A wearable sports fatigue monitoring and reminder system and method based on multi-channel optical signals

Through the wearable motion fatigue monitoring system of multi-path optical signals, combined with multi-path optical fiber sensors and neural network algorithms, the accurate identification of mental and muscle fatigue status is achieved, and the monitoring accuracy and convenience in the existing technology is solved. Problem, providing an efficient rehabilitation monitoring solution.

CN118648877BActive Publication Date: 2024-11-08BEIJING NORMAL UNIV AT ZHUHAI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411132496.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-08
Estimated Expiration
2044-08-19

AI Technical Summary

Technical Problem

The prior art is difficult to accurately monitor the patient's mental fatigue and muscle fatigue status at the same time during the rehabilitation process, and it is difficult to integrate equipment and integrate information sources conflicts, resulting in insufficient accuracy and convenience.

Method used

Wearable motion fatigue monitoring system using multi-pass optical signals, combined with multi-pass optical fiber sensors and mobile portable devices to monitor respiratory ECG and gait data in real time, and use threshold segmentation method and neural network algorithm to separate signal channels and Fatigue status recognition.

Benefits of technology

It realizes accurate identification of mental fatigue and muscle fatigue status, improves monitoring accuracy and convenience, and enables the equipment to work independently without relying on other equipment, and is suitable for the medical monitoring field.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118648877B_ABST
    Figure CN118648877B_ABST
Patent Text Reader

Abstract

The present invention proposes a wearable sports fatigue monitoring and reminder system and method based on multi-channel optical signals, including: a data acquisition module, which collects multiple-channel optical signals through respiratory heart rate monitoring sensors and gait monitoring sensors to obtain physiological information; a data transmission module, which transmits data from the acquisition module to the processing module; a data processing module, which processes the original data information to determine the fatigue state; a feedback interaction module, which reminds doctors and patients of fatigue states in real time and on multiple terminals, and records relevant information and gives feedback after the rehabilitation exercise is completed. The present invention accurately constructs a patient fatigue detection model, proves the effectiveness of fatigue identification through experimental data, and can achieve timely feedback, providing a highly accurate and convenient integrated sports fatigue monitoring and reminder system and method for the medical monitoring field.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of medical monitoring technology, and in particular relates to a wearable sports fatigue monitoring and reminder system and method based on multi-channel optical signals. Background Art

[0002] Sports fatigue has long been one of the issues that need to be focused on during the rehabilitation process. Exercise itself can promote the recovery of patients' physical functions, but patients may be impatient for quick results or doctors' exercise therapy prescriptions are not standardized, which may lead to sports fatigue in patients, impairing the patient's rehabilitation effect. It will also increase the risk of reinjury, reduce the stability and tolerance of tissues such as muscles and ligaments, and cause emotional problems such as fatigue, frustration, and even anxiety and depression due to continuous fatigue. Therefore, the patient's fatigue state is also one of the important assessment indicators during the rehabilitation assessment process.

[0003] During the rehabilitation process, fatigue monitoring can help rehabilitation professionals better understand the patient's fatigue level and adjust the rehabilitation plan in time to avoid excessive fatigue and further injury. Common methods for monitoring sports fatigue status include physiological indicator monitoring and sports performance evaluation. Physiological indicator monitoring can reflect the patient's physiological state by monitoring heart rate, blood pressure, breathing and other indicators, thereby indirectly evaluating their fatigue level. Sports performance evaluation can evaluate the patient's fatigue level by observing the patient's sports performance, gait, etc.

[0004] In recent years, with the development of wearable technology, fatigue monitoring equipment has gradually been used in the field of medical rehabilitation. These devices can monitor the patient's relevant physiological information in real time, provide more accurate data for rehabilitation professionals, and help them better formulate rehabilitation plans. In addition to the common smart bracelets in daily life that can monitor physiological information, there are also some more accurate EEG technologies that can help patients perform rehabilitation exercises. In addition, with the development of fiber optic sensing technology, monitoring physiological information through optical fibers is also a common method in this field. However, smart watch monitoring tools often have limited accuracy, which means that it is difficult for doctors or professional caregivers to grasp the latest and most accurate physiological information of patients; EEG technology tools are limited in convenience, and patients are often required to wear EEG caps and can only do some simple rehabilitation exercises in situ. At the same time, the cost is also high, which is not suitable for ordinary patients; optical fiber physiological information monitoring tools currently often focus on obtaining respiratory heart rate signals alone or sole pressure information alone. The lower-dimensional network features of these physiological information are often not applied, resulting in fewer methods or equipment for fatigue state detection through optical fibers, and these devices are also often more dependent on other monitoring equipment such as electromyography equipment, body temperature sensors, etc., with certain defects in relative accuracy and convenience.

[0005] In summary, the existing technology mainly uses a single-channel optical fiber sensor to collect ECG data to identify mental fatigue or uses a single-channel optical fiber sensor to collect body movement information to identify muscle fatigue status. It is rare to detect both states at the same time; the difficulty lies in the difficulty of achieving integration from the perspective of equipment, and it is more dependent on external circuits, such as external ECG monitoring electrodes, which greatly increases the inconvenience; the other difficulty is the information source conflict and difficulty in effective fusion of multi-sensor systems, and its processing is often at the expense of accuracy.

