A multi-parameter wearable smart optical fiber sensing system

Abstract

The application discloses a kind of multi-parameter wearable smart optical fiber sensing systems, the system includes: fluorescent sensing patch, for detecting target by fluorescence reaction, including biosensor, biosensor is used to react with target and generate fluorescence signal;Data glove, for real-time monitoring finger joint bending state and realizing gesture recognition, including glove and the multiple fiber bending sensors of joint part of being attached to glove, fiber bending sensor reflects finger bending degree in real time by light intensity change, and exports multidimensional light intensity signal;Wearable fiber detection meter is connected with fluorescent sensing patch, data glove, for collecting and analyzing fluorescence signal and bending signal.The multi-parameter wearable optical fiber intelligent sensing system has multi-channel, full-automatic intelligent detection and analysis capability, can simultaneously provide fluorescence, gesture bending and environmental temperature and humidity and other multi-physical and chemical parameters Real-time monitoring and analysis services provide powerful technical support for the application of each field.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to a multi-parameter wearable intelligent fiber optic sensing system. Background Technology

[0002] Gestures are one of the most common forms of communication between humans. For example, sign language is widely used by people with speech impairments. Automotive gesture recognition can be used as a human-computer interaction (HCI) solution, offering a more natural and efficient solution for information exchange between users and devices compared to traditional methods such as mice and / or keyboards. Under the influence of the global pandemic, it plays a crucial role in practical applications, making wearable devices capable of contactless HCI and pathogen detection extremely important.

[0003] Currently, gesture recognition mainly includes vision-based or sensor-based gesture recognition. Vision-based gesture recognition is easily limited by equipment and environmental conditions. For example, during shooting, if the camera light is too bright or too dark, the quality of the captured image will be affected, thus impacting the accuracy of recognition. However, sensor-based recognition technology is unaffected by ambient light and background color; it features stable data acquisition and simple signal processing, overcoming the shortcomings of the aforementioned vision-based recognition technologies.

[0004] Smart gloves equipped with flexible sensors are a viable solution for hand gesture and sign language recognition due to their real-time responsiveness and portability. Furthermore, applications including robot-assisted surgery and rehabilitation, human-computer interaction, and gesture recognition for communication and entertainment have been proven. Various soft sensor technologies have been used in the design of smart gloves, including resistive and capacitive sensors, such as flexure sensors for detecting joint flexion and / or piezoelectric sensors and electromyography (EMG) for detecting finger segmental movements.

[0005] However, gesture recognition systems based on electronic sensors are susceptible to electromagnetic interference, while those based on chemical sensors are complex and expensive. Compared to electronic sensors, optical sensors offer advantages such as ease of large-scale multiplexing and resistance to electromagnetic interference. Various methods have been employed to achieve bending sensing, with fiber optic sensors facilitating continuous multi-point curvature measurement. Researchers have proposed a series of design methods through extensive study of fiber optic sensors, enabling the sensing and detection of bending degree, bending curvature, and, based on these, gestures and movements. However, existing wearable fiber optic sensing and detection devices lack multi-functional sensing capabilities, cannot achieve multi-functional detection, cannot recognize repetitive and continuous gestures, and have a limited range of detected gestures, necessitating improvements in detection speed and accuracy. Summary of the Invention

[0006] In view of the above problems, the present invention provides a multi-parameter wearable intelligent fiber optic sensing system, which realizes efficient and accurate biosensing detection, pathogenic microorganism detection, etc., and enables human-computer interaction with medical equipment through data gloves, providing convenient and accurate technical support for on-site detection and human-computer interaction between medical personnel and related equipment.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] This invention provides a multi-parameter wearable intelligent fiber optic sensing system, comprising:

[0009] A fluorescent sensing patch for detecting a target analyte by means of a fluorescence reaction, including a biosensor that reacts with the target analyte and generates a fluorescence signal;

[0010] A data glove for real-time monitoring of finger joint bending and gesture recognition includes a glove and multiple fiber optic bending sensors attached to the joints of the glove. The fiber optic bending sensors reflect the degree of finger bending in real time through changes in light intensity and output multidimensional light intensity signals.

[0011] A wearable fiber optic detector, connected to the fluorescent sensing patch and the data glove, is used to collect and analyze fluorescence signals and bending signals.

[0012] In one embodiment of the present invention, the fluorescent sensing patch further includes:

[0013] An excitation fiber is used to transmit excitation light to the biosensor;

[0014] A receiving optical fiber is used to capture the fluorescence signal generated by the biosensor;

[0015] The patch substrate, made of flexible material, integrates the biosensor and optical fiber.

[0016] In one embodiment of the present invention, the biosensor is made by freeze-drying a detection reagent onto the surface of filter paper. When the target substance enters the reaction chamber of the biosensor, ultraviolet light excites and generates a fluorescence signal. The fluorescence signal is transmitted through a receiving optical fiber and captured by a miniature camera and converted into a fluorescence image. The change in the image intensity is linearly correlated with the concentration of the target substance.

