradient:

[0051] ; Calculate the adjustment amount and update the weights:

[0052] ;

[0053] This step completes online learning of the physical state of changes in optical boundary parameters caused by the patient's rolling.

[0054] Step 4: Write the newly corrected weights into the tap coefficients of the digital adaptive filter, and then re-apply the current shallow feature vector. Perform convolution operation:

[0055] ;

[0056] The shallow noise component obtained after remapping ( It achieves precise alignment with the current actual deep interference in both physical and temporal dimensions.

[0057] Step 5: Perform differential operation to remove macroscopic tumbling interference, net smooth muscle signal. :

[0058] ;

[0059] Perform full-wave rectification (take the absolute value) on the net signal. ), and input the cutoff frequency as The low-pass filter is assumed to have smooth myoclonus envelope characteristics in the filtered output. This value objectively reflects the absolute mechanical tension of the uterine smooth muscle at the current physical node.

[0060] Step 6: Extract Using historical observation window envelope data at the end of the time frame, perform Hilbert transform to obtain the orthogonal imaginary part. Construct the analytical signal and calculate the instantaneous phase angle:

[0061] radian;

[0062] The unwinding process determined that there was no abrupt change, and the current smooth muscle phase evolution trajectory was recorded as the absolute phase angle. Radius (corresponding to the early diastolic phase of the second cycle).

[0063] Step 7: Map the absolute phase angle to a single period:

[0064] radian;

[0065] The preset safe range for the diaphragm window is known to be... Radius, judgment If the phase falls within this interval, calculate and verify the rate of change of phase angle: radians / period, less than the stability threshold Radius / period, checking the historical register, it was found that this stable diastolic state has been maintained for 3 consecutive cycles, meeting the anti-digital jitter requirement. Therefore, it was determined that the diastolic window matching condition is met.

[0066] Step 8: Determine whether to enter the targeted intervention branch. The preset diastolic center phase angle is known to be... Curve, half-width of relaxation is The radians are used to calculate the degree to which the current phase angle deviates from the relaxation target center, and then substituted into the parabolic power compensation gain function:

[0067] ;

[0068] Generate the final emission command: ;

[0069] The digital-to-analog converter outputs the corresponding voltage to the LED driver, enabling the device to output accurately during the current diastolic window. The radiation power maximizes the blood perfusion volume of the target area while ensuring no thermal damage.

[0070] The core innovation of the above technology lies in abandoning the limitations of traditional frequency domain filtering and proposing a dual-depth light intensity feature spatial projection architecture. By using the shallow reflected light intensity as a reference benchmark for macroscopic body motion interference and introducing a dynamic optimization gradient based on approximation residuals, an adaptive correction closed loop of spatial mapping weights is constructed. This mechanism can accurately extract pure smooth muscle clonus envelope features in strong interference environments such as patient rolling and curling up, thereby effectively solving the problem of motion artifacts masking weak signals in deep target areas.

[0071] To address the characteristics of smooth muscle contraction signals, such as large amplitude fluctuations and non-strictly consistent periods, this paper reconstructs the envelope signal in the one-dimensional time domain to the complex plane through analytical domain reconstruction. This allows for the direct calculation and extraction of the smooth muscle phase evolution trajectory, which is unaffected by local amplitude abrupt changes. The judgment of smooth muscle state is upgraded from the traditional coarse threshold estimation to the precise tracking of continuous phases, providing a high-confidence digital foundation for subsequent target window determination.

[0072] This technology departs from the continuous or blind pulse irradiation methods of existing technologies, strictly limiting the intervention time to the relaxation window of the smooth muscle evolution trajectory. By performing power compensation gain within a small time window when the smooth muscle is in a relaxation trend and the instantaneous phase is stable, it utilizes the physical property of high tissue transmittance during the relaxation period to enable photons to penetrate the target area deeply and maximize the photobiological regulation effect. On the other hand, it maintains the baseline state during the non-relaxation period, effectively avoiding the energy waste caused by ineffective irradiation and the risk of potential accumulated tissue thermal damage. Thus, it achieves intelligent physical intervention with precise targeting and on-demand drug delivery.

[0073] Because single-channel optical detection cannot physically distinguish whether the received light intensity fluctuations originate from periodic clonic movements of the target's deep smooth muscle or from changes in the tissue optical interface caused by bodily actions such as abdominal skin undulations and probe sliding, subsequent signal processing lacks decoupling prior conditions. Therefore, this paper proposes acquiring the superficial reflected light intensity characteristics and deep mixed light intensity characteristics of the human abdominal target area at the current sampling time, as follows:

[0074] Among them: the human abdominal target area refers to the area on the patient's lower abdomen where the phototherapy and signal acquisition device are attached, and its deep layer corresponds to the physical location of the uterine smooth muscle.

[0075] Current sampling time: The current measurement node in a discrete time series, usually represented as a time variable. This refers to the specific physical time required for the analog-to-digital converter (ADC) to perform a single data sampling.

[0076] Superficial reflected light intensity characteristics: After the light source is emitted, the optical signal is reflected by the superficial tissues of the epidermis and dermis of the human skin and received by the detector with a short light source-detector spacing. This characteristic mainly reflects the low-frequency interference caused by changes in epidermal optical coupling and macroscopic displacement of the body. It does not contain deep physiological rhythm information. In specific hardware deployments, the short light source-detector spacing is usually set to 5mm-10mm. The photon migration path at this spacing is mainly concentrated in the epidermis and superficial dermis, which can maximize the capture of interference caused by changes in surface optical coupling due to the relative displacement of the probe.

[0077] Deep mixed light intensity characteristics: After the light source is emitted, it penetrates into the subcutaneous tissue and the smooth muscle layer of the uterus. After scattering and absorption, it returns to the body surface and is quantified by the optical signal received by the long-spacing detector. This characteristic contains not only the contraction state information of the target deep smooth muscle, but also the interference from superficial tissues and macroscopic body movements along the light propagation path. The spacing of the long-spacing detector is usually set to 30mm-45mm. Based on the banana-shaped scattering path model of photons in biological tissues, the maximum penetration depth at this spacing can reach 15mm-25mm, which can effectively penetrate the subcutaneous fat layer of the abdomen and reach the surface of the uterine smooth muscle, thereby obtaining a deep mixed signal containing the mechanical contraction information of the smooth muscle.

[0078] Step 1: At the current sampling time A constant-power near-infrared light is emitted towards the target area of ​​the human abdomen using a light-emitting diode (LED). Two sets of photodetectors (PDs) positioned near and far of the target area simultaneously receive the reflected photons. The near-end PD outputs an analog current signal representing the shallow layer, while the far-end PD outputs an analog current signal representing the deep layer. Subsequently, the signal is processed by a transimpedance amplifier circuit and an analog-to-digital converter circuit to obtain the current shallow layer raw voltage sample value. Compared with deep original voltage sampling value .

[0079] Step 2: Since different patients have different skin color and stratum corneum thickness, different static baseline drifts will be introduced. In order to ensure the accuracy of subsequent spatial projection weight calculation, the original sampled values ​​need to be linearly normalized to extract light intensity features with dimensionless properties.

[0080] First, the shallow layer reflected light intensity features are extracted, using the following formula:

[0081] ;

[0082] in: Indicates the current sampling time The intensity characteristics of the shallow reflected light output;

[0083] Indicates the current sampling time The measured shallow original voltage sample value;

[0084] This represents the baseline for the minimum value of the shallow signal recorded within the preset initialization window;

[0085] This represents the reference value of the shallow signal maximum recorded within the preset initialization window.