[0006] In order to solve these problems, the new generation of wearable fatigue monitoring equipment during rehabilitation should have higher accuracy and convenience. By adopting advanced sensing technology, the individual's physiological state and movement state can be monitored in real time, so as to more accurately identify the fatigue state. At the same time, these devices should have the ability to work independently, without relying on other power supply or monitoring equipment, and can also be managed uniformly by dedicated ports to achieve multi-terminal simultaneous monitoring, which is more suitable for the field of medical monitoring and improves the convenience of use. Summary of the invention

[0007] The purpose of the present invention is to propose a wearable sports fatigue monitoring and reminder system and method based on multi-channel optical signals. The multi-channel optical fiber sensor is combined with a movable portable device to realize device integration. The integrated wearable device can monitor respiratory electrocardiogram and gait data in real time, and further process the data to realize mental fatigue and muscle fatigue state monitoring; and combined with the characteristics of hardware equipment, the threshold segmentation method is used to realize accurate separation of signal channels. Furthermore, the underlying features are extracted through the neural network algorithm, the data is used more effectively, and the accuracy of fatigue state recognition is correspondingly improved.

[0008] To achieve the above-mentioned purpose, the present invention provides a wearable sports fatigue monitoring and reminder system based on multi-channel optical signals, comprising: a data acquisition module, a data transmission module, a data processing module and a feedback interaction module;

[0009] The data collection module is used to collect multiple types of physiological information data of the user; wherein the multiple types of physiological information data include: respiratory physiological information, heart rate physiological information and gait information;

[0010] The data transmission module is used to transmit the collected multiple types of physiological information data of the user to the data processing module;

[0011] The data processing module is used to separate the multiple types of physiological information data into channels using a threshold segmentation method, and to identify the separated physiological information data using a preset mental fatigue model and a muscle fatigue model to determine the patient's fatigue state; wherein the fatigue state includes: mental fatigue state and muscle fatigue state;

[0012] The feedback interaction module is used to remind the user of the fatigue state in real time and provide status feedback.

[0013] Optionally, the data acquisition module includes: a respiratory heart rate monitoring sensor, a gait monitoring sensor and a movable portable device;

[0014] The respiration and heart rate monitoring sensor is used to collect the periodic physiological information of the user's respiration and heart rate;

[0015] The gait monitoring sensor is used to collect gait information of the user;

[0016] The movable portable device is used to collect periodic physiological information and gait information.

[0017] Optionally, the data processing module includes: a path separation unit, a data preprocessing unit and a fatigue determination unit;

[0018] The pathway separation unit is used to perform pathway separation on the multiple types of physiological information data, extract the optical signal and convert it into physiological information;

[0019] The path separation unit includes: a grayscale processing subunit, a threshold segmentation subunit, a mask operation subunit, a contour extraction subunit and a signal segmentation subunit;

[0020] The grayscale processing subunit is used to divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images;

[0021] The threshold segmentation subunit is used to perform threshold segmentation on the grayscale processed image using the Sauvola adaptive local binarization algorithm to obtain the threshold of each signal pathway;

[0022] The mask operation subunit is used to obtain a Boolean matrix corresponding to the image through mask operation based on the threshold of the signal path;

[0023] The contour extraction subunit is used to extract the contour of each signal pathway by marking the connected areas in the Boolean matrix and using the label function in the skimage library to obtain the edge coordinate point set;

[0024] The signal segmentation subunit is used to determine the positions of the images corresponding to different signal channels according to the edge coordinate point set, and segment the overall image signals of different channels by using ROI operation to realize the division of a single image signal path into different independent signal paths;

[0025] The data preprocessing unit is used to preprocess the separated physiological information data;

[0026] The fatigue determination unit is used to analyze the network characteristics of the pre-processed physiological information data to determine the fatigue state of the user;

[0027] The fatigue determination unit includes: a data set construction subunit, a model construction subunit, a training subunit, a mental fatigue state recognition subunit and a muscle fatigue state recognition subunit;

[0028] The data set construction subunit is used to construct a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and the corresponding fatigue levels;

[0029] The model building subunit is used to build a neural network algorithm model;

[0030] The training subunit is used to train the neural network algorithm model using the data set to obtain a mental fatigue model and a muscle fatigue model;

[0031] The mental fatigue state recognition subunit is used to recognize the respiratory physiological information and the heart rate physiological information by using the mental fatigue model to obtain the corresponding mental fatigue state;

[0032] The muscle fatigue state recognition subunit is used to recognize the gait information using the muscle fatigue model to obtain the corresponding muscle fatigue state.

[0033] Optionally, the feedback interaction module includes: a real-time reminder unit and a recording feedback unit;

[0034] The real-time reminder unit is used to specify a preset classical Chinese to remind the user based on the user's fatigue state, and to remind the user by controlling the flash of the smartphone's flashlight and the vibration ringtone;

[0035] The recording and feedback unit is used to record the exercise duration and recovery time of each user, and provide feedback to the staff after the rehabilitation exercise is completed.

[0036] The present invention also proposes a wearable sports fatigue monitoring and reminder method based on multi-channel optical signals, the method comprising:

[0037] Collect multiple types of physiological information data of the user, and use a threshold segmentation method to separate information channels of the multiple types of physiological information data; wherein the multiple types of physiological information data include: respiratory physiological information, heart rate physiological information and gait information;

[0038] Preprocessing the separated physiological information data;

[0039] Based on deep learning, the signal network change characteristics of the preprocessed physiological information data are extracted to evaluate and classify the fatigue state; wherein the fatigue state includes: mental fatigue state and muscle fatigue state;

[0040] Provide real-time reminders and record feedback on multiple terminals based on fatigue status determination.

[0041] Optionally, separating the multiple types of physiological information data into information channels includes:

[0042] Divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images;

[0043] The Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale image to obtain the threshold of each signal pathway.

[0044] Based on the threshold of the signal pathway, the Boolean matrix corresponding to the image is obtained through mask operation to reduce image noise, improve image contrast, and enhance image edge features;

[0045] By marking the connected regions in the Boolean matrix, the label function in the skimage library was used to extract the outline of each signal pathway and obtain the edge coordinate point set;

[0046] The edge coordinate point set is segmented by using ROI operation to divide a single image signal path into several signal paths.