[0017] In one embodiment of the present invention, the fiber optic bending sensor is attached to the metacarpophalangeal joints and proximal interphalangeal joints of the glove using a transparent silicone rubber material.

[0018] In one embodiment of the present invention, the wrist of the glove is also provided with an optical fiber bending sensor to improve the recognition accuracy of repetitive gestures.

[0019] In one embodiment of the present invention, the gesture recognition employs the following algorithm flow:

[0020] The acquired fiber optic images are segmented to obtain the output light intensity signal of each fiber optic bend sensor;

[0021] The forward differential method is used to extract the real-time bending characteristics of the fiber optic bending sensor.

[0022] Eliminate data noise using threshold denoising techniques;

[0023] Detect bending features to identify gesture switching points, and segment the start and end times and gesture switching points as key points;

[0024] The output light intensity signal at the intermediate moment between adjacent key points is retained as key information of the gesture pattern and compressed accordingly.

[0025] The compressed gesture pattern is transmitted to the gesture recognition module based on the radial basis function network model for gesture recognition.

[0026] In one embodiment of the present invention, the gesture recognition specifically includes:

[0027] When the system is operational, the wearable fiber optic sensing system extracts the R-channel pixel intensity from images acquired by a small camera as the output signal Ic of the fiber optic bending sensor; the initial output signal I0 is when the fiber optic bending sensor is not bent; to quantize the output signals of the fiber optic bending sensor array to the same scale, the following equation is used to normalize the output signals:

[0028] I=k·θ+I b

[0029] Where θ is the bending angle of the fiber optic bending sensor, I = 1 – Ic / I0, k = -k′ / I0, k′ is the linear conversion coefficient from the output power to the output light intensity signal of the fiber optic bending sensor, and Ib is the deviation of the linear fitting result;

[0030] I(n) represents the total light intensity signal obtained from the nth sampling of the fiber optic bending sensor array, I j (n) represents the light intensity signal obtained from the nth sampling of the j-th fiber optic bend sensor in the fiber optic bend sensor array, I j(n+1) represents the light intensity signal obtained by the (n+1)th sampling of the j-th fiber bending sensor in the fiber bending sensor array; the gesture switching characteristic curve ΔI(n) is used to represent the degree of change of finger joint movement. If the total light intensity signal change ΔI(n) of the fiber bending sensor array is greater than the threshold, the gesture is considered to have started to change; if ΔI(n) is less than the threshold, the gesture is considered to remain unchanged; during the gesture change, the finger joint angle vector changes with the hand shape, and ΔI(n) changes accordingly; each gesture mode is processed using formula (2);

[0031]

[0032] Where N is the total number of fiber optic bending sensors in the data glove. Forward difference is performed on the multidimensional light intensity signal output by the data glove to extract the bending features of the fiber optic bending sensors in real time. Absolute value operation is performed to eliminate the sign information in the data, so that the amplitude values ​​of the feature waveforms are all positive, simplifying the feature extraction and analysis process. An aggregation function is used to reduce the dimensionality of the multidimensional bending data through linear addition.

[0033] ΔI(n) is processed by threshold denoising:

[0034]

[0035] Where λ is the set light intensity threshold for determining whether a gesture has changed, the gesture switching point is found by detecting the maximum value, and the start and end times and the gesture switching point are used as key points for continuous gesture segmentation; the data between every two key points is an oversampled identical data point; the area near each gesture switching point represents the gesture switching process, ΔI h If (n) remains unchanged, it indicates a stable hand gesture maintenance process;

[0036] The light intensity vector output by the fiber optic bending sensor at the midpoint between adjacent key points is used as the key information of the gesture pattern, and data compression is performed based on this.

[0037] The compressed gesture pattern is transmitted to the gesture recognition module based on the radial basis function network model, and gesture recognition is performed using the trained model.

[0038] In one embodiment of the present invention, the optical fiber probe of the fluorescent sensing patch is chemically etched to remove part of the cladding to enhance the excitation light intensity. The specific etching method is as follows: the optical fiber is immersed in a 1:1 mixture of n-hexane and acetone for 5 seconds, and then cleaned, dried and the etching effect is fixed.

[0039] In one embodiment of the present invention, a wireless communication module is further included. The wireless communication module is connected to the wearable fiber optic detector and is used to transmit the detection data of the wearable fiber optic detector to a remote server and / or personal device in real time.

[0040] In one embodiment of the present invention, the target is a pathogenic microorganism.

[0041] The beneficial effects achieved by this invention are as follows:

[0042] This invention provides a multi-parameter wearable intelligent fiber optic sensing system. By integrating a fluorescent sensing patch, it achieves highly sensitive detection of pathogenic microorganisms in the air; through a data glove, it enables wireless human-machine interaction between the testing personnel and medical equipment; and through the wearable system design, it meets the needs of rapid on-site testing. This system not only achieves fluorescence detection of pathogenic microorganisms in the air, but also innovatively proposes gesture segmentation, compression, and recognition algorithms, thereby achieving efficient recognition of real-time dynamic gestures. Furthermore, the system can also achieve remote data visualization and sharing. In summary, this multi-parameter wearable fiber optic intelligent sensing system possesses multi-channel, fully automated intelligent detection and analysis capabilities, and can simultaneously provide real-time monitoring and analysis services for multiple physicochemical parameters such as fluorescence, bending, and environmental temperature and humidity, providing strong technical support for applications in various fields. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0044] Figure 1 (a) is a schematic diagram of a multi-parameter wearable intelligent fiber optic sensing system; (b) is a schematic diagram of a wearable fluorescent sensing patch structure; and (c) is a schematic diagram of a wearable fiber optic detector structure after etching.