[0086] Then, the deep mixed light intensity features are extracted, using the following formula:

[0087] ;

[0088] in: Indicates the current sampling time The intensity characteristics of the output deep mixed light;

[0089] Indicates the current sampling time The measured deep-seated raw voltage sampling value;

[0090] This represents the minimum reference value of the deep signal recorded within the preset initialization window;

[0091] This represents the reference value of the deep signal maximum recorded within the preset initialization window.

[0092] It should be noted that the preset initialization window is a fixed observation period after the device is started or after the wearing state is stable, such as 10 consecutive sampling periods. To prevent baseline contamination under strong interference conditions, the variance of the signal amplitude within this window will be evaluated before obtaining the extreme value benchmark. Only when the variance is lower than the preset resting threshold, i.e., it is determined that the patient is not currently in a state of violent rolling, will the maximum and minimum voltage values ​​captured within this window be locked as the benchmark. Otherwise, the window will slide backward until the resting condition is met.

[0093] Example 2: The system initialization is set to complete, and the current device is... Sampling rate operation, setting the current sampling time. Retrieve pre-calibrated patient baseline extreme value data:

[0094] Shallow signal minimum benchmark Maximum benchmark ;

[0095] Deep signal minimum benchmark Maximum benchmark ;

[0096] exist At any given moment, the hardware sensors collect data in real time:

[0097] Shallow raw voltage sampling value ;

[0098] Deep raw voltage sampling value ;

[0099] Perform shallow normalized feature extraction calculation:

[0100] ;

[0101] Perform deep normalized feature extraction calculation:

[0102] ;

[0103] The final output of step one is: The shallow reflected light intensity characteristic value at the current sampling time has been successfully obtained. The characteristic value of deep mixed light intensity is And store the above values ​​into a register.

[0104] In the aforementioned technology, the physical spatial transmission characteristic of near-infrared light in biological tissues, where the longer the light source-detector distance (SDS), the deeper the detection depth, is utilized to simultaneously construct a dual-depth detection architecture. The short-pitch channel captures pure shallow interference as a noise reference source, while the long-pitch channel captures a mixed source containing both the target signal and interference. This achieves spatial layering of the signal at the physical acquisition front end. The two independent light intensity characteristics obtained through this layering provide a direct and objective data input basis for subsequent spatial mapping and adaptive differential stripping, ensuring the reliability of weak smooth muscle contraction feature extraction.

[0105] Because superficial bodily motion disturbances inevitably undergo nonlinear amplitude attenuation and phase hysteresis when propagating to deeper tissues, directly subtracting the superficial signal from the deep signal, without spatial transfer function calibration, not only fails to eliminate artifacts but also introduces secondary computational noise, leading to the destruction of the true smooth muscle contraction characteristics. Therefore, this paper proposes spatial projection of the superficial reflected light intensity characteristics based on the current spatial mapping weights to obtain initial simulated noise characteristics, and calculates the approximation residual between the deep mixed light intensity characteristics and the initial simulated noise characteristics, as detailed below:

[0106] The approximation residuals between the deep mixed light intensity characteristics and the initial simulated noise characteristics are calculated, specifically including:

[0107] The shallow reflected light intensity characteristics at the current sampling time and its historical times are used to construct a multi-dimensional reference signal vector;

[0108] Calculate the inner product of the multidimensional reference signal vector and the current spatial mapping weight vector to obtain the scalar projection value, which serves as the initial simulated noise feature;

[0109] Time delay evaluation and timing alignment processing are performed on the deep mixed light intensity characteristics and the initial simulated noise characteristics;

[0110] The instantaneous amplitude of the temporally aligned deep hybrid light intensity feature is subtracted from the instantaneous amplitude of the initial simulated noise feature to obtain a temporal difference sequence, which serves as the approximation residual.

[0111] Wherein: Current spatial mapping weight: a mathematical coefficient representing the attenuation and phase shift experienced by local light as it propagates from the shallow layer to the deep layer in biological tissue. In the discrete time domain, this weight exists in the form of a column vector containing multiple tap coefficients.

[0112] Spatial projection: The mathematical mapping process of transforming optical features collected in one spatial dimension (shallow tissue interface) to another spatial dimension (deep smooth muscle interface) according to a specific transfer relationship (weight). Its essence is the inner product operation of multidimensional feature vectors and weight vectors.

[0113] Initial simulation noise characteristics: Macroscopic body motion interference values ​​that exist in deep tissues at the current moment, estimated based on current shallow interference data and existing spatial transmission experience (weights).

[0114] Approximation residual: The numerical difference between the actual measured deep mixed light intensity and the initial simulated noise characteristics estimated by the system, used to measure the estimation error of the current spatial mapping weights.

[0115] Step 1: Due to the time lag in the optical response of biological tissues, the current deep interference is related not only to the current shallow state but also to the past shallow state. Therefore, it is necessary to extract the shallow reflected light intensity features at the current sampling time and its historical times, constructing a multi-dimensional column vector. The specific formula is as follows:

[0116] ;

[0117] in: To be at the current sampling time Constructed multidimensional reference signal vector;

[0118] The current sampling time The shallow reflected light intensity characteristics;

[0119] The dimension of the preset spatial mapping weight vector (i.e., the order of the adaptive filter).

[0120] This represents the transpose operation of a matrix or vector.

[0121] Step 2: Extract the current spatial mapping weight vector stored in the register and calculate its inner product with the multidimensional reference signal vector. This inner product operation completes the spatial projection of shallow features to deeper dimensions. The specific formula is as follows:

[0122] ;

[0123] in: The obtained scalar projection value is the initial simulated noise characteristic;

[0124] The current sampling time The spatial mapping weight vector is represented as follows:

[0125] ;

[0126] in, The first in the weight vector Tap coefficient.

[0127] Step 3: Due to the slight inherent delay in the hardware response and signal processing links during photon transmission in both long and short channels, direct subtraction would lead to phase misalignment. This fixed physical delay needs to be calculated and compensated for. The specific formula is as follows:

[0128] ;

[0129] in: The intensity characteristics of the deep mixed light after time alignment;

[0130] To translate the deep mixed light intensity characteristics of a specific period to a historical moment;

[0131] This is the relative time delay constant extracted based on dual-channel hardware calibration.

[0132] Step 4: Subtract the estimated instantaneous amplitude of the initial simulated noise feature from the instantaneous amplitude of the time-aligned deep hybrid intensity feature, and output the time-domain deviation. The specific formula is as follows:

[0133] ;

[0134] in: This is the approximation residual at the current sampling time in the output time-domain difference sequence.

[0135] Example 3: Setting the preset spatial mapping weight vector dimension The current sampling time is Time delay constant (Assuming the hardware has achieved zero-latency synchronization).

[0136] Retrieve the shallow reflected light intensity characteristics at the current and the two previous sampling times:

[0137] , , ;

[0138] Generate column vectors: .

[0139] Retrieve the current weighting coefficients from the register: , , 0.

[0140] Generate column vectors: .

[0141] Calculate spatial projection (initial simulation noise characteristics):

[0142] .

[0143] The current instantaneous value of the deep mixed light intensity characteristic is .

[0144] Calculate the approximation residual: .

[0145] Output: The final calculated approximation residual for the current time step is... .