[0047] Optionally, preprocessing the separated physiological information data includes:

[0048] Performing baseline correction on the separated physiological information data;

[0049] Perform signal denoising on the corrected data;

[0050] Perform sliding window segmentation processing on the denoised data;

[0051] The segmented data were subjected to signal normalization.

[0052] Optionally, extracting the signal network change characteristics of the preprocessed physiological information data based on deep learning and performing fatigue state assessment classification includes:

[0053] Constructing a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and corresponding fatigue levels;

[0054] Build a neural network algorithm model;

[0055] Using the data set to train the neural network algorithm model to obtain a mental fatigue model and a muscle fatigue model;

[0056] Using the mental fatigue model to identify the respiratory physiological information and the heart rate physiological information, and obtaining the corresponding mental fatigue state;

[0057] The muscle fatigue model is used to identify the gait information to obtain the corresponding muscle fatigue state.

[0058] Optionally, building a neural network algorithm model includes:

[0059] Modify the input layer in the MobileNetV3 algorithm model to form the neural network algorithm model;

[0060] Using the data set to train the neural network algorithm model includes:

[0061] Processing the data set into a one-dimensional time-domain series;

[0062] The input one-dimensional time domain sequence is amplified and sampled through convolution, and the sampled data is processed through several blocks;

[0063] Overlap the feature matrices output in the blocks, then perform convolution and dimensionality reduction, use the activation function to act on the feature vector after dimensionality reduction, and complete data feature extraction;

[0064] A multi-layer perception structure is introduced in the fully connected layer to complete the classification.

[0065] Optionally, the Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale processed image, and the threshold of each signal pathway is obtained, including:

[0066] For each pixel in the grayscale image, the local mean is calculated in the defined neighborhood according to the pre-formula;

[0067]

[0068] in, represents the total number of pixels in the neighborhood, represents the light intensity at the pixel of the corresponding coordinate in the local neighborhood, i represents the i-th pixel point processed in the neighborhood, Represents the local pixel intensity mean in the neighborhood;

[0069] Based on the local mean, obtaining the local standard deviation in the same neighborhood;

[0070]

[0071] in, represents the local standard deviation within the same neighborhood;

[0072] Obtain the adaptive threshold of each pixel based on the pixel neighborhood mean and local standard deviation;

[0073]

[0074] in, Represents the adaptive threshold of the pixel (x, y), Indicates custom correction parameters, 0 <k<1, Represents the dynamic range of pixel intensity, usually determined by the grayscale range of the image;

[0075] According to the original intensity value of the pixel and the corresponding adaptive threshold For comparison, if the original intensity value of the pixel is higher than the adaptive threshold, it is recorded as 1, otherwise it is recorded as 0. Generally, the same channel signal will be recorded as the same threshold.

[0076] The present invention has the following beneficial effects:

[0077] In terms of hardware, the device is integrated by combining multi-channel optical fiber sensors with portable devices. The integrated wearable device can monitor respiratory, electrocardiographic and gait data in real time, and further process the data to monitor mental fatigue and muscle fatigue status.

[0078] In data processing, the present invention combines the characteristics of hardware equipment and uses a threshold segmentation method to achieve accurate separation of signal channels. Furthermore, the underlying features are extracted through a neural network algorithm to make more effective use of data, and the accuracy of fatigue state recognition is correspondingly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0080] Figure 1 A schematic diagram of the structure of a wearable sports fatigue monitoring and reminder system based on multi-channel optical signals according to an embodiment of the present invention;

[0081] Figure 2 A schematic diagram of the structure of a data acquisition module and related devices according to an embodiment of the present invention;

[0082] Figure 3 A schematic flow chart of a method for sports fatigue monitoring and feedback according to an embodiment of the present invention;

[0083] Figure 4 This is a schematic diagram of the conversion of the original optical signal according to an embodiment of the present invention;

[0084] Figure 5This is a flow chart of step S100 of the method for monitoring and providing feedback on exercise fatigue according to an embodiment of the present invention;

[0085] Figure 6 This is a flow chart of step S200 of the method for monitoring and providing feedback on exercise fatigue according to an embodiment of the present invention;

[0086] Figure 7 A schematic diagram of extracting time domain signals of respiration and heart rate according to an embodiment of the present invention;

[0087] Figure 8 A schematic diagram of gait frequency domain signal extraction according to an embodiment of the present invention;

[0088] Fig. 9 This is a schematic diagram of the algorithm model training process of step S400 of the method for monitoring and providing feedback on exercise fatigue according to an embodiment of the present invention;

[0089] Fig.10 This is a schematic diagram of confusion matrix results for mental fatigue state recognition according to an embodiment of the present invention;

[0090] Fig.11 This is a schematic diagram of confusion matrix results for muscle fatigue state recognition according to an embodiment of the present invention;

[0091] Fig.12 This is a schematic diagram of network parameters for modifying the input layer of the MobileNetV3 algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0092] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0093] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0094] In this embodiment, fatigue states are divided into muscle fatigue and mental fatigue. The former is due to the decline in physical function caused by continuous body movement, and the latter is due to the decline in physical performance caused by long-term cognitive activities. In the medical rehabilitation process, due to the complexity of the patient's physical condition, it is very likely that the two fatigue states will appear simultaneously or separately. Therefore, the two fatigue states are not simply merged. This is also different from the traditional method of only using physiological information such as breathing and heart rate to identify mental fatigue states or two mixed fatigue states. This embodiment can more accurately distinguish and identify muscle fatigue and mental fatigue states, which helps doctors or caregivers to understand individual conditions more accurately and help achieve personalized diagnosis. For different types of fatigue, fatigue relief is performed in different ways.