[0045] Figure 2 This is a flowchart of the workflow of a real-time gesture recognition system.

[0046] Figure 3 (a) is a schematic diagram of the plastic optical fiber before and after etching; (b) is a schematic diagram of the wearable fiber optic fluorescence sensing system.

[0047] Figure 4 This is a picture of the actual data gloves.

[0048] Figure 5 (a) is a physical diagram of the multi-parameter sensing system; (b) is a diagram of the system worn; and (c) is a diagram of remote data display via mobile phone.

[0049] Figure 6 This is a schematic diagram of the visualization interface of a multi-parameter wearable intelligent fiber optic sensing system.

[0050] Figure 7 (a) is a schematic diagram of fiber optic fluorescence image; (b) is a schematic diagram of camera parameter optimization.

[0051] Figure 8 The figures show the performance test results of the wearable fluorescent sensor patch, where (a) is a comparison of the test results of the device and the commercial device microplate reader; and (b) is a stability test figure.

[0052] Figure 9 This is a dynamic test diagram of pathogenic bacteria.

[0053] Figure 10 (a) in the image is a gesture image; (b) is gesture data.

[0054] Figure 11 The diagram shows the static gesture recognition results based on RBFNN, where (a) is a schematic diagram of the confusion matrix of the training set output results; (b) is a schematic diagram of the confusion matrix of the validation set output results; and (c) is a schematic diagram of the gesture recognition accuracy.

[0055] Figure 12 This is a schematic diagram of the confusion matrix with the largest eigenvalue of the gesture switching point.

[0056] Figure 13 The diagram shows the real-time gesture recognition data processing procedure, where (a) is a data glove sampling data diagram that changes with gestures; (b) is a data segmentation signal diagram; (c) is an extracted gesture feature data diagram; and (d) is a gesture recognition diagram based on RBFNN.

[0057] Figure 14 The diagram shows the data processing procedure for real-time gesture recognition with repetitive features. (a) is a sampled data diagram of the glove that changes with the gesture; (b) is a data segmentation signal diagram; (c) is a diagram of the extracted gesture feature data; and (d) is a gesture recognition diagram based on RBFNN. Detailed Implementation

[0058] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the invention, are intended to cover non-exclusive inclusion.

[0060] In the description of the embodiments of this invention, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this invention, "multiple" means two or more, unless otherwise explicitly defined.

[0061] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least some embodiments of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0062] In the medical field, for example, when healthcare workers wear protective suits equipped with fluorescent detection capabilities, the fluorescent sensors respond rapidly upon encountering potentially pathogenic microorganisms. These sensors identify the microorganisms using unique fluorescent markers and issue an alarm, providing immediate safety warnings. This technology not only enhances medical safety but also provides doctors with an intuitive and rapid method for identifying biological risk factors. In the interaction between doctors and medical equipment, the application of gesture recognition technology further improves work efficiency. Through precise gesture recognition, doctors can easily operate and query equipment without being distracted by complex interfaces. Especially in emergency situations, gesture recognition technology becomes a crucial bridge for communication between doctors and equipment, ensuring rapid response and accurate execution of medical procedures.

[0063] Intelligent fiber optic sensing systems, with their superior multi-signal detection capabilities, including diverse detection such as fluorescence and bending, and their highly integrated and automated recognition features, demonstrate broad application prospects. To further expand their application scope, this invention provides a multi-parameter wearable intelligent fiber optic sensing system that integrates wearable fluorescence detection technology and gesture recognition technology. This integrated multi-parameter wearable intelligent fiber optic sensing system not only achieves accurate identification of biological risk factors but also provides precise analysis of continuous gestures, potentially driving the intelligent development of wearable devices, biomedicine, environmental monitoring, and other fields, bringing people a safer, more convenient, and more efficient life experience.

[0064] like Figure 1As shown in (a), a multi-parameter wearable intelligent fiber optic sensing system includes a data glove, a fluorescent sensing patch, and a wearable fiber optic detector. This system can capture the light intensity information output by the data glove in real time, thereby achieving high-precision recognition of hand gestures. Simultaneously, by monitoring the fluorescence response of a biosensor under specific excitation light, the system effectively detects the presence of pathogens in exposed environments. Utilizing the fiber optic light intensity images captured by the wearable fiber optic detector, the system can provide comprehensive intelligent detection and analysis services with multiple channels and multiple physicochemical parameters. Particularly noteworthy is that all analysis results can be transmitted in real time to mobile devices or personal servers via wireless communication technology, greatly enhancing the convenience and real-time performance of data processing. This system can not only detect pathogenic microorganisms such as bacteria and viruses in the air, but also achieve dynamic, efficient, and intelligent recognition of hand gestures, thereby realizing wireless human-computer interaction. The system also enables remote data visualization and sharing.