[0146] In the aforementioned technique, multidimensional vector inner product operation (spatial projection) is introduced. Convolutional algebra operations are performed using shallow feature vectors containing historical states and multi-order weight vectors to simulate the physical scattering and attenuation process of photons in the tissue medium. Through time alignment processing, the phase of the dual-channel signals is strictly synchronized, and an interference estimation model that conforms to the objective laws of tissue optical propagation is constructed. The model outputs an approximate residual that accurately reflects the current weight estimation accuracy. This residual directly contains the unfiltered net smooth muscle signal and estimation error, providing a high-confidence target scalar for subsequent steps of error feedback-based gradient correction.

[0147] Because the physical movements (such as rolling and breathing) of patients during dysmenorrhea attacks are highly time-varying and non-stationary, the path attenuation coefficient and phase delay of superficial interfering photons propagating to the deep target area are constantly changing. Using static or pre-set fixed empirical weights will fail to cope with this dynamic physical deformation, leading to artifact removal failure. Therefore, this paper proposes determining the optimization gradient based on the characteristics of shallow reflected light intensity and the approximate residual, and correcting the pre-set spatial mapping weights based on the optimization gradient, including:

[0148] Calculate the product of the shallow reflected light intensity characteristics and the approximation residual at the current sampling time to obtain the optimization gradient characterizing the direction of interference correlation;

[0149] The spatial mapping weight adjustment amount is obtained by adjusting the gain of the optimization gradient through the convergence step size.

[0150] The spatial mapping weight adjustment is accumulated and added to the current spatial mapping weight to obtain the corrected spatial mapping weight.

[0151] Among them: Optimization gradient: The optimization gradient is a vector parameter used to characterize the direction and magnitude of the spatial mapping weights that should be adjusted in the current state. Its essence is the cross-correlation between input interference (shallow features) and output error (approximation residual), pointing in the direction of decreasing mean square value of approximation residual.

[0152] Convergence step size: A scalar coefficient used to calibrate the magnitude of the optimization gradient. This coefficient determines the rate at which the system updates its weights within a single sampling period. A value that is too large can easily lead to system divergence, while a value that is too small can cause the system to lag in tracking dynamic disturbances.

[0153] Spatial mapping weight adjustment amount: The numerical change that the optimization gradient actually affects the current spatial mapping weight after the convergence step size gain is adjusted.

[0154] Step 1: Extract the multidimensional reference signal vector representing the current and historical states generated in Step 2, i.e., the shallow reflected light intensity feature vector, and perform a scalar multiplication operation with the approximation residual obtained at the current sampling time. This operation quantifies the correlation between the current input interference signal and the final estimation error. The specific formula is as follows:

[0155] ;

[0156] in: To be at the current sampling time The calculated optimization gradient vector;

[0157] To be at the current sampling time The constructed multidimensional reference signal vector is the shallow reflected light intensity feature sequence;

[0158] To be at the current sampling time The approximate residual is obtained.

[0159] Step 2: Retrieve the historical and current approximation residual sequences within the preset observation time window, and calculate the time average of their instantaneous energy (squared value). The specific formula is as follows:

[0160] ;

[0161] in: To be at the current sampling time The obtained residual energy measure coefficients;

[0162] This is the number of sampling periods included within the preset observation time window;

[0163] These are the approximate residual sequence values ​​for the current and historical sampling times.

[0164] Step 3: Traditional adaptive filtering algorithms (such as the standard LMS algorithm) use a fixed convergence step size. This mechanism has an inherent physical contradiction: if a large step size is set, the system converges quickly and can respond to violent bodily movements, but this leads to a large steady-state imbalance and a lot of residual noise after convergence. If a small step size is set, the steady-state accuracy is high, but when faced with rapid tumbling caused by sudden pain, the system tracking speed lags significantly, resulting in severe signal distortion. Therefore, a dynamic convergence step size based on error energy negative feedback is proposed, specifically including:

[0165] Extract the instantaneous average energy of the approximation residual within the preset observation time window, and use it as the residual energy measure coefficient to characterize the degree of prediction bias;

[0166] The residual energy measurement coefficients are input into a preset monotonically increasing mapping function to calculate the step gain;

[0167] Calculate the product of the step gain and the preset baseline step size to generate the final convergence step size.

[0168] The specific formula for the monotonically increasing mapping function is:

[0169] ;

[0170] in, Step size gain;

[0171] This is the preset maximum gain limit;

[0172] The residual energy measurement coefficient;

[0173] The preset energy threshold;

[0174] The preset curve steepness coefficient, and .

[0175] Among them: Error energy negative feedback regulation: a closed-loop control strategy that continuously monitors the square mean of the system output error (approaching residual) (i.e. energy) and uses this as a basis to dynamically adjust the system's learning rate (convergence step size). When the error energy increases, the learning rate is increased, and when it decreases, the learning rate is decreased.

[0176] Residual energy measure coefficient: The time average of the instantaneous squared values ​​of the approximation residual within a specific time window, used to quantify the overall prediction bias level of the current spatial mapping model.

[0177] Monotonically increasing mapping function: a mathematical transformation model whose output value increases strictly with the increase of the input value. In this application, it is specifically manifested as a sigmoid function with a saturation upper limit and a nonlinear transition region.

[0178] Baseline step size: A pre-defined constant scalar that represents the base weight update rate when not subject to dynamic gain adjustment.

[0179] Step gain: A dimensionless product factor calculated by the mapping function, used to scale the baseline step size in real time.

[0180] The residual energy measurement coefficients obtained in the previous step are input into a preset monotonically increasing mapping function to calculate the dynamic amplification factor for the current disturbance state. By introducing a sigmoid-type monotonically increasing mapping function with the approximation of the instantaneous residual energy as the independent variable, the severity of the current external disturbance is assessed by calculating the residual energy measurement coefficients within the time window. This drives the convergence step size to switch nonlinearly and continuously smoothly between the set upper and lower limits, effectively resolving the contradiction between convergence rate and steady-state accuracy. When the patient experiences violent macroscopic physical movements (residual energy exceeding the threshold), the dynamic amplification factor is calculated. When ), the mapping function quickly approaches The step size is adaptively amplified to achieve rapid tracking and fitting of the physical boundary of mutation. When the patient's body is stable (the residual energy is below the threshold), the mapping function decays exponentially and the step size is restored to close to the lower limit of the baseline, ensuring steady-state high fidelity of the separation of deep weak smooth muscle features.

[0181] Step 4: Multiply the step size gain by the baseline step size to obtain the actual convergence step size used at the current time, and use this step size to adjust the gain of the optimization gradient, outputting the weight adjustment amount. The specific formula is as follows:

[0182] ;

[0183] in: To be at the current sampling time The generated spatial mapping weight adjustment vector;

[0184] The preset baseline step size constant for the system;

[0185] The step size gain obtained above;

[0186] To find the optimal gradient vector.

[0187] Step 5: Add the obtained spatial mapping weight adjustment to the current weight to complete the iteration. The specific formula is as follows:

[0188] ;

[0189] in: This is the corrected spatial mapping weight vector that will be used in the next sampling period;

[0190] This is the spatial mapping weight vector at the current moment;

[0191] This is the spatial mapping weight adjustment vector for the current moment.

[0192] Example 4: Continuing from the calculation example in Example 3, the current sampling time is... Given:

[0193] Multidimensional reference signal vector ;

[0194] Approaching residual ;

[0195] Current weight vector .

[0196] Assume the time window includes the current moment and the two previous moments ( ), and the recorded historical residual squares are , 4.

[0197] System preset constant settings: reference step size ;

[0198] Maximum gain limit ;

[0199] Energy threshold ;

[0200] Curve steepness coefficient .