[0095] like Figure 1 As shown, this embodiment proposes a wearable sports fatigue monitoring and reminder system based on multi-channel optical signals, including: a data acquisition module, a data transmission module, a data processing module and a feedback interaction module; the data acquisition module is used to collect multiple types of physiological information data of the user; the data transmission module is used to transmit the collected multiple types of physiological information data of the user to the data processing module; the data processing module is used to judge the patient's fatigue state based on the multiple types of physiological information data; the feedback interaction module is used to remind the user of the fatigue state in real time and provide status feedback.

[0096] Furthermore, the data acquisition module includes: a respiratory and heart rate monitoring sensor, a gait monitoring sensor and a movable portable device; the respiratory and heart rate monitoring sensor is used to collect the user's periodic physiological information of breathing and heart rate; the gait monitoring sensor is used to collect the user's gait information; the movable portable device is used to collect the collected periodic physiological information and gait information.

[0097] Further, the data processing module includes: a path separation unit, a data preprocessing unit and a fatigue determination unit;

[0098] A pathway separation unit is used to perform pathway separation on multiple types of physiological information data using a threshold segmentation method, extract optical signals and convert them into physiological information;

[0099] The channel separation unit includes: a grayscale processing subunit, a threshold segmentation subunit, a mask operation subunit, a contour extraction subunit and a signal segmentation subunit;

[0100] A grayscale processing subunit is used to divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images;

[0101] A threshold segmentation subunit is used to perform threshold segmentation on the grayscale processed image using the Sauvola adaptive local binarization algorithm to obtain the threshold of each signal pathway;

[0102] A mask operation subunit, used for obtaining a Boolean matrix corresponding to the image through mask operation based on a threshold of a signal path;

[0103] The contour extraction subunit is used to extract the contour of each signal pathway by marking the connected areas in the Boolean matrix and using the label function in the skimage library to obtain the edge coordinate point set;

[0104] The signal segmentation subunit is used to determine the positions of the images corresponding to different signal channels according to the edge coordinate point set, and segment the overall image signals of different channels using ROI calculations to divide a single image signal path into different independent signal paths;

[0105] The Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale image, and the thresholds of each signal pathway are obtained, including:

[0106] For each pixel in the grayscale image, the local mean is calculated in the defined neighborhood according to the pre-formula;

[0107]

[0108] in, represents the total number of pixels in the neighborhood, Represents the light intensity at the pixel of the corresponding coordinate in the local neighborhood, i represents the i-th pixel point processed, Represents the local pixel intensity mean in the neighborhood;

[0109] Based on the local mean, obtain the local standard deviation in the same neighborhood;

[0110]

[0111] in, represents the local standard deviation within the same neighborhood;

[0112] Obtain the adaptive threshold of each pixel based on the pixel neighborhood mean and local standard deviation;

[0113]

[0114] in, Represents the adaptive threshold of the pixel (x, y), Indicates custom correction parameters, 0 <k<1, Represents the dynamic range of pixel intensity;

[0115] The original intensity value of the pixel is compared with the corresponding adaptive threshold. If the original intensity value of the pixel is higher than the adaptive threshold, it is recorded as 1, otherwise it is recorded as 0. The same channel signal will be recorded as the same threshold.

[0116] A data preprocessing unit, used for preprocessing the separated physiological information data;

[0117] A fatigue determination unit, used to analyze the network characteristics of the pre-processed physiological information data to determine the user's fatigue state;

[0118] The fatigue determination unit includes: a data set construction subunit, a model construction subunit, a training subunit, a mental fatigue state recognition subunit and a muscle fatigue state recognition subunit;

[0119] The data set construction subunit is used to construct a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and the corresponding fatigue degrees;

[0120] A model building subunit, used to build a neural network algorithm model;

[0121] Building a neural network algorithm model includes:

[0122] Modify the input layer in the MobileNetV3 algorithm model so that the MobileNetV3 algorithm model can process one-dimensional time domain data and form a neural network algorithm model;

[0123] Using data sets to train neural network algorithm models includes:

[0124] Process the data set into a one-dimensional time-domain series;

[0125] The input one-dimensional time domain sequence is amplified and sampled through convolution, and the sampled data is processed through several blocks;

[0126] Overlap the feature matrices output in the blocks, then perform convolution and dimensionality reduction, use the activation function to act on the feature vector after dimensionality reduction, and complete data feature extraction;

[0127] Introduce a multi-layer perception structure in the fully connected layer to complete the classification;

[0128] A training subunit, used to train the neural network algorithm model using the data set to obtain a mental fatigue model and a muscle fatigue model;

[0129] A mental fatigue state recognition subunit is used to recognize respiratory physiological information and heart rate physiological information using a mental fatigue model to obtain a corresponding mental fatigue state;

[0130] The muscle fatigue state recognition subunit is used to recognize gait information using a muscle fatigue model and obtain the corresponding muscle fatigue state.

[0131] Further, the feedback interaction module includes: a real-time reminder unit and a recording feedback unit;

[0132] The real-time reminder unit is used to specify preset textual language for reminder based on the user's fatigue status, and to give reminders by controlling the flashing of the smartphone flash and the vibration ringtone; the recording feedback unit is used to record the exercise duration and recovery time of each user, and provide feedback to the staff after the rehabilitation exercise is completed.

[0133] More specifically, the present embodiment provides a wearable sports fatigue monitoring and reminder system based on multi-channel optical signals, which includes: a data acquisition module S10, a data transmission module S20, a data processing module S30, and a feedback interaction module S40;

[0134] Data acquisition module S10, reference Figure 2 A detailed schematic diagram is given, including a respiratory heart rate monitoring sensor, a gait monitoring sensor, and a smart phone for collecting monitoring sensor data, wherein the respiratory heart rate monitoring sensor and the gait monitoring sensor are both implemented by built-in optical fiber sensors;

[0135] The respiratory and heart rate monitoring sensor is composed of elastic textile clothing and an optical fiber sensor. The optical fiber sensor is fixed to the inner side of the elastic textile clothing by Velcro and tape, close to the patient's chest.

