Enclosed space infrared induction head physiology monitoring system

Abstract

The invention relates to the technical field of physiological feature non-inductive monitoring and multi-modal sensor information fusion, in particular to a closed space infrared induction head physiology monitoring system which comprises a multi-modal detection acquisition module, an optical path dynamic compensation module, a thermal inertia separation processing module and a cross-modal signal self-healing module. According to the method, simple data stacking of traditional multiple sensors is abandoned, a strict hardware-level clock synchronization mechanism is established at the front end, and closed-loop optical path dynamic gain compensation is executed by combining a depth displacement vector obtained by time-of-flight distance measurement with a square inverse proportion attenuation law of illumination intensity, so that the accuracy of optical path dynamic gain compensation is improved. Motion artifact interference caused by spatial displacement is eliminated from an optical object base, thermal inertia physical differences between the environment in a closed cabin and human physiological activities are deeply excavated, and thermal noise baselines in an extremely severe environment are accurately stripped by constructing a time sequence dual-channel filter of a specific time window, so that the accuracy of thermal noise detection is improved. And high-fidelity extraction of the weak respiratory rhythm is realized.

Description

Technical Field

[0001] This invention relates to the field of non-invasive physiological feature monitoring and multimodal sensor information fusion technology, specifically to an infrared sensing head physiological monitoring system for enclosed spaces. Background Technology

[0002] With the development of intelligent technologies for enclosed spaces such as smart cockpits, hyperbaric oxygen chambers, and contactless sleep monitoring, non-contact physiological characteristic monitoring (such as heart rate and respiratory rhythm detection) is becoming a research hotspot in the industry due to its advantage of being wearable without any burden. Currently, mainstream non-contact monitoring systems mainly rely on near-infrared imaging technology based on photoplethysmography (rPPG) and far-infrared thermal imaging technology based on thermodynamic distribution.

[0003] However, existing non-contact physiological monitoring systems face the following insurmountable technical shortcomings in complex, enclosed space environments: 1. In environments with dynamic disturbances or enclosed spaces allowing free movement of the target, the target is highly susceptible to spatial displacement in the depth direction, such as leaning forward or backward. Because light intensity decreases dramatically with increasing propagation distance, this physical displacement causes severe amplitude modulation (i.e., motion artifacts) in the reflected light intensity received by the near-infrared camera. Existing technologies often employ purely software-based blind source separation or smoothing algorithms for post-processing digital filtering. These methods not only struggle to handle significant distance abrupt changes but also easily filter out genuine high-frequency physiological pulse characteristics, leading to severe distortion in heart rate extraction.

[0004] Second, far-infrared thermal imaging technology is highly susceptible to drastic temperature changes in enclosed spaces when extracting facial respiratory rhythms (e.g., vehicle windows heated by direct sunlight or air conditioning blowing directly on the face). The overall baseline drift of the ambient temperature often far exceeds the subtle temperature fluctuations generated by human respiratory airflow in the subnasal region. Existing systems lack effective mechanisms for separating large-scale environmental thermal noise from minute physiological thermal radiation, easily leading to respiratory characteristic signals being completely overwhelmed by the environmental background temperature drift.

[0005] Third, multimodal sensors in existing systems often operate in isolation, performing simple comparisons only at the final result level, lacking deep data coupling and complementary mechanisms at the underlying level. When a key sensor (such as a near-infrared component) experiences signal interruption due to extreme physical obstruction or violent target movement, the system can usually only output intermittent physiological parameters or directly trigger false alarm shutdown, failing to guarantee the rigid requirement for continuous monitoring data in high-level security scenarios.

[0006] Therefore, there is an urgent need for a highly robust monitoring scheme that can resist earthquakes from the physical level, is immune to environmental temperature drift, and has the ability to self-heal across modal signals. Summary of the Invention

[0007] To achieve the above objectives, the present invention provides the following technical solution: This invention provides an infrared sensing head physiological monitoring system for enclosed spaces, comprising: a multimodal detection and acquisition module for simultaneously acquiring three-dimensional depth displacement vectors, near-infrared photoelectric pulse wave signals, and far-infrared thermal radiation sequence of a target face within an enclosed space; Optical path dynamic compensation module: connected to the multimodal detection and acquisition module, used to use the three-dimensional depth displacement vector as a compensation reference to dynamically adjust the exposure parameters or digital gain of near-infrared detection, so as to output a reference pulse wave signal that eliminates motion artifacts; Thermal inertia separation processing module: For the far-infrared thermal radiation map sequence, based on the physical thermal inertia difference between the cabin environment background and the target's physiological activities, a time-series dual-channel filter is constructed to subtract and strip the low-frequency environmental thermal noise baseline from the total thermal radiation map, and extract the pure target respiratory rhythm signal. Cross-modal signal self-healing module: connected to the optical path dynamic compensation module and the thermal inertia separation processing module respectively, used to monitor the signal-to-noise ratio of the reference pulse wave signal in real time. When the signal-to-noise ratio is lower than the preset safety threshold, the phase features of the respiratory rhythm signal are extracted and substituted into the preset cardiopulmonary coupling rhythm model. Mathematical interpolation and curve smoothing are performed on the missing heart rate data segments to output continuous and uninterrupted physiological characteristic parameters.