[0065] like Figure 1 As shown in (b), the fluorescence sensing patch includes a biosensor, an excitation fiber, a receiving fiber, and a patch substrate. The biosensor reacts with the target analyte and generates a fluorescence signal. The excitation fiber transmits excitation light to the biosensor, and the receiving fiber captures the fluorescence signal generated by the biosensor under the action of the target analyte. The patch substrate is made of a flexible material and integrates the biosensor and the fiber. The biosensor can react with the pathogenic microorganisms being detected, generating fluorescence signals through various principles, which are then captured by the fiber. The fluorescence sensing patch is provided with an external light source. The fluorescence signal acquired by the fiber is focused by filters, collimating lenses, etc., and a camera or similar device captures the fluorescence image for processing and analysis. The biosensor is made by freeze-drying specific detection reagents onto the surface of filter paper. When small droplets containing pathogenic bacteria in the environment enter the reaction chamber of the biosensor through the micropores of the patch, ultraviolet light irradiates the reaction chamber of the biosensor via the excitation fiber. The target analyte reacts with the biosensor to generate a fluorescence signal, which is transmitted through an optical fiber array and finally captured by a miniature camera of a fiber optic fluorometer and converted into a fluorescence image. The change in image intensity is directly related to the concentration of the target analyte, providing a reliable basis for pathogen detection.

[0066] like Figure 1As shown in (c), the data glove design is based on the tight integration of fiber optic bend sensors with the glove body. The data glove includes: multiple fiber optic bend sensors, bonded to the metacarpophalangeal and proximal interphalangeal joints of the glove via transparent silicone rubber material; and an additional fiber optic bend sensor at the wrist to improve the accuracy of repetitive gesture recognition. The sensors reflect the degree of finger bending in real time through changes in light intensity and output multidimensional light intensity signals. Soft, transparent silicone rubber material is used to precisely bond the fiber optic bend sensors to the joints on the glove, ensuring that the sensors can respond in real time to bending changes caused by flexion or extension of the finger joints. Each sensor independently monitors the bending state of the metacarpophalangeal or proximal interphalangeal joints of five fingers and reflects the degree of finger bending through changes in light intensity. Furthermore, the additional fiber optic bend sensor at the wrist further improves the accuracy of repetitive gesture recognition.

[0067] like Figure 2 As shown, the workflow of the sign language recognition system mainly includes three core stages: gesture segmentation, gesture sequence compression, and gesture recognition. In the data preprocessing stage, the system first performs fine segmentation on the acquired fiber optic images to obtain the output light intensity signal of each fiber optic bending sensor. Subsequently, the system uses the forward difference method to extract the real-time bending features of the fiber optic bending sensors and eliminates potential data noise through threshold denoising technology.

[0068] To accurately segment continuous gesture data, the system detects bending features to identify gesture transition points and uses the start and end times along with these transition points as key points for segmentation. To improve the efficiency and accuracy of gesture recognition, the system introduces an efficient data compression method that retains the output light intensity signal at the intermediate time between adjacent key points as crucial information for the gesture pattern. Finally, the compressed gesture pattern is transmitted to the RBFNN gesture recognition module for accurate recognition. The specific process is as follows:

[0069] When the system is operational, the wearable fiber optic sensing system extracts the R-channel pixel intensity from images acquired by a small camera as the output signal Ic of the fiber optic bending sensor; the initial output signal I0 is when the fiber optic bending sensor is not bent; to quantize the output signals of the fiber optic bending sensor array to the same scale, the following equation is used to normalize the output signals:

[0070] I=k·θ+I b

[0071] Where θ is the bending angle of the fiber optic bending sensor, I = 1 – Ic / I0, k = -k′ / I0, k′ is the linear conversion coefficient from the output power to the output light intensity signal of the fiber optic bending sensor, and Ib is the deviation of the linear fitting result;

[0072] I(n) represents the total light intensity signal obtained from the nth sampling of the fiber optic bending sensor array, I j (n) represents the light intensity signal obtained from the nth sampling of the j-th fiber optic bend sensor in the fiber optic bend sensor array, I j (n+1) represents the light intensity signal obtained by the (n+1)th sampling of the j-th fiber bending sensor in the fiber bending sensor array; the gesture switching characteristic curve ΔI(n) is used to represent the degree of change of finger joint movement. If the total light intensity signal change ΔI(n) of the fiber bending sensor array is greater than the threshold, the gesture is considered to have started to change; if ΔI(n) is less than the threshold, the gesture is considered to remain unchanged; during the gesture change, the finger joint angle vector changes with the hand shape, and ΔI(n) changes accordingly; each gesture mode is processed using formula (2);

[0073]

[0074] Where N is the total number of fiber optic bending sensors in the data glove. Forward difference is performed on the multidimensional light intensity signal output by the data glove to extract the bending features of the fiber optic bending sensors in real time. Absolute value operation is performed to eliminate the sign information in the data, so that the amplitude values ​​of the feature waveforms are all positive, simplifying the feature extraction and analysis process. An aggregation function is used to reduce the dimensionality of the multidimensional bending data through linear addition.