[0136] Since the patient's breathing, heart rate, etc. will cause a certain periodic vibration on the chest, which can act on the concave notch of the optical fiber sensor, the curvature of the concave notch of the optical fiber sensor will change under the action of the external vibration force, and then the optical power attenuation will also change, that is, light loss corresponding to the external force will be generated. The patient's periodic physiological information can be extracted from the change information of the light loss, thereby realizing the function of monitoring the patient's breathing and heart rate;

[0137] The gait monitoring sensor consists of an insole and three fiber optic sensors embedded therein. The fiber optic sensors are fixed by hot melt adhesive and tape, and the sensing area can cover the entire plantar pressure area of ​​the patient.

[0138] When the patient stands, walks or runs, the soles of the feet touch the ground, and the concave notch of the optical fiber sensor bends due to pressure, which manifests as fluctuations in optical power. The patient's gait information can be obtained based on this optical power change, i.e., optical loss.

[0139] When the patient is in different mental or muscular states, the patient's breathing, heart rate and gait will have different states. According to this change, further processing can obtain the fatigue state;

[0140] The smartphone plays many roles in the system. The smartphone flashlight is used to provide light source, and the intensity of the flashlight can be adjusted on the smartphone to adjust the intensity of the light source. The smartphone camera module is used as a receiver to receive optical signals.

[0141] The above-mentioned optical fiber sensor is installed on the camera module of the smartphone through a special optical fiber fixing module, so that four optical fiber signal pathways are directly transmitted to the camera, and it is not easy to fall, thus ensuring stability.

[0142] The data transmission module S20 also mainly relies on the Wi-Fi module of the smartphone for transmission. The system uses the smartphone’s camera to collect video data in real time through the application iVCam. The video data is transmitted from the smartphone to the PC via Wi-Fi.

[0143] By constructing a data transmission module in this way, low-latency, high-speed transmission and high-quality real-time video recording can be achieved. At the same time, in order to reduce the use of computer CPU / GPU, the system completely uses the video data recorded by the smartphone for processing without re-encoding the video. The collected video data is usually a series of original image frames and is stored in a designated folder of the PC. The system implements the continuous query function of video data through a loop structure, and further writes a video data processing program to achieve real-time video reading and data processing functions, after which the image can be processed to obtain more information.

[0144] The data processing module S30 mainly processes the original image information by an external computer, obtains the patient's physiological information, and determines his fatigue state through a specific algorithm model on this basis.

[0145] The feedback interaction module S40 mainly includes two functions: real-time reminder function and feedback recording function;

[0146] The real-time reminder function is designed to help doctors understand the patient's physical condition and intervene in time. It can also remind the patient to take proper rest and avoid excessive exercise.

[0147] Specifically, when a patient experiences muscle fatigue or mental fatigue, the physician's monitoring end, i.e., the external computer display end, will remind the patient through designated classical Chinese. For example, the classical Chinese is as follows: "At 15:30, the patient experienced muscle fatigue, but not mental fatigue. Please remind and intervene in time." In addition, it can also be customized to suit patients in different medical rehabilitation scenarios.

[0148] Patients are reminded through the vibration of their cell phones and the flashing of their smartphones’ lights. The system controls the intensity of the smartphone audio and lights through built-in programs.

[0149] The flash frequency of the smartphone flash is 0.2Hz, and the flash brightness is the maximum intensity of the flash. When the PC detects that the patient is in any fatigue state, in addition to the monitoring system sending a reminder to the doctor in a conspicuous position, the smartphone flash intensity and audio are also controlled at the same time to remind the patient to rest through visual and auditory means, which is effective and not too stimulating. Moreover, this combination of visual and audio reminder can help patients to timely detect their personal physical conditions while also taking care of elderly patients with poor vision or hearing or some disabled patients.

[0150] The recording and feedback function will record the patient's exercise duration and fatigue recovery time each time, and provide feedback to the physician after the rehabilitation exercise is completed, helping the physician to provide personalized diagnosis and prescribe exercise prescriptions that are more suitable for the patient's physical condition.

[0151] The present invention also proposes a wearable sports fatigue monitoring and reminder method based on multi-channel optical signals, the method comprising:

[0152] Collect multiple types of physiological information data of the user, and use a threshold segmentation method to separate information channels of the multiple types of physiological information data; wherein the multiple types of physiological information data include: respiratory physiological information, heart rate physiological information and gait information;

[0153] Preprocessing the separated physiological information data;

[0154] Based on deep learning, the signal network change characteristics of the preprocessed physiological information data are extracted to evaluate and classify the fatigue state; wherein the fatigue state includes: mental fatigue state and muscle fatigue state;

[0155] Provide real-time reminders and record feedback on multiple terminals based on fatigue status determination.

[0156] Furthermore, separating the multiple types of physiological information data into information channels includes:

[0157] Divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images;

[0158] The Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale image to obtain the threshold of each signal pathway.

[0159] Based on the threshold of the signal pathway, the Boolean matrix corresponding to the image is obtained through mask operation;

[0160] By marking the connected regions in the Boolean matrix, the label function in the skimage library was used to extract the outline of each signal pathway and obtain the edge coordinate point set;

[0161] The edge coordinate point set is segmented by using ROI operation to divide a single image signal path into several signal paths.

[0162] Furthermore, preprocessing the separated physiological information data includes:

[0163] Performing baseline correction on the separated physiological information data;

[0164] Perform signal denoising on the corrected data;

[0165] Perform sliding window segmentation processing on the denoised data;

[0166] The segmented data were subjected to signal normalization.