[0008] Furthermore, the multimodal detection and acquisition module includes: The time-of-flight ranging sensor, a near-infrared camera module with a center wavelength of 850nm or 940nm, an uncooled infrared thermal imaging module, and a global clock generator; The global clock generator is electrically connected to the time-of-flight ranging sensor, the near-infrared camera assembly, and the uncooled infrared thermal imaging assembly, respectively, and is used to send hardware trigger pulses of the same frequency to the three components, so that the acquired three-dimensional depth displacement vector, near-infrared photoelectric pulse wave signal, and far-infrared thermal radiation map sequence maintain strict frame synchronization alignment on the time axis.

[0009] Furthermore, when acquiring the three-dimensional depth displacement vector, the multimodal detection and acquisition module specifically performs the following steps: The initial three-dimensional point cloud data of the target's face is acquired using the time-of-flight ranging sensor. Rigid feature regions of the face are extracted as reference anchor points, and their initial three-dimensional coordinates are recorded. : During continuous monitoring, the rigid feature region is tracked in real time at the current moment. 3D coordinates ; To eliminate high-frequency speckle noise from the ranging sensor, a sliding window smoothing strategy is used to calculate the current time. 3D depth displacement vector :

[0010] in, The preset number of sliding window frames, for The three-dimensional coordinates of the rigid feature region at that moment.

[0011] Furthermore, the optical path dynamic compensation module includes a depth projection unit and a gain calculation unit; The depth projection unit is used to project the three-dimensional depth displacement vector. Projecting the data onto the physical optical axis of the near-infrared camera assembly, a scalarized change in distance along the optical axis is obtained. ; The gain calculation unit is used to calculate the current moment based on the inverse square law of light intensity decay. Dynamic gain compensation coefficient The specific calculation formula is as follows:

[0012] in, The system presets the amplification gain for the target face at the initial reference time. Initial three-dimensional coordinates The initial absolute distance from the near-infrared camera component.

[0013] Furthermore, the optical path dynamic compensation module will adjust the dynamic gain compensation coefficient. The signal is sent in real time to the underlying automatic gain control circuit of the near-infrared camera component, or it is directly multiplied by the amplitude of the currently acquired original photoelectric pulse wave signal as a signal multiplier to output the reference pulse wave signal that eliminates spatial displacement modulation artifacts.

[0014] Furthermore, the thermal inertial separation processing module extracts the target respiratory rhythm signal by specifically including the following steps: The subnasal feature region of the target is located from the far-infrared thermal radiation image sequence, and the temporal variation sequence of the average temperature of all pixels in this region is extracted and recorded as the original thermal radiation sequence. ; Based on the high thermal inertia physical characteristics of a closed space environment, a large-scale sliding time window is constructed to extract the slowly changing baseline of environmental thermal noise. The calculation formula is as follows:

[0015] in, The preset background baseline time window length, and The value is greater than 3 times the target minimum effective respiratory cycle.

[0016] Furthermore, the original thermal radiation sequence is subtracted from the slowly varying baseline in the time domain to obtain the initial respiratory characteristic sequence after eliminating environmental thermal interference. :

[0017] The initial sequence of the respiratory features The input is fed into a preset physiological frequency bandpass filter, the passband frequency of which is limited to [specific value]. ,in and Corresponding to the lower and upper limits of the normal human respiratory rate, respectively, the filtered signal outputs a smooth and pure target respiratory rhythm signal.

[0018] Furthermore, when the cross-modal signal self-healing module monitors the signal-to-noise ratio of the reference pulse wave signal in real time, it specifically includes: Set a pulse wave quality assessment time window, extract the AC component amplitude and DC component amplitude of the reference pulse wave signal within the window in real time, and calculate the ratio of the two as the current signal quality index. The current signal quality index is compared with a preset safety threshold. If the current quality index is continuous... If the number of calculation cycles is lower than the safety threshold, it is determined that the near-infrared photoelectric pulse wave signal has been severely blocked or interfered with by violent motion. The system determines that data loss has occurred and triggers the cross-modal interpolation compensation mechanism.

[0019] Furthermore, the pre-defined cardiopulmonary coupled rhythm model used in the cross-modal interpolation compensation mechanism includes the following specific mathematical interpolation and curve smoothing steps: Extract the average heart rate within a historical normal time window prior to the occurrence of data loss. and synchronized average respiratory rate Calculate the individualized cardiopulmonary baseline coupling coefficient :

[0020] During the time segment where data is missing, extract the instantaneous frequency of the pure target respiratory rhythm signal continuously output by the thermal inertial separation processing module. and respiratory phase characteristics .