[0201] Perform a scalar multiplication between the shallow feature vectors and the approximation residuals:

[0202] .

[0203] Calculate the current squared residual: .

[0204] Calculate the mean within the observation window:

[0205] .

[0206] Substituting into the formula for the monotonically increasing mapping function:

[0207] Calculation of the exponent:

[0208] .

[0209] Known natural constant Gain calculation:

[0210] .

[0211] Generate the final convergence step size: .

[0212] Calculate the weight adjustment:

[0213] ;

[0214] ;

[0215] Adjusted weights:

[0216] ;

[0217] .

[0218] Due to the currently calculated residual energy ( The threshold value exceeded the set safety threshold. The system determined that it was experiencing some degree of physical interference. Therefore, it triggered gain adjustment through a negative feedback mechanism, adjusting the originally fixed reference step size (…). Dynamically zoomed in This makes the adjustment of the corrected weight vector more significant compared to when no energy gain is used, thus giving the model faster convergence tracking ability.

[0219] In the aforementioned technology, through an error feedback mechanism, relevant information between the approximation residual and the input shallow features is extracted at each sampling point. Using the gradient descent principle, by setting a convergence step size and smoothing gain, the weight state is driven in real time toward minimizing the approximation residual. Thus, when faced with unpredictable changes in the physiological detection interface caused by the patient's body movements, it can reconverge and fit a new spatial interference model with an extremely short time delay, ensuring the robustness and accuracy of subsequent smooth muscle feature separation.

[0220] The initial simulated noise features are calculated based on outdated weights retained from the previous sampling period. Since the optical boundaries of biological tissues undergo drastic changes during somatic movements caused by dysmenorrhea, directly using this initial simulated noise for signal differentiation will leave significant misalignment errors, i.e., a fundamental leak in the algorithm, resulting in residual motion artifacts in the final separated smooth muscle signal. Therefore, a method is proposed to reconstruct the shallow reflected light intensity features based on corrected spatial mapping weights, outputting a shallow noise component characterizing macroscopic somatic movement interference, specifically including:

[0221] The corrected spatial mapping weights are updated to the tap coefficients of the preset finite impulse response adaptive filter;

[0222] The shallow reflected light intensity characteristics of the continuous time series are input into the finite impulse response adaptive filter after updating the tap coefficients for convolution filtering calculation, and the shallow noise component is output.

[0223] Among them: Reconstruction mapping: a spatial transformation process that recalculates the original input variables using iteratively updated mapping coefficients with higher parameter accuracy. Unlike spatial projection, which uses historical weight estimation, reconstruction mapping uses the latest corrected weights.

[0224] Shallow noise component: The final output after parameter optimization, which mathematically and physically closely approximates the actual macroscopic body motion interference signal in the current deep mixed light intensity characteristics.

[0225] Finite Impulse Response (FIR) Adaptive Filter: A digital filter structure whose output is only related to the current and a finite number of historical input values, and whose internal multiplication coefficients (tap coefficients) can be dynamically adjusted according to external feedback signals.

[0226] Tap coefficients: Gain product factors on each delay branch within an FIR filter. In this application, tap coefficients are mathematically equivalent to the component elements of the spatially mapped weights.

[0227] Convolution filtering calculation: an algebraic operation in discrete-time signal processing, which is represented by the sliding multiplication of the input signal sequence with the tap coefficient sequence of the filter and the summation.

[0228] Step 1: The control unit overwrites the corrected spatial mapping weight vector obtained in Step 3 into the register of the preset Finite Impulse Response (FIR) adaptive filter, completing the hardware or software-level state update of the tap coefficients. The specific formula is as follows:

[0229] ;

[0230] in: This is the parameter vector consisting of all the tap coefficients contained in the updated FIR adaptive filter;

[0231] This refers to the corrected spatial mapping weight vector generated in step three.

[0232] Step 2: Input the multidimensional reference signal vector constructed from the shallow reflection intensity features of the continuous time series into the FIR adaptive filter with updated tap coefficients, and perform discrete convolution calculation, i.e., calculate the inner product of the multidimensional reference signal vector and the latest tap coefficient vector, and output the final interference estimate. The specific formula is as follows:

[0233] ;

[0234] in: The shallow noise component output represents the macroscopic bodily motion interference contained in the current deep signal;

[0235] This is the transpose of the updated FIR adaptive filter tap coefficient vector;

[0236] To be at the current sampling time The constructed multidimensional reference signal vector is the current and historical shallow reflection light intensity feature sequence.

[0237] Example 5: Continuing from the calculation example in Example 4, the order of the FIR filter is set. The current sampling time is Given:

[0238] Corrected spatial mapping weight vector ;

[0239] Current shallow multidimensional reference signal vector .

[0240] Write the weights into the FIR filter:

[0241] Tap coefficient 0: ;

[0242] Tap 1 coefficient: ;

[0243] Tap 2 coefficient: ;

[0244] Obtain the tap coefficient vector .

[0245] vector with vector Perform convolution (sum of inner products):

[0246] ;

[0247] .

[0248] Output result: The final output shallow noise component is Compare the initial simulated noise characteristics obtained using the old weights ( The updated noise component values ​​were adaptively reduced. This tiny increment is the adaptive compensation made by the system based on the deep actual feedback, and this value prepares the precise numerical value for the next step of completely removing interference.

[0249] In the aforementioned technique, by introducing posterior computation logic, the latest weights obtained after correction based on the current time-approaching residual (representing the current error gradient) are substituted back into the current input features to perform FIR convolution. This real-time gradient-based remapping is equivalent to using the latest understanding of the current physical state to process the current input data, effectively eliminating the inherent time delay error of single-step iteration in the adaptive system. The output shallow noise component not only approximates the actual interference in amplitude, but also achieves strict alignment with the interference components in the deep signal in phase, providing a high-precision reference benchmark for subsequent steps to implement pure physical quantity differential stripping.

[0250] While simple time-domain subtraction removes macroscopic motion artifacts, the remaining net signal still contains disordered high-frequency fluctuations such as microvascular blood flow pulsation and high-frequency optical scattering noise. These original signals, with alternating positive and negative properties, lack clear periodic boundaries and cannot be directly used to determine the macroscopic contraction and relaxation phases of smooth muscle. Therefore, a differential stripping method is proposed to separate the deep mixed light intensity features from the shallow noise components, yielding smooth muscle clonus envelope features characterizing the mechanical contraction state of the target smooth muscle. Specifically, these features include:

[0251] After removing the shallow noise component from the deep mixed light intensity features, the denoised net smooth muscle signal sequence is obtained.

[0252] The net smooth muscle signal sequence is subjected to full-wave rectification to obtain the absolute amplitude sequence of the signal;

[0253] The absolute amplitude sequence is smoothed by using a low-pass filter with a preset cutoff frequency, and the slowly varying energy profile of the signal is extracted to obtain the smooth muscle clonus envelope feature.

[0254] Among them: Differential stripping: In the discrete time domain, the mathematical operation process of subtracting the estimated equivalent noise signal from the measured mixed signal containing the target signal and noise, thereby separating the target physiological signal.

[0255] Net smooth muscle signal sequence: The intermediate time-domain signal obtained after removing macroscopic somatic motion interference (shallow noise component) from the deep mixed light intensity characteristics. This sequence contains high-frequency electrophysiological fluctuations and low-frequency mechanical contraction information of uterine smooth muscle.