[0167] Furthermore, based on deep learning, the signal network change characteristics of the preprocessed physiological information data are extracted to perform fatigue state assessment and classification, including:

[0168] Constructing a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and corresponding fatigue levels;

[0169] Build a neural network algorithm model;

[0170] Using the data set to train the neural network algorithm model to obtain a mental fatigue model and a muscle fatigue model;

[0171] Using the mental fatigue model to identify the respiratory physiological information and the heart rate physiological information, and obtaining the corresponding mental fatigue state;

[0172] The muscle fatigue model is used to identify the gait information to obtain the corresponding muscle fatigue state.

[0173] Furthermore, building a neural network algorithm model includes:

[0174] Modify the input layer in the MobileNetV3 algorithm model to form the neural network algorithm model;

[0175] Using the data set to train the neural network algorithm model includes:

[0176] Processing the data set into a one-dimensional time-domain series;

[0177] The input one-dimensional time domain sequence is amplified and sampled through convolution, and the sampled data is processed through several blocks;

[0178] Overlap the feature matrices output in the blocks, then perform convolution and dimensionality reduction, use the activation function to act on the feature vector after dimensionality reduction, and complete data feature extraction;

[0179] A multi-layer perception structure is introduced in the fully connected layer to complete the classification.

[0180] Furthermore, the Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale processed image, and the thresholds of each signal pathway are obtained, including:

[0181] For each pixel in the grayscale image, the local mean is calculated in the defined neighborhood according to the pre-formula;

[0182]

[0183] in, represents the total number of pixels in the neighborhood, represents the light intensity at the pixel of the corresponding coordinate in the local neighborhood, i represents the i-th pixel point processed in the neighborhood, Represents the local pixel intensity mean in the neighborhood;

[0184] Based on the local mean, obtaining a local standard deviation in the same neighborhood, and using the local standard deviation in the same neighborhood to measure the change of pixel intensity;

[0185]

[0186] in, represents the local standard deviation within the same neighborhood;

[0187] Obtain the adaptive threshold of each pixel based on the pixel neighborhood mean and local standard deviation;

[0188]

[0189] in, Represents the adaptive threshold of the pixel (x, y), Indicates custom correction parameters, 0 <k<1, It represents the dynamic range of the standard deviation, which is set according to the grayscale level of the input image;

[0190] According to the original intensity value of the pixel and the corresponding adaptive threshold For comparison, if the original intensity value of the pixel is higher than the adaptive threshold, it is recorded as 1, otherwise it is recorded as 0. Generally, the same channel signal will be recorded as the same threshold.

[0191] like Figure 3 As shown, a flow chart of the method for sports fatigue monitoring and feedback is provided, which reflects the main process of the data processing module and the feedback interaction module, including the following steps:

[0192] S100: Separating the multi-channel optical signal data collected by the sensor to obtain corresponding channel information;

[0193] Since there are multiple sensors in the sensing system, specifically including one signal path of the respiratory and heart rate monitoring sensor and three signal paths of the gait monitoring device, a single frame of image information acquired by the smartphone contains multiple sensor signals;

[0194] In order to monitor multi-sensor signals, the system needs to obtain different sensor response signals from single-frame image signals and process them separately. In this sensor system, due to the existence of connecting devices, the position of the signal path is fixed, and different signal paths are located in Figure 4 At different light spots in the left image, the image segmentation method can be used to separate different areas of an image, and finally multiple signal pathways can be separated from a single signal pathway;

[0195] S200: signal preprocessing;

[0196] The original sensor signal will be polluted by various noises, such as ambient light, temperature fluctuations, electromagnetic interference, and the patient's subconscious body tremors, which will make the measurement data affected by errors and noise. Therefore, it is necessary to pre-process the original motion signal before further analysis;

[0197] Signal preprocessing mainly consists of the following steps: baseline correction, denoising, sliding window segmentation, and normalization;

[0198] S300: signal filtering to obtain physiological signals;

[0199] By using different filters, the respiration, heart rate and gait frequency signals can be extracted from different signal paths respectively; for the respiration and heart rate sensor part, since respiration and heartbeat have different frequency ranges, this embodiment selects a filter range larger than the normal range to extract the heart rate and respiration frequency, the heart rate range is set to 1-2.5 Hz, the respiration frequency range is set to 0.1-0.8 Hz, and the frequency ranges are non-overlapping;

[0200] For the gait monitoring device, a filter in the range of 0.5-3.0 Hz is used to cover a variety of gait modes, obtain gait signals such as walking, running, and jumping, and further obtain corresponding information such as pressure and step length;

[0201] S400: Extract signal network change characteristics based on deep learning to achieve fatigue status assessment and classification;

[0202] This step includes two parts: neural network model construction and algorithm model training;

[0203] The neural network model is constructed using the MobileNetV3 neural network algorithm in deep learning. This is a lightweight deep neural network algorithm designed for efficient mobile and edge device deployment, emphasizing high accuracy and reducing computational complexity.

[0204] Since the MobileNetV3 algorithm focuses on efficiency, it is an ideal choice for scenarios with limited computing resources. In addition, the signal value of each channel in this sensor system consists of continuous grayscale values ​​of the image, while the data type that the neural network needs to process is one-dimensional time domain sequence data. This convolutional neural network architecture is particularly good at image classification tasks, so it is very suitable for the algorithm model of this system.