[0021] Furthermore, based on the physiological modulation characteristics of respiratory sinus arrhythmia, the cardiopulmonary reference coupling coefficient is utilized. Estimate the instantaneous heart rate node sequence within the missing time segment. :

[0022] in, It is a periodic regulatory function that follows the alternating phases of inhalation and exhalation. Using the actual effective heart rate data before and after the missing data segment as boundary anchors, the estimated instantaneous heart rate node sequence is... As an intermediate node, a cubic spline interpolation algorithm is used for curve fitting, outputting a continuous, smooth physiological characteristic parameter curve without obvious step changes.

[0023] Beneficial effects This invention abandons the simple data stacking of traditional multi-sensor systems, establishes a strict hardware-level clock synchronization mechanism at the front end, and combines the depth displacement vector obtained by time-of-flight ranging with the inverse square law of light intensity attenuation to perform closed-loop optical path dynamic gain compensation, eliminating motion artifact interference caused by spatial displacement from the optical substrate. At the same time, it deeply explores the thermal inertial physical differences between the enclosed cabin environment and human physiological activities, and accurately strips the thermal noise baseline under extreme conditions by constructing a time-series dual-channel filter for a specific time window, achieving high-fidelity extraction of weak respiratory rhythms. It establishes a cross-modal mathematical interpolation self-healing model using the "cardiopulmonary coupling effect" in medical physiology. When the near-infrared pulse wave signal is briefly lost due to extreme physical interference, the system can immediately call the far-infrared respiratory rhythm for instantaneous curve fitting and smooth reconstruction. Thus, without adding any additional hardware redundancy, it achieves all-weather, continuous, uninterrupted, and high-confidence imperceptible monitoring of physiological characteristics in a complex, turbulent, and temperature-changing enclosed space. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a flowchart illustrating the operational logic of the optical path dynamic compensation module of the present invention. Figure 3 This is a data processing flowchart of the thermal inertia separation processing module of the present invention; Figure 4 This is a flowchart of the mathematical interpolation compensation process for the cross-modal signal self-healing module of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings: Example: like Figures 1-4 As shown, an infrared-sensing head physiological monitoring system for enclosed spaces includes: Multimodal detection and acquisition module: used to simultaneously acquire three-dimensional depth displacement vectors, near-infrared photoelectric pulse wave signals, and far-infrared thermal radiation sequence of the target face in an enclosed space; Optical path dynamic compensation module: connected to the multimodal detection and acquisition module, used to use the three-dimensional depth displacement vector as a compensation reference to dynamically adjust the exposure parameters or digital gain of near-infrared detection, so as to output a reference pulse wave signal that eliminates motion artifacts; Thermal inertia separation processing module: For the far-infrared thermal radiation map sequence, based on the physical thermal inertia difference between the cabin environment background and the target's physiological activities, a time-series dual-channel filter is constructed to subtract and strip the low-frequency environmental thermal noise baseline from the total thermal radiation map, and extract the pure target respiratory rhythm signal. Cross-modal signal self-healing module: connected to the optical path dynamic compensation module and the thermal inertia separation processing module respectively, used to monitor the signal-to-noise ratio of the reference pulse wave signal in real time. When the signal-to-noise ratio is lower than the preset safety threshold, the phase features of the respiratory rhythm signal are extracted and substituted into the preset cardiopulmonary coupling rhythm model. Mathematical interpolation and curve smoothing are performed on the missing heart rate data segments to output continuous and uninterrupted physiological characteristic parameters.

[0028] Furthermore, the specific operation procedure of the multimodal detection and acquisition module is as follows: In specific embodiments of the present invention, the internal architecture, collaborative working mechanism, and data processing flow of the multimodal detection and acquisition module are described in detail. As the front-end sensing center of the entire closed-space physiological monitoring system, the multimodal detection and acquisition module's main responsibility is to non-contactly and synchronously acquire the spatial topological state, subcutaneous microvascular optical characteristics, and facial thermodynamic distribution of the target object in complex and variable closed physical environments (such as vehicle cabins, hyperbaric oxygen chambers, or sleep chambers).