[0256] Full-wave rectification: a signal processing method that flips the negative amplitude portion of a bipolar (containing both positive and negative amplitudes) AC signal sequence, converting it entirely into positive amplitude, i.e., taking the absolute value of the original sequence.

[0257] Absolute amplitude sequence: The non-negative numerical sequence generated after the net smooth muscle signal is rectified by full-wave rectification.

[0258] Slowly varying energy profile: By filtering out high-frequency oscillation components in the signal, the remaining geometric boundary line reflects the slow fluctuation trend of the overall energy of the signal on the time axis.

[0259] Smooth muscle spasm envelope characteristics: The final output low-frequency physiological signal curve characterizing the periodic mechanical spasm and relaxation state of uterine smooth muscle.

[0260] Step 1: At the current sampling time, extract the time-aligned deep mixed light intensity features and subtract the shallow noise component output in Step 4 to obtain the denoised net signal. The specific formula is as follows:

[0261] ;

[0262] in: To be at the current sampling time The calculated values ​​of the net smooth muscle signal sequence;

[0263] The deep mixed light intensity characteristics are obtained by shifting a specific period to a historical moment (time-aligned).

[0264] To be at the current sampling time The shallow noise component of the output.

[0265] Step 2: Since the net smooth muscle signal includes high-frequency scattering changes at the tissue interface, its baseline may fluctuate around zero. To calculate the energy envelope, the absolute value of this sequence is taken. The specific formula is as follows:

[0266] ;

[0267] in: To be at the current sampling time The extracted absolute amplitude sequence values;

[0268] It is an absolute value mathematical operator;

[0269] This represents the value of the net smooth muscle signal sequence obtained in the previous step.

[0270] Step 3: To filter out the cardiac pulse wave (PPG fundamental frequency) and high-frequency electrophysiological noise while preserving the low-frequency characteristics representing the rhythmic contraction of smooth muscle, a low-pass digital filter with a preset cutoff frequency is used to perform convolution smoothing on the absolute amplitude sequence. The specific formula is as follows:

[0271] ;

[0272] in: For the final output at the current sampling time The smooth muscle clonus envelope feature;

[0273] The order of the preset low-pass filter;

[0274] For the first low-pass filter The unit impulse response coefficient (its cutoff frequency is strictly set within the typical contraction frequency band of uterine smooth muscle, which corresponds to the slow wave rhythm of human smooth muscle, and its actual physical frequency is usually distributed in the extremely low frequency range of 0.01Hz to 0.08Hz. Therefore, the cutoff frequency of the low-pass filter is configured to be slightly higher than the upper limit of this frequency band (e.g., set to 0.1Hz) to ensure that while filtering out cardiac pulse waves (PPG) with frequencies typically in the range of 1.0Hz to 1.5Hz and higher frequency power frequency interference, the slowly varying energy profile characterizing smooth muscle spasm is preserved without loss).

[0275] This represents the sampled values ​​of the absolute amplitude sequence at the current and historical delay nodes.

[0276] Example 6: Continuing from the calculation example in Example 5, the current sampling time is set to... Given the system state:

[0277] Temporally aligned deep mixing intensity features ;

[0278] Shallow noise components ;

[0279] Let's assume the low-pass filter is a simple second-order mean filter. Its coefficient is set as , And the absolute amplitude after rectification at the previous sampling time is known. .

[0280] Perform differential stripping:

[0281] .

[0282] Perform full-wave rectification:

[0283] Take the absolute value: .

[0284] Perform low-pass smoothing to extract the envelope:

[0285] Perform convolution: ;

[0286] .

[0287] Output: The final calculated smooth muscle clonus envelope feature value is... This value represents the level of mechanical tension of uterine smooth muscle at the current physical time point.

[0288] In the aforementioned technology, after performing high-precision differential subtraction, cascaded full-wave rectification technology is used to uniformly convert bipolar fluctuations into an energy accumulation form. Subsequently, a low-pass filter matching a specific physiological frequency band is applied to forcibly suppress all high-frequency interference components and fit the external energy profile of the signal. This transforms the complex original optical fluctuations into a low-frequency envelope curve with a simple shape and clear characterization. This envelope feature eliminates interference from non-target physiological rhythms and directly and objectively reflects the slow evolution process of the mechanical tension of the target uterine smooth muscle, providing a pure data source for subsequent construction of analytical signals and accurate phase measurement.

[0289] Differences in individual tissue transmittance among patients, variations in probe attachment pressure, and even signal attenuation at different times within the same physiological cycle can all lead to significant drift in the absolute amplitude of the extracted envelope features. Relying on a fixed amplitude threshold to determine when smooth muscle is in relaxation is highly susceptible to missed or incorrect detections due to baseline drift. Therefore, this paper proposes reconstructing the smooth muscle clonic envelope features within a preset observation time window using an analytical domain to obtain the smooth muscle phase evolution trajectory.

[0290] Perform a Hilbert transform on the smooth muscle clonus envelope features to obtain the corresponding orthogonal imaginary part signal components;

[0291] Using the smooth muscle clonus envelope feature as the real part and the orthogonal imaginary signal component as the imaginary part, an analytic signal in complex form is constructed.

[0292] The instantaneous phase angle of the analytic signal is calculated, and the instantaneous phase angle at which a phase jump occurs is unwound to generate a continuous smooth muscle phase evolution trajectory.

[0293] Among them: the preset observation time window refers to a finite-length historical data extraction interval set in time-domain signal processing. Since extracting phase features requires examining the local periodicity of the signal, it is necessary to simultaneously call the current sampling time and its multiple continuous sampling points on the historical time axis.

[0294] Analytic domain reconstruction: The process of extending a time-domain signal that originally only has real number representation to a complex plane containing both real and imaginary parts through mathematical transformation.

[0295] Hilbert transform: a specific linear mathematical operator used to strictly apply Hilbert transform to all frequency components of the original signal. (i.e., negative) The phase shift does not change its original amplitude spectrum characteristics.

[0296] Orthogonal imaginary part signal component: The numerical sequence output after the original envelope feature is transformed by Hilbert, and this sequence is mathematically orthogonal to the original envelope feature.

[0297] An analytic signal is a sequence of complex functions without negative frequency components, consisting of the original real-domain signal as the real part and its Hilbert transform as the imaginary part.

[0298] Instantaneous phase angle: The angle between the complex vector corresponding to the analytic signal in the complex plane coordinate system and the positive direction of the real axis. Its range is usually limited by mathematical functions. or between.

[0299] Phase jump and unwinding processing: Due to the inherent range limitation of inverse trigonometric function calculations, when the actual physiological phase crosses... or 2 At the boundary, the calculated angle will abruptly revert to its previous state. Or 0 (i.e., a transition occurs). The unwinding process is based on the phase difference between adjacent sampling points, through automatic compensation of 2... An algorithm that converts the discontinuous phase of a package into a continuous cumulative phase that monotonically increases with physical time by multiples of the given number of times.

[0300] Smooth muscle phase evolution trajectory: A numerical sequence of phase angles generated after unwinding, representing the continuous changes in the state of uterine smooth muscle during multiple contraction-relaxation cycles.

[0301] Step 1: Perform Discrete Hilbert Transform calculation on the smooth muscle clonus envelope feature data segments accumulated within the preset observation time window to generate phase lag. The auxiliary signal. The specific formula is as follows:

[0302] ;

[0303] in: To be at the current sampling time The output quadrature imaginary part of the signal;

[0304] This represents the discrete Hilbert transform operator (which in engineering is usually represented as a generalized bandpass filter with specific coefficients).