[0205] Furthermore, the original MobileNetV3 algorithm cannot process one-dimensional time domain sequence data, so it is necessary to modify the network parameters of the input layer to process the time domain data;

[0206] The MobileNetV3 algorithm extracts the network change characteristics of the signal through convolution, introduces a multi-layer perception structure for classification, and can realize classification judgment;

[0207] To realize the recognition of mental and muscle fatigue states, the model needs to be trained accordingly to realize the determination of mental fatigue states and muscle fatigue states;

[0208] S500: Perform multi-terminal real-time reminders and record feedback based on fatigue status determination;

[0209] Based on the trained fatigue state recognition model, the system can obtain the patient's current physiological state information;

[0210] The feedback interaction module determines the patient's physiological status information. When muscle fatigue or mental fatigue occurs, the system will give feedback reminders;

[0211] At the physician's monitoring end, i.e., the external computer display end, a reminder will be given in a specified language, including relevant timestamps, fatigue status information, and relevant guidance;

[0212] The patient side uses a combination of sound and light to remind the patient to take a proper rest or slow down the pace of exercise;

[0213] When the patient's relevant physiological information begins to return to normal, the reminder stops, and the background still records the relevant physiological information and time in real time;

[0214] Once the patient has fully recovered, the system will record it as an exercise cycle, including exercise duration and recovery time, and provide feedback to the physician so that a more appropriate exercise prescription can be issued.

[0215] refer to Figure 5In specific implementation, the above step S100 can be implemented by the following steps:

[0216] S110: Image grayscale processing;

[0217] The original multi-channel optical signal is received by the signal receiver, i.e., the camera module of the smartphone, and the data is in the form of a video signal, i.e., a continuous image frame. Therefore, the video signal needs to be segmented into images frame by frame for subsequent processing.

[0218] After the image is extracted, it is necessary to grayscale the image, so that the system can directly extract the optical signal information for processing and utilization, reducing the amount of calculation. The system uses the weighted average method for grayscale processing;

[0219] S120: Threshold segmentation using Sauvola adaptive local binarization algorithm;

[0220] After extracting a single frame image from the set, considering the existence of multiple pathways and possible differences in light intensity, the Sauvola adaptive local binarization algorithm is used for threshold segmentation. This algorithm can extract the threshold of each pathway through the neighborhood mean and neighborhood variance;

[0221] For each pixel in the image, the local mean is calculated within the defined neighborhood according to the following formula:

[0222]

[0223] Where I(x i ,y i ) is the light intensity at the pixel of the corresponding coordinate in the local neighborhood, and N is the total number of pixels in the neighborhood;

[0224] The local standard deviation stddev(x,y) in the same neighborhood is used to measure the change in pixel intensity. The calculation formula is as follows:

[0225]

[0226] Finally, the adaptive threshold T(x,y) of each pixel can be obtained according to the Sauvola formula:

[0227]

[0228] Where T(x,y) is the adaptive threshold of the pixel, k is the custom correction parameter, and R is the dynamic range of the pixel intensity, which usually depends on the grayscale range of the image. The application of the Sauvola algorithm enables the binary image to be effectively segmented into different areas with different local illumination conditions. This adaptive threshold segmentation method is beneficial to images with uneven illumination, making it suitable for scenes with different contrasts in different areas of the image. By adjusting the parameters k and R, the system can fine-tune the sensitivity and adaptability of the algorithm.

[0229] S130: Edge feature enhancement and local image segmentation;

[0230] After obtaining the threshold of each channel, the Boolean matrix corresponding to the image can be obtained through mask operation, which can reduce image noise, improve image contrast, and enhance the edge features of the image;

[0231] By marking the connected regions in the Boolean matrix, the label function in the skimage library can be used to extract the outline of each signal pathway and obtain the edge coordinate point set;

[0232] Finally, the specific positions of the four signal channels corresponding to the image can be locked according to the edge coordinate point set, and the ROI operation is used to segment the overall image signal of the four channels, so that a single image signal path can be divided into four independent signal paths. The program is used to read the signal of each signal path separately, and then the read signal is further processed.

[0233] Reference Figure 6 In specific implementation, the above step S200 can be implemented by the following steps:

[0234] S210: performing baseline correction on the original response signal;

[0235] After obtaining the signal after pathway separation, for the original response signal of each sensor, the detrend function is used to fit a linear function to the signal data by the least squares method;

[0236] S220: signal denoising;

[0237] Wavelet forced denoising is used to remove high-frequency noise in the original signal. The original signal is decomposed into two layers using the db5 wavelet basis function, and the detail coefficient is set to zero to filter out high-frequency noise.

[0238] S230: Sliding window segmentation;

[0239] The signal after baseline removal and noise reduction is segmented using a window size of 200. There is an overlap of 100 samples between consecutive windows. This window size is convenient for fast Fourier transform and Welch's method to calculate certain features. In this embodiment, feature extraction with a 50% overlapping sliding window is most suitable.

[0240] S240: signal normalization processing;

[0241] When collecting signals, the signals need to be normalized to eliminate the differences caused by inconsistent light intensity collected from different patients;

[0242] The system uses the Max-Min normalization method to normalize the signal so that the signal value is uniformly normalized between -1 and 1. While eliminating differences, the normalization process retains the relative relationships of the original data and maps these relative relationships to a more processable range, which has a positive impact on the convergence and performance of the machine learning algorithm.

[0243] Above, after the signal path segmentation and preprocessing, the signal can be further filtered to obtain the patient's breathing, heart rate and gait information;

[0244] Figure 7 The figure is a time domain signal extraction diagram of respiration and heart rate according to an embodiment. Respiration and heart rate can be displayed as power spectrum density of 0.30 Hz and 0.97 Hz, respectively, corresponding to 18 cpm (counts per minute) and 58.2 bpm (beats per minute). A medical heart rate monitor is used as a reference to measure respiration and heart rate. The results show that the relative error of respiration measurement is less than 4.5%, and the relative error of heart rate measurement is less than 5.4%, which proves its measurement accuracy.