[0029] To achieve the above objectives, the module is highly integrated in its hardware architecture, primarily consisting of a Time-of-Flight (ToF) sensor, a near-infrared camera assembly, an uncooled infrared thermal imaging assembly, and a global clock generator. During physical deployment, these sensor components are rigidly fixed to the same optical base to ensure a high degree of overlap in their fields of view (FOV) and a fixed extrinsic parameter matrix correlation in the physical space coordinate system. Specifically, the near-infrared camera assembly preferably employs a near-infrared light source with a center wavelength of 850nm or 940nm and a matching CMOS image sensor. This specific wavelength band is chosen because near-infrared light in this band can effectively penetrate the epidermal tissue of the human face and reach the dermis layer rich in capillaries, thereby sensitively capturing minute changes in local blood volume caused by the periodic beating of the heart (i.e., photoplethysmography pulse wave), completely unaffected by sudden changes in visible light (such as flickering streetlights or cabin lighting switches) within an enclosed space. Meanwhile, the infrared thermal imaging component uses an uncooled microbolometer to balance the compactness of the device with a high sensitivity response to weak thermal radiation from the human body surface (usually in the far-infrared band of 8 to 14 micrometers).

[0030] In existing technologies, simply piecing together multiple sensors often leads to uncontrollable phase differences in the data over time, causing subsequent cross-modal fusion algorithms to fail. To completely overcome this deficiency, this invention creatively introduces a strict timing synchronization mechanism based on underlying hardware into the multimodal detection and acquisition module. The global clock generator is connected to the external trigger pins of the time-of-flight ranging sensor, the near-infrared camera component, and the uncooled infrared thermal imaging component via hardware-level electrical wiring. After system startup, the global clock generator synchronously sends hardware trigger pulses with microsecond-level precision to all the aforementioned sensors at a preset fixed frequency (e.g., 30Hz or 60Hz). This means that the three sets of sensors simultaneously expose and sample at the instant they receive the same rising edge signal, thereby ensuring absolute frame-level alignment of the output 3D depth point cloud, near-infrared image frame, and far-infrared thermal map frame over time. This physical-level synchronization provides an unbreakable data alignment foundation for subsequent signal spatial compensation and cross-modal self-healing.

[0031] After completing high-precision synchronous data acquisition, the multimodal detection and acquisition module needs to perform real-time calculation of spatial displacement. When the target object is in a closed space, it will inevitably produce movements such as leaning forward, tilting backward, or turning its head. To accurately quantify this spatial displacement, the initial three-dimensional point cloud data of the target's face is first acquired using the time-of-flight ranging sensor. To avoid misjudgment of local depth caused by the target opening its mouth to speak or changes in facial expressions (such as the bulging of the cheekbones due to smiling), a built-in three-dimensional morphological analysis algorithm is used to actively optimize and extract the "rigid feature region" of the face as a reference anchor point in the point cloud. This rigid feature region is preferably the upper part of the bridge of the nose or the brow bone region, where human skeletal deformation is minimal. The system records the three-dimensional coordinate center point of this rigid feature region in the initial monitoring state, and records it as the initial three-dimensional coordinate. .

[0032] As monitoring continues, the system tracks and updates the rigid feature region in real time across consecutive time slices. 3D coordinates However, in confined spaces (especially vehicle cabins or oxygen chambers containing numerous metallic reflective surfaces), the light pulses emitted by ToF sensors are highly susceptible to multipath reflection interference, resulting in high-frequency speckle noise in the output point cloud data. If the displacement is obtained by directly subtracting the coordinates of two adjacent frames, this high-frequency noise will be rapidly amplified, causing oscillations in the subsequent dynamic compensation system.

[0033] To eliminate the inherent high-frequency speckle noise in ranging, this module introduces a time-series sliding window smoothing strategy with memory effect before data output. Specifically, the system does not rely solely on single-frame data to calculate displacement, but rather extracts the current time frame... and its preceding consecutive The coordinate data of each historical moment is smoothed within a time window. The system calculates the current moment. 3D depth displacement vector The specific mathematical model is as follows:

[0034] in, This represents the preset number of frames for the sliding window (usually set between 5 and 15 frames based on the frequency characteristics of unconscious micro-movements in daily life). for The system calculates the absolute three-dimensional coordinates of the rigid feature region at any given time. Through this moving average mechanism and the joint calculation by subtracting the reference coordinates, the system can completely filter out the physical-level noise from the sensor and output an extremely smooth three-dimensional depth displacement vector sequence that truly reflects the macroscopic displacement of the target object's torso and head.

[0035] Finally, the multimodal detection and acquisition module will synchronize the near-infrared video stream, the far-infrared thermal radiation sequence, and the three-dimensional depth displacement vector with high signal-to-noise ratio after noise-resistant calculation, all through hardware-level synchronization. The data is then packaged and transmitted to the optical path dynamic compensation module.

[0036] Furthermore, the specific implementation process of the optical path dynamic compensation module is as follows: This module achieves data-level interconnection with the aforementioned multimodal detection and acquisition module, aiming to completely eliminate the photoelectric signal amplitude modulation interference (i.e. motion artifacts) introduced by the random spatial displacement of the target object in the enclosed space, starting from the underlying logic of physical optics, thereby reconstructing a pure pulse wave signal with extremely high baseline stability.