[0075] ΔI(n) is processed by threshold denoising:

[0076]

[0077] Where λ is the set light intensity threshold for determining whether a gesture has changed, the gesture switching point is found by detecting the maximum value, and the start and end times and the gesture switching point are used as key points for continuous gesture segmentation; the data between every two key points is an oversampled identical data point; the area near each gesture switching point represents the gesture switching process, ΔI h If (n) remains unchanged, it indicates a stable hand gesture maintenance process.

[0078] The light intensity vector output from the fiber optic bending sensor at the midpoint between adjacent key points is used as key information for the gesture pattern, and data compression is performed based on this. Finally, the compressed gesture pattern is transmitted to the RBFNN gesture recognition module, where a trained model is used for accurate gesture recognition.

[0079] To enhance excitation light intensity and improve detection sensitivity and accuracy, the fiber optic probe is etched to remove part of the cladding. Chemical etching of plastic optical fibers offers advantages such as simplicity and low cost; a 1:1 solution of n-hexane and acetone can be used to etch plastic optical fiber sensing probes.

[0080] The specific steps are as follows: Immerse the portion of the plastic optical fiber to be etched into the solution, ensuring full contact for uniform etching. Then, immerse the plastic optical fiber in the solution for 5 seconds to achieve the desired etching depth. After immersion, immediately remove the plastic optical fiber and thoroughly rinse its surface with clean water to remove any residual chemical solvents or etching agents. During rinsing, ensure sufficient water flow and time to avoid any chemical residue. Finally, place the cleaned plastic optical fiber in an oven at 55°C for 2 hours to dry. This process helps ensure the plastic optical fiber is completely dry and fixes the etching effect, thus completing the entire etching process. Side images of the plastic optical fiber before and after etching are shown below. Figure 3 As shown in (a) in the figure.

[0081] like Figure 3 As shown in (b), the wearable fiber optic fluorescence sensing system includes plastic optical fibers, a wearable fluorescence sensing patch, and a wearable fiber optic intensity demodulator. The wearable fluorescence sensing patch is made by freeze-drying a biosensor onto filter paper in a reaction chamber. Each plastic optical fiber sensing probe is made of four excitation fibers and three receiving fibers spaced apart and placed in the reaction chamber. The 0.5 mm diameter fiber is used as the excitation fiber to transmit the LED ultraviolet light signal and excite the fluorescence signal of the biosensor. The 0.25 mm diameter fiber is used as the receiving fiber to transmit the fluorescence signal to the miniature camera of the fluorometer for fluorescence signal acquisition.

[0082] A data glove is made by attaching fiber optic bend sensors to a knitted glove for continuous hand motion detection. The correspondence between sensor numbers and finger joints is as follows: Figure 4 As shown.

[0083] like Figure 5 As shown, by integrating a fluorescent patch and a data glove simultaneously into a wearable fiber optic intensity demodulator, the fluorescent sensing patch is attached to the surface of the clothing for detecting pathogens in ambient air. This realizes a multi-parameter wearable fiber optic intelligent sensing system for simultaneous detection of fluorescence and bending signals.

[0084] like Figure 6 As shown, a web application (wireless communication module) was designed to achieve remote visualization of data from a multi-parameter wearable fiber optic intelligent sensing system. When the system is working, the data detected by the system can be transmitted to a remote server or personal device via the wireless communication module, and the detection results can be displayed in real time.

[0085] Exposure time t b The selected s times are 1s, 2s, 6s, and 10s, with Iso values ​​chosen from [100, 200, 300, 400, 500, 600, 700, 800, 900, 1000]. The fiber optic images acquired under these parameters are shown below. Figure 7As shown in (a) above, the extracted average pixel intensity is as follows: Figure 7 As shown in (b) above, H represents the high-intensity fluorescence reagent, and L represents the low-intensity reagent. The bar chart represents the average pixel difference between the two reagents under corresponding camera parameters. It can be seen that at iso=500, t b The average pixel intensity is highest at 10s, at which point the fiber optic fluorometer has the highest sensitivity. Therefore, when iso = 500, t... b =10s was selected as the camera parameter setting for subsequent experiments.

[0086] like Figure 8 As shown, the performance of the wearable fluorescent sensing patch was tested. To evaluate the system's fluorescence detection performance, FITC was diluted to six different concentrations (10 pM, 100 pM, 1 nM, 10 nM, 100 nM, and 1 μM) for testing, and compared with a commercial microplate reader at excitation and emission wavelengths of 485 nm and 525 nm, respectively. Figure 8 As shown in (a), the horizontal axis represents the fluorescence test results of the ELISA reader, and the vertical axis represents the fluorescence intensity measured by the device. The test results of the device and the ELISA reader (SynergyH1) show a good linear relationship, demonstrating excellent performance similar to the ELISA reader. High-concentration (1 μM) and low-concentration (1 nM) fluorescence signals were measured at 15-second intervals for 1 hour, and the change in fluorescence intensity over time is shown in Figure (a). Figure 8 As shown in (b) in the figure, it exhibits good stability.