[0305] The output of step five at the current sampling time The smooth muscle clonus envelope feature.

[0306] Step 2: Within the microprocessor, a complex data structure is created. The actually measured and extracted envelope features are set as real axis data, and the calculated orthogonal imaginary part signal components are set as imaginary axis data. The specific formula is as follows:

[0307] ;

[0308] in: To be at the current sampling time The constructed complex form of the analytic signal;

[0309] It is the basic imaginary unit in complex number operations (satisfying) );

[0310] This represents the real part of the analytic signal;

[0311] This is for analyzing the imaginary part of the signal.

[0312] Step 3: Based on the constructed analytic signal, calculate the argument of the complex variable in the complex plane at the current moment. The specific formula is as follows:

[0313] ;

[0314] in: To be at the current sampling time The calculated instantaneous phase angle (this original angle is constrained by the boundary of the inverse trigonometric function calculation);

[0315] It is the arctangent mathematical function.

[0316] Step 4: Perform differential calculations on the instantaneous phase angles along the continuous time axis, monitoring for any jumps exceeding the physical law threshold. When a phase fallback is detected, an absolute continuous phase is generated by superimposing compensation constants. The specific formula logic is as follows:

[0317] if ,but:

[0318] ;

[0319] in: This is the final output value of the smooth muscle phase evolution trajectory at the current moment after unwinding.

[0320] The instantaneous phase angle at the previous sampling time;

[0321] This is an integer-type cycle number compensation variable that is incremented and accumulated based on the number of detected transitions.

[0322] The instantaneous phase angle on a continuous time axis is differentially calculated to monitor whether a jump threshold exceeding the physical laws occurs.

[0323] When a positive periodic transition occurs, i.e., if Then positive compensation will be performed: ,and ;

[0324] When a reverse backoff jump occurs due to transient noise, i.e., if Then reverse compensation will be performed: ,and This ensures that the generated smooth muscle phase evolution trajectory remains continuous under any physical fluctuations.

[0325] Example 7: Following Example 6, the current sampling time is set to... Given:

[0326] Current moment smooth muscle clonus envelope characteristics ;

[0327] Instantaneous phase angle at the previous sampling time radians, current compensation variable .

[0328] The historical envelope data segment is input into the Hilbert filter, and after convolution, the filtered output value at the current time is obtained: assuming the calculated value is... .

[0329] Combine the real and imaginary parts: .

[0330] Calculate the argument using the arctangent function:

[0331] radian.

[0332] Calculate the phase difference between adjacent phases: .

[0333] Determine if a transition occurs: because (Right now This indicates that no abrupt change across quadrant cycles has occurred, and no further additions are needed. value.

[0334] The absolute continuous phase at that moment is:

[0335] radian.

[0336] Output: The final value of the smooth muscle phase evolution trajectory at the current moment is obtained. The radian value not only represents the specific phase angle currently in a cycle, but also records the historically accumulated state of the cycle's evolution.

[0337] In the aforementioned technology, an orthogonal analytic signal model is established by using Hilbert transform, transforming one-dimensional time-domain amplitude information into a two-dimensional vector in the complex plane. Then, through inverse trigonometric function operations and dewinding logic, the rotation angle of this vector is extracted as a single evaluation criterion, thereby realizing the dimensional leap of signal features from the amplitude domain to the phase domain. The phase parameter only reflects the evolution state of the physiological cycle and has robustness to absolute changes in signal strength and static baseline drift. The monotonic evolution trajectory of this reconstructed output provides a highly consistent mathematical comparison benchmark for subsequently setting strict diastolic window time nodes.

[0338] Because a single phase threshold determination lacks consideration of dynamic trends, simply relying on the phase angle entering a certain interval to determine the relaxation period can easily lead to false triggering of relaxation periods when faced with high-frequency noise spikes or smooth muscle in a rapid deformation phase, resulting in frequent jitter in phototherapy output commands. Therefore, this paper proposes a method to determine whether the sustained convergence of the smooth muscle phase evolution trajectory relative to a preset relaxation period reference phase meets preset relaxation window matching conditions, specifically including:

[0339] Calculate the instantaneous phase change rate of the smooth muscle phase evolution trajectory;

[0340] Determine whether the instantaneous phase at the current sampling time falls within the preset phase interval that characterizes the relaxation state of uterine smooth muscle, and determine whether the absolute value of the instantaneous phase change rate is less than the preset stability threshold.

[0341] If the above conditions are met in multiple consecutive sampling periods, then the preset slack window matching condition is satisfied.

[0342] Among them: preset diastolic reference phase: a pre-set standard phase angle center value that characterizes the mechanical relaxation state of uterine smooth muscle during the menstrual cycle.

[0343] Preset phase range: The allowable phase angle deviation range defined around the preset relaxation period reference phase. Falling into this range indicates that the smooth muscle is currently in a relaxation trend.

[0344] Instantaneous phase change rate: The difference between adjacent sampling periods in the phase evolution trajectory, which is physically equivalent to the instantaneous angular velocity of the phase and is used to measure the intensity of smooth muscle contraction or relaxation.

[0345] Stability threshold: A pre-defined constant scalar used to limit the upper limit of the absolute value of the instantaneous phase change rate in order to screen out the time points when the smooth muscle state tends to level off.

[0346] Sustained convergence: A state index that takes into account the instantaneous phase state and its first derivative (rate of change) and maintains it within a preset interval on the time axis.

[0347] Relaxation window matching conditions: a combination of logic and continuous numerical thresholds that determine whether the current time period is suitable for triggering photon energy power compensation intervention.

[0348] Step 1: Extract the smooth muscle phase evolution trajectory values ​​from the current sampling time and the previous sampling time, as output in Step 6, and calculate the current instantaneous phase change rate using first-order backward differencing. The specific formula is as follows:

[0349] ;

[0350] in: To be at the current sampling time The calculated instantaneous phase change rate;

[0351] The current sampling time The smooth muscle phase evolution trajectory value;

[0352] This represents the smooth muscle phase evolution trajectory value at the previous sampling time.

[0353] Step 2: Due to the output of Step 6 For a monotonically increasing absolute phase that has undergone unwinding, in order to determine its specific position within a single contraction-dipulation cycle, a modulo operation needs to be performed on it to map it back to the base cycle, and a dual-condition judgment needs to be made.

[0354] First, perform the periodic mapping operation:

[0355] ;

[0356] in: Map the current sampling time to The instantaneous phase angle of a single cycle within the base cycle;

[0357] For constants Modulo operator.

[0358] Then, determine whether the current state simultaneously satisfies the following set of logical conditions:

[0359] Condition 1 (Phase Interval Matching):

[0360] ;

[0361] Condition 2 (convergence of rate of change):

[0362] ;

[0363] in: The lower limit of a preset phase interval characterizing the diastolic state;

[0364] The upper limit of a preset phase interval characterizing the diastolic state;

[0365] It is an absolute value operator;

[0366] This is the preset stability threshold.

[0367] Define the single-step logic output at the current moment. If both conditions one and two above are true, then ;otherwise .

[0368] Step 3: To avoid false triggering caused by transient digital noise, perform a time window accumulation judgment on the single-step logic output sequence generated in the previous step. The specific formula is as follows:

[0369] ;

[0370] in: It is a continuously accumulated logical value;

[0371] The preset continuous sampling period threshold (i.e., the length of the preset observation time window);

[0372] This represents the single-step logic output value for the current and historical sampling times.