[0037] In terms of system operation logic, the optical path dynamic compensation module is internally divided into a depth projection unit and a gain calculation unit.

[0038] First, those skilled in the art should understand that the three-dimensional depth displacement vector output by the front-end module... This includes spatial motion components of the target in the world coordinate system, encompassing three degrees of freedom: X (horizontal), Y (vertical), and Z (depth). However, in optical imaging systems, the core variable that truly causes the drastic attenuation of light from the light source reaching the face and reflecting back to the image sensor (i.e., the entire optical path) is merely the depth displacement along the optical axis of the camera component. Therefore, the primary task of the depth projection unit is to reduce the dimension of this three-dimensional vector. Specifically, the system pre-calibrates the physical optical axis normal vector of the near-infrared camera component, and the projection unit uses spatial analytical geometric operations such as vector dot product to reduce the dimension of the three-dimensional depth displacement vector. The optical path is precisely projected onto the physical optical axis of the near-infrared camera component, thereby filtering out the lateral and longitudinal translation components that have minimal impact on light intensity attenuation, and extracting the scalarized change in optical axis distance that truly dominates the optical path change. .like A positive value indicates that the target is moving away from the camera; a negative value indicates that the target is moving closer to the camera.

[0039] After obtaining accurate The data is then fed into the gain calculation unit. This unit is the core of this module. It breaks away from the conventional approach of relying on pure black-box algorithms for software waveform smoothing in existing technologies, and instead directly introduces the classic photometric physical law—the inverse square law of light intensity attenuation. According to this physical law, the total light intensity of a point light source (or a near-infrared light-emitting diode with a specific divergence angle) illuminating the target surface and reflected back to the receiver is strictly inversely proportional to the square of the optical path distance.

[0040] Based on this physical mechanism, the gain calculation unit constructs a dynamic gain inverse compensation model. The system assumes that the initial absolute distance between the target face and the near-infrared camera component at the initial reference time (i.e., the time before significant displacement) is... In order to obtain an image with the best signal-to-noise ratio, the system presets a baseline amplification gain. When the target undergoes depth displacement, in order to ensure that the amplitude of the AC signal of the effective photoplethysmography (PPG) wave received on the target surface of the near-infrared camera component remains constant, the system must apply an amplification factor that is completely inversely proportional to the light intensity attenuation. Therefore, this unit calculates the current moment in real time. Dynamic gain compensation coefficient The exact calculation formula is expressed as follows:

[0041] When the target is far from the camera When the optical signal attenuates exponentially on a quadratic basis, the system instantly calculates a compensation coefficient that amplifies exponentially on a quadratic basis. ;vice versa.

[0042] deriving the dynamic gain compensation coefficient Subsequently, this embodiment provides two execution paths for applying it to the original signal to adapt to different levels of hardware computing power platforms: In the first preferred execution path (hardware-level closed-loop compensation), the optical path dynamic compensation module uses the underlying driver interface to transfer the coefficient. The signal is sent directly and in real-time to the automatic gain control circuitry or exposure control register within the near-infrared camera assembly. This means that the system dynamically adjusts the sensor's light-sensing gain at the analog circuit level before the photoelectric signal has even completed analog-to-digital conversion (ADC). The significant advantage of this closed-loop control is that it fundamentally prevents the interruption of highlight signals caused by a target suddenly approaching the lens, and also avoids the weak signal being drowned out by background noise due to a target moving away, thus ensuring signal quality from the source of optical sampling.

[0043] In the second execution path (digital back-end multiplier compensation), if the hardware sensor does not support high-frequency low-level register erasure and rewriting, the system will use this coefficient. As a digital signal multiplier, it performs direct digital multiplication with the instantaneous amplitude of the original photoelectric pulse wave extracted from the near-infrared image sequence by the back-end algorithm.

[0044] Regardless of the path employed, through the close collaboration between depth projection and inverse solving of physical laws, the optical path dynamic compensation module successfully and completely separates three-dimensional spatial variables from the fragile physiological optical signal, outputting a reference pulse wave signal to the system that eliminates spatial displacement modulation artifacts. The baseline of this reference pulse wave signal exhibits extremely high smoothness; even when the target object undergoes violent and irregular back-and-forth swaying within a closed space, the morphology of its AC pulsation components remains highly intact.

[0045] Furthermore, the specific implementation process of the thermal inertia separation processing module is as follows: This module interacts directly with the uncooled infrared thermal imaging component in the aforementioned multimodal detection and acquisition module. Its core purpose is to accurately extract pure human respiratory rhythm signals under harsh conditions such as severe background thermal radiation interference in enclosed spaces (e.g., direct airflow from vehicle air conditioners, temperature and pressure changes in hyperbaric oxygen chambers).