[0087] like Figure 9 As shown, based on the designed wearable fiber optic fluorescence sensing system, 10 4 Real-time detection of Salmonella at CFU / ml was performed. The dynamic testing function of the designed detection system was verified; the fluorescence intensity gradually increased with increasing reaction time, and the reaction rate differed between different concentrations, demonstrating strong distinguishability.

[0088] This invention utilizes a fabricated data glove for systematic data acquisition. For example... Figure 10 As shown in (a), 26 English letter gestures from "A" to "Z" and one stationary gesture "`" were selected as the research subjects. To ensure the rigor and completeness of the data, each gesture was repeated 5 times, and 60 samples were collected in each repetition, ultimately forming a detailed dataset containing 8100 samples. The collected data are as follows: Figure 10 As shown in (b) in the figure. The horizontal axis represents different gestures, the vertical axis represents the output response of the multi-channel sensor corresponding to each gesture, and the color intensity represents the light intensity output by the sensor.

[0089] Of the 8100 gesture data samples collected, there were 5670 training set samples and 2430 test set samples. The confusion matrix for the recognition results of all 27 gesture categories in the training set is as follows: Figure 11 As shown in (a) of the figure, the horizontal axis represents the actual gesture labels, and the vertical axis represents the gesture results output by the model. Data with blue backgrounds represents the number of samples correctly recognized for each gesture, while data with gray backgrounds represents the number of samples misclassified as other gestures. For example, if the actual gesture is "A", all 211 training samples were correctly recognized. Of the 200 training samples for gesture "B", 196 were correctly recognized, 1 was misclassified as "A", and 3 were misclassified as "C". Similarly, Figure 11 In the diagram, (b) represents the confusion matrix of the gesture recognition results on the validation set. The training and validation recognition accuracy for each gesture is shown in the figure. Figure 11 As shown in (c) of the diagram. For the training set, the recognition accuracy for the eight gestures "A", "E", "N", "O", "Q", "R", "V", and "W" was 100%. The recognition accuracy for the seven gestures "J", "M", "P", "S", "T", "U", and "`" was above 99%. The recognition accuracy for the eight gestures "B", "D", "G", "I", "K", "L", "X", and "Y" was between 96.4% and 99%. The recognition accuracy for the four gestures "C", "F", "H", and "Z" was between 93.1% and 95%. Among them, the recognition accuracy for "Z" was the lowest, at 93.1%. For the validation set, the recognition accuracy of 13 gestures, including "A", "B", "E", "MR", "T", "V", "W", and "Y", was 100%. The recognition accuracy of 9 gestures, including "D", "FH", "JL", "S", and "U", ranged from 95.6% to 99%. The recognition accuracy of 3 gestures, including "C", "I", and "X", ranged from 91.7% to 93.3%. The recognition accuracy of gesture "Z" was the lowest, at 87.8%.

[0090] Overall, the training accuracy for the 27 gestures was 98.1%, and the test accuracy was 97.8%. This means that the RBFNN model correctly recognized approximately 98.13% of the gesture data in the training dataset. This is a relatively high accuracy, indicating that the model learned effective features from the data during training and classified these features well. The test accuracy of 97.8% indicates that the model correctly recognized approximately 97.78% of the gesture data in the test dataset. Although the test accuracy is slightly lower than the training accuracy, it is still a high value, indicating that the model maintains good performance even on unseen data. Considering both training and test accuracies, the RBFNN model demonstrates good generalization ability in the gesture recognition task, effectively learning from training data and applying it to test data.

[0091] To select a suitable threshold, the data from switching between all 27 different gestures were processed, resulting in 27*26=702 gesture switching feature curves. The 702 maxima were then extracted as the corresponding gesture switching feature values. The confusion matrix for different gesture switching feature values ​​is shown below. Figure 12 As shown in the figure, the horizontal axis represents the initial gesture, and the vertical axis represents the final gesture after the change. The data points corresponding to the horizontal and vertical axes are the gesture switching feature values ​​from the initial gesture to the final gesture. For example, the feature value for switching from gesture "A" to gesture "B" is 0.98, and the feature value for switching to "C" is 0.8. Experimental results show that when switching between the 27 target gestures, the minimum switching feature value between any two different gestures is 0.3. Therefore, 0.25 is selected as the threshold, with a margin of 0.05.

[0092] The gesture segmentation threshold and static gesture recognition model have been determined, and the real-time gesture recognition algorithm proposed in this invention will be verified.