[0373] like If the condition is met, the preset expansion window matching condition is determined; otherwise, it is determined that the condition is not met.

[0374] Example 8: Following Example 7, the current sampling time is set to... Given:

[0375] Current phase trajectory 8 radians, the phase trajectory at the previous moment radian.

[0376] System preset parameters:

[0377] Lower limit of the preset diastolic range radians, upper limit radian;

[0378] Preset stability threshold radians / period;

[0379] The required number of consecutively satisfied periods The logic outputs of the first two cycles are known to be... , .

[0380] Find the rate of change:

[0381] radian.

[0382] Execution mapping and state determination:

[0383] Modulo mapping: radian;

[0384] Judgment condition one: , established;

[0385] Judgment condition two: , established;

[0386] The current logical output is obtained. .

[0387] Continuous convergence check:

[0388] Calculate the accumulated value: .

[0389] Output result: judgment Is it equal to : If the smooth muscle phase evolution trajectory at the current moment satisfies the preset relaxation window matching condition, the system status flag will be set to "True" and passed to the subsequent instruction generation module.

[0390] In the aforementioned technology, a high-confidence triggering and error-proofing mechanism is constructed by introducing a dual constraint logic of first derivative (rate of change) and time integration (multi-cycle continuous verification). This means that not only must the phase position meet the standard, but the phase movement speed must also be slow enough, and this stable state must be maintained for a certain duration on the time axis. This mechanism effectively eliminates transient phase fluctuation interference caused by the patient's breathing or minor tremors, ensuring that the diastolic window is only confirmed when the uterine smooth muscle enters a true and stable physical relaxation period, thus providing a safe and reliable premise for the targeted power output in step eight.

[0391] Traditional open-loop phototherapy devices use fixed power output. During the tetanic contraction phase of uterine smooth muscle, the dense muscle fibers cause a significant decrease in photon transmittance. At this time, continuous strong light radiation not only fails to effectively penetrate to the ischemic target area but also causes local tissue heat accumulation and thermal pain. During the smooth muscle relaxation phase, because the fixed power is not adaptively increased, it cannot provide sufficient photon flux to maximize the microvascular dilation effect, resulting in low intervention efficacy. Therefore, this paper proposes a method that calculates power compensation gain based on the smooth muscle phase evolution trajectory, generates and outputs targeted intervention luminescence commands, otherwise maintains the current intervention state, as detailed below:

[0392] Among them: Power compensation gain: On top of the base optical output power, an additional dynamic optical power value is added based on the specific relaxation depth of the uterine smooth muscle. This gain is variable and aims to dynamically adjust the amount of photons injected during relaxation.

[0393] Targeted intervention emission command: A digital control word or analog voltage signal generated by a microprocessor or control unit and sent to the underlying hardware emission driving circuit (such as an LED constant current driving chip). This command determines the actual optical radiation flux output by the light source device at the current physical moment.

[0394] Current intervention state: refers to the baseline operating mode maintained by the system when the relaxation window condition is not detected. It is usually set to a safety baseline low power output (used to maintain the light source required for basic optical physiological detection) or zero power output state.

[0395] Step 1: Read the single-step logic output conclusion generated in Step 7 (i.e., the Boolean value indicating whether the diastolic window matching condition is met). If the logic judgment is negative (False), the system executes the standby branch, directly assigning the set baseline operating power to the light emission instruction register to maintain the basic light emission state of the underlying hardware, and the operation of this cycle ends; if the logic judgment is positive (True), then proceed to the next calculation and execution branch.

[0396] Step 2: In the execution branch, extract the instantaneous phase angle of the current single cycle and calculate its mathematical deviation from the preset diastolic reference phase center point to quantify the current diastolic depth of the smooth muscle. The specific formula is as follows:

[0397] ;

[0398] in: To be at the current sampling time The calculated phase deviation;

[0399] It is an absolute value operator;

[0400] To be at the current sampling time The extracted single-cycle instantaneous phase angle (already mapped to in step seven) (within the basic cycle)

[0401] This is the preset nominal value of the center of the diastolic reference phase.

[0402] Step 3: Based on the obtained phase deviation, calculate the compensation power to be allocated at the current moment using a preset nonlinear mapping function (usually a quadratic decay model). This logic ensures that maximum compensation is applied when the relaxation is most complete (phase angle is closest to the center nominal value). The specific formula is as follows:

[0403] ;

[0404] in: To be at the current sampling time The generated power compensation gain;

[0405] The maximum power compensation gain constant preset for the system;

[0406] The phase deviation at the current moment;

[0407] It is the preset phase angle half-width constant of the slack window (i.e., half of the total phase angle span of the slack window).

[0408] Step 4: Linearly superimpose the calculated power compensation gain with the reference operating power to generate the final command value, and output it to the hardware light-emitting module through the bus interface. The specific formula is as follows:

[0409] ;

[0410] in: To be at the current sampling time The generated targeted intervention luminescence command;

[0411] The baseline operating power set for the system;

[0412] This is the calculated power compensation gain.

[0413] Example 9: Following Example 8, the current sampling time is set to... Given:

[0414] The logical conclusion is: the preset window opening matching condition is met (the execution branch is triggered).

[0415] Current instantaneous phase angle of a single cycle radian.

[0416] System preset parameters:

[0417] nominal value of the reference phase center during diastole radian;

[0418] half-width constant of phase angle of the slack window radian;

[0419] Maximum power compensation gain constant ;

[0420] Branch decision: If the logic is "true", the system enters the dynamic compensation gain calculation process.

[0421] Calculate the phase deviation: radian.

[0422] Find the power compensation gain:

[0423] ;

[0424] .

[0425] Generate light emission instructions:

[0426] .

[0427] Output result: The final calculated and generated value is The targeted intervention emission command is sent to the LED driver after digital-to-analog conversion, so that the emitter can accurately project the corresponding radiation power onto the human target area at the current moment.

[0428] In the aforementioned technology, a closed-loop feedback control mechanism is established to dynamically allocate light power based on the mechanical phase of smooth muscle. Through digital logic gating, power is allowed to rise only within a safe diastolic window. Combined with a secondary attenuation model, the output command is continuously and smoothly quantified based on the degree to which the phase angle approaches the target depth. This enables on-demand drug delivery in the field of physical phototherapy. On the one hand, maintaining a low baseline power during the non-intervention window eliminates the physical risks of thermal damage and increased spasm, reducing the overall energy consumption of the system. On the other hand, achieving targeted and precise injection of high-dose photons during diastole significantly improves the physical conversion efficiency of photobiological modulation (PBM) in improving local microcirculation.

[0429] A precise menstrual pain relief system based on red and near-infrared light includes:

[0430] The optical signal acquisition module is used to acquire the shallow reflected light intensity characteristics and deep mixed light intensity characteristics of the target area on the human abdomen at the current sampling time.

[0431] The active interference suppression module is used to spatially project the shallow reflected light intensity features based on the current spatial mapping weights to obtain the initial simulated noise features, and calculate the approximation residual between the deep mixed light intensity features and the initial simulated noise features; determine the optimization gradient based on the shallow reflected light intensity features and the approximation residual, and correct the current spatial mapping weights based on the optimization gradient; reconstruct the shallow reflected light intensity features based on the corrected spatial mapping weights, and output the shallow noise component representing macroscopic body motion interference;

[0432] The smooth muscle contraction feature extraction module is used to perform differential stripping on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle.

[0433] The phase trajectory analysis module is used to perform analytical domain reconstruction of the smooth muscle clonus envelope features within a preset observation time window to obtain the smooth muscle phase evolution trajectory.