[0046] As is known to those skilled in the art, in a confined space, changes in ambient temperature and the temperature characteristics of human physiological activities exhibit drastically different frequency responses over time. This is attributed in physics to the difference in "thermal inertia." The interior walls, seats, or overall air volume of a vehicle cabin possess enormous heat capacity and thermal inertia, and their temperature rise or fall is a slow, low-frequency gradual process. Conversely, the alternation frequency of human breathing airflow is relatively high (typically 10 to 30 times per minute), and the heat capacity of the airflow is extremely small. The heat exchange formed in the facial region under the nose exhibits typical high-frequency periodic fluctuations. This module is based on this difference in thermal inertia in physics to construct a time-domain dual-channel separation algorithm.

[0047] In the specific data processing flow, the system first receives the far-infrared thermal radiation map sequence. Using a built-in facial feature point recognition algorithm, the system accurately locates the target's subnasal region (i.e., the area directly exposed to respiratory airflow) within the thermal map. The system calculates the average temperature of all pixels within this region in each frame and arranges them along the time axis to form a raw thermal radiation sequence that includes environmental thermal noise and minute fluctuations in respiratory activity. .

[0048] In order to hide in The system completely removes the slow baseline drift by introducing a large-scale sliding time window mechanism to extract the environmental background baseline representing high thermal inertia. The system calculates this slowly changing baseline in real time. The specific mathematical calculation model is as follows:

[0049] In this model, This represents the preset background baseline time window length. The time window length here... These are not arbitrarily set empirical values, but rather strictly constrained by the physiological Nyquist sampling theorem and boundary conditions. This is to ensure the accuracy of the calculated baseline. It only reflects slow changes in ambient temperature and does not mistakenly "average" out the actual respiratory peaks; the system is strictly limited. The time span represented must be greater than three times the "minimum effective respiratory cycle" that the target physiologically might occur. For example, if the system sets the minimum effective respiratory rate for humans to be 8 breaths per minute (i.e., a cycle of 7.5 seconds), then the time span of the sliding window must be forcibly set to at least 22.5 seconds. This rigid binding of parameters based on physiological cycles ensures that the mathematical filter will not overfit when faced with various abnormal respiratory rhythms.

[0050] Accurately extract the environmental thermal noise baseline Next, the system enters the time-domain difference stage. The thermal inertial separation processing module processes the original thermal radiation sequence at the current moment. Corresponding slowly changing baseline Point-by-point temporal difference subtraction is performed to obtain the initial sequence of respiratory characteristics after eliminating environmental thermal interference. :

[0051] After the above subtraction calculation, the temperature curve, which was originally severely tilted upward or downward due to the influence of ambient temperature rise or fall, was instantly "flattened" to the horizontal baseline of zero mean, completely eliminating the low-frequency thermal drift caused by air conditioning or sun exposure.

[0052] Finally, in order to eliminate the electrical noise of the far-infrared sensor itself and the high-frequency sudden thermal interference caused by the movement of people around, the system will "flatten" the initial sequence of respiratory characteristics. The input is fed into a preset physiological frequency bandpass filter. The passband frequency of the bandpass filter is strictly limited to [specific range]. Between. Among them, These correspond to the lower limit (e.g., 0.15Hz, approximately 9 breaths / min) and upper limit (e.g., 0.6Hz, approximately 36 breaths / min) of normal human respiratory rate, respectively. All noise exceeding this frequency band is suppressed. Ultimately, the module outputs an extremely smooth, pure target respiratory rhythm signal in the form of a standard sine wave to the end of the system.

[0053] Furthermore, the specific implementation process of the cross-modal signal self-healing module is as follows: This module performs deep data handshakes with the optical path dynamic compensation module and the thermal inertia separation processing module, respectively. It aims to achieve seamless and smooth splicing of data streams by utilizing the underlying coupling physiological mechanism of the cardiopulmonary system when irreversible data breaks occur in the near-infrared photoelectric pulse wave due to extreme physical obstruction or violent movement, ensuring that the system outputs 100% continuous physiological characteristic parameters to the terminal.

[0054] During the normal operation phase of the system, the cross-modal signal self-healing module initially functions as a "sentinel." The system sets a length of... The system defines a pulse wave quality assessment time window. Within this window, the system extracts the AC component (reflecting blood volume pulsation amplitude) and DC component (reflecting background illumination and static tissue absorption) of the reference pulse wave signal output from the previous stage in real time, and calculates the ratio (AC / DC), which is defined as the current signal quality index. The system then compares the currently calculated SQI with a preset safety threshold. A comparison is performed. If the target object experiences only slight shaking, the preceding "optical path dynamic compensation module" is sufficient to maintain the SQI above the threshold; however, if the current SQI is continuously... If the calculation cycle falls below the safety threshold (e.g., the user turns their head sharply away from the camera for several seconds), the system will determine that the near-infrared signal has been severely interfered with, resulting in a loss of valid data.