[0093] like Figure 13 As shown, the real-time gesture recognition process is divided into four modules: gesture data acquisition, continuous gesture data segmentation, data compression, and gesture recognition. First, "GESTURE RECOGNITION" is selected as the target gesture for data acquisition, and the data is as follows: Figure 13 As shown in (a), the output data of the ten fiber optic bending sensors on the data glove corresponding to the finger joints changes in real time as the gesture changes. Figure 13 As shown in (b) above, continuous gesture data segmentation is used to segment continuous gestures, with a segmentation time of 8.3 ms. Figure 13 As shown in (c), compression of continuous gestures is achieved by using data from the intermediate moments of the gesture switching point. Finally, as... Figure 13 As shown in (d), the compressed gesture data is input into the RBFNN model for gesture recognition, and the recognition result is output within 11ms. In summary, this system achieves real-time recognition of the "GESTURERECOGNITION" continuous gesture within 19.3ms.

[0094] In practical gesture recognition, repetitive gestures are inevitably encountered. To recognize gestures with repetitive features, a fiber optic bending sensor was added to the wrist of the data glove. When a gesture is required, the wrist bends. By detecting the output of the wrist sensor, gestures containing repetitive features are recognized. The gesture "I am a bookkeeper" was selected as the target gesture for data collection, and the data is as follows: Figure 14As shown in (a), the output data from ten fiber optic bending sensors on the data glove corresponding to the finger joints changes in real time as the gesture changes. Channel 11 represents the wrist sensor. Figure 14 As shown in (b), without adding a wrist sensor, the switching feature points of repetitive gestures cannot be detected. By incorporating the wrist sensor output data from channel 11 into the gesture feature curve processing, the switching feature points of repetitive gestures "O", "K", and "E" can be detected. The proposed method was used to segment continuous gestures, with a segmentation time of 8.3 ms. Figure 14 As shown in (c), compression of continuous gestures is achieved by using data from the intermediate moments of the gesture switching point. Finally, as... Figure 14 As shown in (d), the compressed gesture data is input into the RBFNN model for gesture recognition, and the recognition result is output within 10ms. In summary, this system achieves real-time recognition of the repetitive gesture "I am a bookkeeper" within 18.3ms.

[0095] The continuous gesture recognition algorithm based on a flexible fiber optic bending sensor proposed in this invention achieves an accuracy of 97.8% when recognizing 26 English letters and the space gesture, significantly outperforming similar sensors, which are mostly limited to static gesture recognition. This algorithm not only provides new insights for continuous gesture recognition research but also offers strong technical support for practical applications due to its high accuracy and fast recognition speed. Its wearable system has a simple structure, low maintenance costs, and ensures a good user experience. Furthermore, the flexibility and sensitivity of the flexible fiber optic sensor, combined with forward differential, threshold segmentation, and the RBFNN algorithm, effectively handles complex gestures, demonstrating unique advantages in specific applications.

[0096] In summary, this invention provides a comprehensive multi-parameter wearable fiber optic intelligent sensing system, integrating key components such as a fluorescent sensing patch, a data glove based on a fiber optic bend sensor, and a wearable fiber optic detector. This system not only achieves fluorescent detection of pathogens in the air but also innovatively proposes gesture segmentation, compression, and recognition algorithms, thereby achieving efficient recognition of real-time gestures. To enable remote visualization and sharing of data, a wireless communication module is also included. During system operation, all detection data can be transmitted in real time to a remote server or personal device via the wireless communication module, ensuring that users can access the latest detection results anytime, anywhere. Regarding the performance of the fluorescent sensing patch, after optimizing camera parameters and testing with FITC, the results show a good linear relationship with commercial ELISA readers, with R... 2=0.9823, and exhibits excellent stability. Simultaneously, the system successfully achieved real-time dynamic detection of Salmonella. In gesture recognition, the system not only achieved static recognition of 27 types of gesture data with an accuracy rate of 97.78%, but also accurately recognized real-time continuous gestures containing repetitive features, with a processing time of only 18.3ms, demonstrating extremely high real-time performance and accuracy. Overall, this multi-parameter wearable fiber optic intelligent sensing system possesses multi-channel, fully automated intelligent detection and analysis capabilities, simultaneously providing real-time monitoring and analysis services for multiple physicochemical parameters such as fluorescence, bending, and environmental temperature and humidity, providing strong technical support for applications in various fields.