[0434] The targeted intervention decision module is used to determine whether the continuous convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic reference phase meets the preset diastolic window matching condition; if yes, it calculates the power compensation gain based on the smooth muscle phase evolution trajectory, generates a targeted intervention luminescence command and outputs it; otherwise, it maintains the current intervention state.

[0435] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for precisely relieving menstrual cramps based on red and near-infrared light, characterized in that, Includes the following steps: Obtain the shallow reflected light intensity characteristics and deep mixed light intensity characteristics of the human abdominal target area at the current sampling time; Based on the current spatial mapping weights, the shallow reflected light intensity characteristics are spatially projected to obtain the initial simulated noise characteristics, and the approximation residual between the deep mixed light intensity characteristics and the initial simulated noise characteristics is calculated. The optimization gradient is determined based on the shallow reflection intensity characteristics and the approximation residual, and the current spatial mapping weights are corrected based on the optimization gradient. The shallow reflected light intensity characteristics are reconstructed and mapped based on the corrected spatial mapping weights, and the shallow noise component representing macroscopic body motion interference is output. Differential stripping is performed on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle. The smooth muscle clonus envelope features within the preset observation time window are reconstructed using the analytical domain to obtain the smooth muscle phase evolution trajectory; Determine whether the sustained convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic reference phase satisfies the preset diastolic window matching condition; If so, the power compensation gain is calculated based on the smooth muscle phase evolution trajectory, a targeted intervention luminescence command is generated and output; otherwise, the current intervention state is maintained.

2. The method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 1, characterized in that: The optimization gradient is determined based on the shallow reflection intensity characteristics and the approximation residual, and the preset spatial mapping weights are corrected based on the optimization gradient, including: Calculate the product of the shallow reflected light intensity characteristics and the approximation residual at the current sampling time to obtain the optimization gradient characterizing the direction of interference correlation; The spatial mapping weight adjustment amount is obtained by adjusting the gain of the optimization gradient through the convergence step size. The spatial mapping weight adjustment is accumulated and added to the current spatial mapping weight to obtain the corrected spatial mapping weight.

3. The method for precise relief of dysmenorrhea based on red and near-infrared light according to claim 2, characterized in that: The convergence step size is a dynamic convergence step size adjusted based on error energy negative feedback, specifically including: Extract the instantaneous average energy of the approximation residual within the preset observation time window, and use it as the residual energy measure coefficient to characterize the degree of prediction bias; The residual energy measurement coefficients are input into a preset monotonically increasing mapping function to calculate the step gain; Calculate the product of the step gain and the preset baseline step size to generate the final convergence step size.

4. The method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 3, characterized in that: The specific formula for the monotonically increasing mapping function is as follows: ； in, Step size gain; This is the preset maximum gain limit; The residual energy measurement coefficient; The preset energy threshold; The preset curve steepness coefficient, and .

5. The method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 1, characterized in that: The approximation residuals between the deep mixed light intensity characteristics and the initial simulated noise characteristics are calculated, specifically including: The shallow reflected light intensity characteristics at the current sampling time and its historical times are used to construct a multi-dimensional reference signal vector; Calculate the inner product of the multidimensional reference signal vector and the current spatial mapping weight vector to obtain the scalar projection value, which is used as the initial simulated noise feature; Time delay evaluation and timing alignment processing are performed on the deep mixed light intensity characteristics and the initial simulated noise characteristics; The instantaneous amplitude of the temporally aligned deep hybrid light intensity feature is subtracted from the instantaneous amplitude of the initial simulated noise feature to obtain a temporal difference sequence, which serves as the approximation residual.

6. The method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 1, characterized in that: The shallow reflected light intensity characteristics are reconstructed based on the corrected spatial mapping weights, outputting a shallow noise component characterizing macroscopic bodily motion interference, specifically including: The corrected spatial mapping weights are updated to the tap coefficients of the preset finite impulse response adaptive filter; The shallow reflected light intensity characteristics of the continuous time series are input into the finite impulse response adaptive filter after updating the tap coefficients for convolution filtering calculation, and the shallow noise component is output.

7. The method for precise relief of dysmenorrhea based on red and near-infrared light according to claim 1, characterized in that: Differential stripping is performed on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features characterizing the mechanical contraction state of the target smooth muscle, specifically including: After removing the shallow noise component from the deep mixed light intensity features, the denoised net smooth muscle signal sequence is obtained. The net smooth muscle signal sequence is subjected to full-wave rectification to obtain the absolute amplitude sequence of the signal; The absolute amplitude sequence is smoothed by using a low-pass filter with a preset cutoff frequency, and the slowly varying energy profile of the signal is extracted to obtain the smooth muscle clonus envelope feature.

8. The method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 1, characterized in that: The smooth muscle phase evolution trajectory is obtained by reconstructing the smooth muscle clonus envelope features within a preset observation time window using an analytical domain: Perform a Hilbert transform on the smooth muscle clonus envelope features to obtain the corresponding orthogonal imaginary part signal components; Using the smooth muscle clonus envelope feature as the real part and the orthogonal imaginary signal component as the imaginary part, an analytic signal in complex form is constructed. The instantaneous phase angle of the analytic signal is calculated, and the instantaneous phase angle at which a phase jump occurs is unwound to generate a continuous smooth muscle phase evolution trajectory.

9. A method for precisely relieving menstrual cramps based on red and near-infrared light according to claim 1, characterized in that: Determining whether the sustained convergence of the smooth muscle phase evolution trajectory relative to a preset diastolic baseline phase satisfies a preset diastolic window matching condition specifically includes: Calculate the instantaneous phase change rate of the smooth muscle phase evolution trajectory; Determine whether the instantaneous phase at the current sampling time falls within the preset phase interval that characterizes the relaxation state of uterine smooth muscle, and determine whether the absolute value of the instantaneous phase change rate is less than the preset stability threshold. If the above conditions are met in multiple consecutive sampling periods, then the preset slack window matching condition is satisfied.

10. A precise dysmenorrhea relief system based on red and near-infrared light, characterized in that, include: The optical signal acquisition module is used to acquire the shallow reflected light intensity characteristics and deep mixed light intensity characteristics of the target area on the human abdomen at the current sampling time. The active interference suppression module is used to spatially project the shallow reflected light intensity features based on the current spatial mapping weights to obtain the initial simulated noise features, and calculate the approximation residual between the deep mixed light intensity features and the initial simulated noise features; it determines the optimization gradient based on the shallow reflected light intensity features and the approximation residual, and corrects the current spatial mapping weights based on the optimization gradient. The shallow reflected light intensity characteristics are reconstructed and mapped based on the corrected spatial mapping weights, and the shallow noise component representing macroscopic body motion interference is output. The smooth muscle contraction feature extraction module is used to perform differential stripping on the deep mixed light intensity features and the shallow noise components to obtain the smooth muscle clonus envelope features that characterize the mechanical contraction state of the target smooth muscle. The phase trajectory analysis module is used to perform analytical domain reconstruction of the smooth muscle clonus envelope features within a preset observation time window to obtain the smooth muscle phase evolution trajectory. The targeted intervention decision module is used to determine whether the continuous convergence of the smooth muscle phase evolution trajectory relative to the preset diastolic reference phase meets the preset diastolic window matching condition; if yes, it calculates the power compensation gain based on the smooth muscle phase evolution trajectory, generates a targeted intervention luminescence command and outputs it; otherwise, it maintains the current intervention state.

Images