[0055] At this critical moment, the system does not trigger any alarms or health diagnosis logic, but immediately initiates a cross-modal interpolation compensation mechanism. The physical and biological basis of this mechanism lies in the well-known phenomenon of "respiratory sinus arrhythmia," which states that the human heartbeat and respiration are strictly coupled under the regulation of the medullary cardiovascular center: the heart rate slightly increases compensatorily during inspiration and slightly decreases compensatorily during expiration. This module transforms this medical physiological phenomenon into a purely computer mathematical interpolation model.

[0056] The specific implementation steps are as follows: First, the system backtracks to the historical data cache and extracts the high-confidence average heart rate within a normal historical time window before the data loss occurred (e.g., 60 seconds before the loss occurred). and the average respiratory rate extracted synchronously within the same historical period. Based on this, the system calculates the individualized "cardiopulmonary baseline coupling coefficient" of the target object under its current physiological state. ,Right now .

[0057] Subsequently, for the time segment where the current heart rate data is missing, the system cross-modally retrieves the pure target respiratory rhythm signal, which is stably output by the preceding "thermal inertia separation processing module" even in extremely harsh thermal environments. The system extracts the instantaneous frequency of this respiratory rhythm within the missing segment. and respiratory phase characteristics (in (Indicates whether the current phase is inspiratory or expiratory).

[0058] Using the above parameters, the system substitutes them into a preset cardiopulmonary coupling rhythm model to estimate the instantaneous heart rate node sequence within the missing time segment. The specific estimation formula is as follows:

[0059] in, The system has a pre-defined periodic adjustment function (e.g., a small sinusoidal modulation term synchronized with the respiratory phase) to simulate the fine-tuning of heart rate caused by inhalation and exhalation. Through the above estimation, the system actually uses the current real respiratory fluctuation waveform to proportionally map and reconstruct the corresponding heartbeat waveform profile.

[0060] Finally, to ensure that there are no abrupt abrupt changes between the reconstructed data segments and the actual valid data segments before and after them, the system introduces a boundary smoothing mechanism. The system uses the actual valid heart rate data points before and after the missing data segments as fixed "boundary anchor points" to smooth the estimated instantaneous heart rate node sequence. As an "intermediate free node," a cubic spline interpolation algorithm is used for curve fitting globally. Since cubic spline interpolation can ensure that the curve achieves second derivative continuity at the connection point (i.e., positional continuity, slope continuity, and curvature continuity), the system ultimately outputs an extremely smooth physiological characteristic parameter curve without any breaks or spikes.

[0061] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A closed-space infrared sensing head physiological monitoring system, characterized in that, include: Multimodal detection and acquisition module: used to simultaneously acquire three-dimensional depth displacement vectors, near-infrared photoelectric pulse wave signals, and far-infrared thermal radiation sequence of the target face in an enclosed space; Optical path dynamic compensation module: connected to the multimodal detection and acquisition module, used to use the three-dimensional depth displacement vector as a compensation reference to dynamically adjust the exposure parameters or digital gain of near-infrared detection, so as to output a reference pulse wave signal that eliminates motion artifacts; Thermal inertia separation processing module: For the far-infrared thermal radiation map sequence, based on the physical thermal inertia difference between the cabin environment background and the target's physiological activities, a time-series dual-channel filter is constructed to subtract and strip the low-frequency environmental thermal noise baseline from the total thermal radiation map, and extract the pure target respiratory rhythm signal. Cross-modal signal self-healing module: connected to the optical path dynamic compensation module and the thermal inertia separation processing module respectively, used to monitor the signal-to-noise ratio of the reference pulse wave signal in real time. When the signal-to-noise ratio is lower than the preset safety threshold, the phase features of the respiratory rhythm signal are extracted and substituted into the preset cardiopulmonary coupling rhythm model. Mathematical interpolation and curve smoothing are performed on the missing heart rate data segments to output continuous and uninterrupted physiological characteristic parameters.

2. The closed-space infrared sensing head physiological monitoring system according to claim 1, characterized in that, The multimodal detection and acquisition module includes: The time-of-flight ranging sensor, a near-infrared camera module with a center wavelength of 850nm or 940nm, an uncooled infrared thermal imaging module, and a global clock generator; The global clock generator is electrically connected to the time-of-flight ranging sensor, the near-infrared camera assembly, and the uncooled infrared thermal imaging assembly, respectively, and is used to send hardware trigger pulses of the same frequency to the three components, so that the acquired three-dimensional depth displacement vector, near-infrared photoelectric pulse wave signal, and far-infrared thermal radiation map sequence maintain strict frame synchronization alignment on the time axis.