[0097] Although the invention has been described with reference to preferred embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, the technical features mentioned in the various embodiments can be combined in any manner as long as there is no structural conflict. The invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A multi-parameter wearable intelligent fiber optic sensing system, characterized in that, include: A fluorescent sensing patch for detecting a target analyte by means of a fluorescence reaction, including a biosensor that reacts with the target analyte and generates a fluorescence signal; A data glove for real-time monitoring of finger joint bending and gesture recognition includes a glove and multiple fiber optic bending sensors attached to the joints of the glove. The fiber optic bending sensors reflect the degree of finger bending in real time through changes in light intensity and output multidimensional light intensity signals. A wearable fiber optic detector, connected to the fluorescent sensing patch and the data glove, is used to collect and analyze fluorescence signals and bending signals. The gesture recognition specifically includes: When the system is in operation, the wearable fiber optic sensing system extracts the R-channel pixel intensity as the output signal of the fiber optic bend sensor by acquiring images from a small camera. The initial output signal when the fiber optic bending sensor is not bent. 0; To quantize the output signal of the fiber optic bending sensor array to the same scale, the output signal is normalized using the following equation: in The bending angle of the fiber optic bending sensor. , , ′ is the linear conversion coefficient from the output power to the output light intensity signal of the fiber optic bending sensor. It represents the deviation from the linear fitting results; I ( n ) represents the first [value] of the fiber optic bending sensor array. n The total light intensity signal obtained from the sampling. I j ( n ) represents the first element in the fiber optic bending sensor array. j Fiber optic bending sensor n The light intensity signal obtained from the second sampling. I j ( n +1) indicates the first (in the fiber optic bending sensor array) j Fiber optic bending sensor n The light intensity signal obtained from +1 sampling; the feature curve is switched using gestures. This indicates the degree of change in finger joint movement; if the total light intensity signal of the fiber optic bending sensor array changes... If the value exceeds the threshold, the gesture is considered to have begun to change; if... If the value is less than a threshold, the gesture is considered unchanged; during the gesture change, the finger joint angle vector changes with the hand shape, and... As it changes; each gesture pattern is processed using formula (2); Where N is the total number of fiber optic bending sensors in the data glove. Forward difference is performed on the multidimensional light intensity signal output by the data glove to extract the bending features of the fiber optic bending sensors in real time. Absolute value operation is performed to eliminate the sign information in the data, so that the amplitude values ​​of the feature waveforms are all positive, simplifying the feature extraction and analysis process. An aggregation function is used to reduce the dimensionality of the multidimensional bending data through linear addition. Denoising pairs using thresholding Processing: in To determine whether a gesture has changed, a light intensity threshold is set. Maximum value detection identifies the gesture transition point, and the start and end times along with the gesture transition point serve as key points for continuous gesture segmentation. The data between any two key points represents an oversampled identical data point. The area surrounding each gesture transition point represents the gesture transition process. If it remains unchanged, it indicates a stable gesture maintenance process; The light intensity vector output by the fiber optic bending sensor at the midpoint between adjacent key points is used as the key information of the gesture pattern, and data compression is performed based on this. The compressed gesture pattern is transmitted to the gesture recognition module based on the radial basis function network model, and gesture recognition is performed using the trained model.

2. The multi-parameter wearable intelligent fiber optic sensing system according to claim 1, characterized in that, The fluorescent sensing patch further includes: An excitation fiber is used to transmit excitation light to the biosensor; A receiving optical fiber is used to capture the fluorescence signal generated by the biosensor; The patch substrate, made of flexible material, integrates the biosensor and optical fiber.

3. The multi-parameter wearable intelligent fiber optic sensing system according to claim 2, characterized in that, The biosensor is made by freeze-drying detection reagents onto the surface of filter paper. When the target substance enters the reaction chamber of the biosensor, ultraviolet light excites it to generate a fluorescence signal. The fluorescence signal is transmitted through a receiving optical fiber and captured by a miniature camera and converted into a fluorescence image. The change in the image intensity is linearly correlated with the concentration of the target substance.

4. The multi-parameter wearable intelligent fiber optic sensing system according to claim 3, characterized in that, The fiber optic bending sensor is attached to the metacarpophalangeal joints and proximal interphalangeal joints of the glove using transparent silicone rubber material.

5. A multi-parameter wearable intelligent fiber optic sensing system according to claim 4, characterized in that, The glove is also equipped with a fiber optic bend sensor at the wrist to improve the accuracy of recognizing repetitive gestures.

6. The multi-parameter wearable intelligent fiber optic sensing system according to claim 5, characterized in that, The gesture recognition uses the following algorithm: The acquired fiber optic images are segmented to obtain the output light intensity signal of each fiber optic bend sensor; The forward differential method is used to extract the real-time bending characteristics of the fiber optic bending sensor. Eliminate data noise using threshold denoising techniques; Detect bending features to identify gesture switching points, and segment the start and end times and gesture switching points as key points; The output light intensity signal at the intermediate moment between adjacent key points is retained as key information of the gesture pattern and compressed accordingly. The compressed gesture pattern is transmitted to the gesture recognition module based on the radial basis function network model for gesture recognition.

7. A multi-parameter wearable intelligent fiber optic sensing system according to claim 6, characterized in that, The optical fiber probe of the fluorescent sensing patch is chemically etched to remove part of the cladding to enhance the excitation light intensity. The specific etching method is as follows: the optical fiber is immersed in a 1:1 mixture of hexane and acetone for 5 seconds, and then cleaned, dried and the etching effect is fixed.

8. The multi-parameter wearable intelligent fiber optic sensing system according to claim 1, characterized in that, It also includes a wireless communication module, which is connected to the wearable fiber optic detector and is used to transmit the detection data of the wearable fiber optic detector to a remote server and / or personal device in real time.

9. A multi-parameter wearable intelligent fiber optic sensing system according to any one of claims 1-8, characterized in that, The target substance is a pathogenic microorganism.

Images