3. The closed-space infrared sensing head physiological monitoring system according to claim 2, characterized in that, When acquiring the three-dimensional depth displacement vector, the multimodal detection and acquisition module specifically performs the following steps: The initial three-dimensional point cloud data of the target's face is acquired using the time-of-flight ranging sensor. Rigid feature regions of the face are extracted as reference anchor points, and their initial three-dimensional coordinates are recorded. : During continuous monitoring, the rigid feature region is tracked in real time at the current moment. 3D coordinates ; To eliminate high-frequency speckle noise from the ranging sensor, a sliding window smoothing strategy is used to calculate the current time. 3D depth displacement vector : in, The preset number of sliding window frames, for The three-dimensional coordinates of the rigid feature region at that moment.

4. The closed-space infrared sensing head physiological monitoring system according to claim 3, characterized in that, The optical path dynamic compensation module includes a depth projection unit and a gain calculation unit; The depth projection unit is used to project the three-dimensional depth displacement vector. Projecting the data onto the physical optical axis of the near-infrared camera assembly, a scalarized change in distance along the optical axis is obtained. ; The gain calculation unit is used to calculate the current moment based on the inverse square law of light intensity decay. Dynamic gain compensation coefficient The specific calculation formula is as follows: in, The system presets the amplification gain for the target face at the initial reference time. Initial three-dimensional coordinates The initial absolute distance from the near-infrared camera component.

5. The closed-space infrared sensing head physiological monitoring system according to claim 4, characterized in that, The optical path dynamic compensation module will adjust the dynamic gain compensation coefficient. The signal is sent in real time to the underlying automatic gain control circuit of the near-infrared camera component, or it is directly multiplied by the amplitude of the currently acquired original photoelectric pulse wave signal as a signal multiplier to output the reference pulse wave signal that eliminates spatial displacement modulation artifacts.

6. The infrared sensing head physiological monitoring system for a closed space according to claim 5, characterized in that, When the thermal inertial separation processing module extracts the target respiratory rhythm signal, it specifically includes the following steps: The subnasal feature region of the target is located from the far-infrared thermal radiation image sequence, and the temporal variation sequence of the average temperature of all pixels in this region is extracted and recorded as the original thermal radiation sequence. ; Based on the high thermal inertia physical characteristics of a closed space environment, a large-scale sliding time window is constructed to extract the slowly changing baseline of environmental thermal noise. The calculation formula is as follows: in, The preset background baseline time window length, and The value is greater than 3 times the target minimum effective respiratory cycle.

7. The closed-space infrared sensing head physiological monitoring system according to claim 6, characterized in that, The original thermal radiation sequence is subtracted from the slowly varying baseline in the time domain to obtain the initial respiratory characteristic sequence after eliminating environmental thermal interference. : The initial sequence of the respiratory features The input is fed into a preset physiological frequency bandpass filter, the passband frequency of which is limited to [specific value]. ,in and Corresponding to the lower and upper limits of the normal human respiratory rate, respectively, the filtered signal outputs a smooth and pure target respiratory rhythm signal.

8. The closed-space infrared sensing head physiological monitoring system according to claim 7, characterized in that, The cross-modal signal self-healing module, when monitoring the signal-to-noise ratio of the reference pulse wave signal in real time, specifically includes: Set a pulse wave quality assessment time window, extract the AC component amplitude and DC component amplitude of the reference pulse wave signal within the window in real time, and calculate the ratio of the two as the current signal quality index. The current signal quality index is compared with a preset safety threshold. If the current quality index is continuous... If the number of calculation cycles is lower than the safety threshold, it is determined that the near-infrared photoelectric pulse wave signal has been severely blocked or interfered with by violent motion. The system determines that data loss has occurred and triggers the cross-modal interpolation compensation mechanism.

9. The closed-space infrared sensing head physiological monitoring system according to claim 8, characterized in that, The pre-defined cardiopulmonary coupled rhythm model used in the cross-modal interpolation compensation mechanism includes the following specific mathematical interpolation and curve smoothing steps: Extract the average heart rate within a historical normal time window prior to the occurrence of data loss. and synchronized average respiratory rate Calculate the individualized cardiopulmonary baseline coupling coefficient : During the time segment where data is missing, extract the instantaneous frequency of the pure target respiratory rhythm signal continuously output by the thermal inertial separation processing module. and respiratory phase characteristics .

10. A closed-space infrared sensing head physiological monitoring system according to claim 9, characterized in that, Based on the physiological modulation characteristics of respiratory sinus arrhythmia, the cardiopulmonary reference coupling coefficient is utilized. Estimate the instantaneous heart rate node sequence within the missing time segment. : in, It is a periodic regulatory function that follows the alternating phases of inhalation and exhalation. Using the actual effective heart rate data before and after the missing data segment as boundary anchors, the estimated instantaneous heart rate node sequence is... As an intermediate node, a cubic spline interpolation algorithm is used for curve fitting, outputting a continuous, smooth physiological characteristic parameter curve without obvious step changes.

Images