[0245] Figure 8 This is a diagram of gait frequency domain signal extraction according to an embodiment. In the diagram, the gait frequencies of walking and running are 0.65 Hz and 1.55 Hz respectively, which are within the normal human gait frequency range, proving the accuracy of gait monitoring measurement.

[0246] Furthermore, on the basis of the available physiological data, to implement step S400, the following steps are required:

[0247] First, build a neural network algorithm model:

[0248] The MobileNetV3 algorithm used in this system builds the model. The algorithm usually consists of a backbone network with an inverse residual block, which includes a depth-separable convolution and a linear bottleneck structure. In the case of one-dimensional data, the input will be regarded as a single-channel sequence, so the input layer needs to be modified to process one-dimensional time domain sequence data, such as Fig.12 As shown, the specific steps include:

[0249] 1) The input one-dimensional time domain sequence is amplified and sampled through convolution, and the sampled data is processed through multiple blocks;

[0250] In each block, the data is first convolved to extract the initial features, and then the feature matrix enters the squeezing and excitation process. In this process, the features are concentrated and then multiplied by the convolved feature vector by two nonlinear activation functions Relu and h-swish to obtain a new feature matrix and output it.

[0251] Among them, the Relu function and h-swish function are as follows:

[0252]

[0253] 2) Overlap the feature matrices output in the blocks, then perform convolution and dimensionality reduction, use the activation function to act on the feature vector after dimensionality reduction, and complete data feature extraction;

[0254] 3) Finally, a multi-layer perception (MLP) structure is introduced in the fully connected layer to complete the classification;

[0255] Unlike the support vector machine algorithm, MobileNetV3 does not need to extract statistical features based on samples. Its features can be obtained by performing a series of depth-separable convolutions on the network itself.

[0256] Further, after completing the neural network algorithm model, refer to Fig. 9 In a specific implementation, the algorithm model training of step S400 can be implemented by the following steps:

[0257] S410: inform the subjects of the relevant instructions and complete the fatigue scale assessment;

[0258] In this embodiment, the physiological data of individual fatigue status are collected by designing mental fatigue recognition experiments and muscle fatigue experiments, and model training and testing are carried out; before the experiment begins, the experimenter is required to inform the subjects of the experimental objectives and the overall overview of the experimental methods, and sign the corresponding written informed consent; the purpose of this experiment is not explained in detail to the subjects, because early disclosure may cause the subjects to intentionally control their own behavior, unable to reflect the real situation and provide valid data; this partial disclosure is also in line with the relevant experimental requirements and standards, and after the experiment is completed, corresponding supplementary explanations will be given; after the written informed consent is signed, the subjects are required to complete the fatigue scale to evaluate their experimental status in advance, in order to meet the experimental requirements;

[0259] S420: mental fatigue and muscle fatigue training data collection;

[0260] In this embodiment, the mental fatigue recognition experiment specifically includes two parts, a non-mental fatigue state experiment and a mental fatigue state experiment;

[0261] Specifically, in the non-fatigue state experiment, the subjects were asked to sit down and watch a video of their interest to keep themselves alert. The whole process took ten minutes, and relevant physiological data were collected as model training materials.

[0262] In the fatigue state experiment, the subjects were required to read boring English literature and hypnotic music in a quiet environment and enter the fatigue hypnotic state as quickly as possible. The induction process took 30 minutes. After the induction process was completed, the data collection process continued for 10 minutes. The subjects were required to fill in a form every 2 minutes to score and record their fatigue level according to the corresponding standards;

[0263] The muscle fatigue identification experiment was divided into two parts: the non-muscle fatigue state experiment and the muscle fatigue state experiment. In the non-muscle fatigue experiment, the subjects were asked to walk at a comfortable speed for 10 minutes. The participants were asked to rest before the experiment to avoid any potential fatigue effects. In order to simulate the real lower limb muscle fatigue task, before the muscle fatigue state experiment, each participant had to do squats for 10 minutes. After the induction part, the participants were asked to repeat the experiment in the non-muscle fatigue state to collect gait data in the fatigue state. During the whole experiment, the participants reported their fatigue scores every 2 minutes.

[0264] S430: Divide the data into a training set and a test set to verify the effectiveness;

[0265] After the data collection was completed, the mental fatigue and muscle fatigue models were trained and tested for validation;

[0266] In the mental fatigue part, the respiration and heart rate data are divided into two groups: fatigue state and non-fatigue state according to the scoring criteria. The original data set is segmented and labeled based on the requirements of the MobileNetV3 neural network, and then divided into a test set and a training set at a ratio of 2:8. In this embodiment, the original respiration and heart rate data are used, and the trained model achieves an accuracy of 0.94 on the test set. The confusion matrix score is referenced Fig.10 , which shows that the model can effectively monitor the mental fatigue state of human body in sitting position;

[0267] In the muscle fatigue part, according to previous studies, muscle fatigue can be divided into multiple levels. Therefore, this embodiment divides the fatigue level into 4 levels (0:0-6, 1:7-12, 2:13-16, 3:17-20) according to the different distribution of RPE scores (fatigue perception score). Similarly, according to the requirements of the MobileNetV3 neural network, the original data set obtained after segmentation and annotation is divided into a test set and a training set at a ratio of 2:8. The accuracy of the trained model on the test set reaches 0.95, and the confusion matrix score reference Fig.11 , indicating that the trained model can classify different fatigue levels using gait signals.

[0268] The present embodiment provides a wearable sports fatigue monitoring and reminder system and method based on multi-channel optical signals. The sensor used is mainly composed of a polymer optical fiber sensor. The optical fiber has a signal modulation part, namely, a light loss part. This part is extremely sensitive to changes in optical signals. Compared with ordinary sensors, this feature also makes it more sensitive and more accurate when sensing weak signals such as breathing and heart rate. The sensor does not require a supporting structure to keep the optical fiber fixed, which can ensure the patient's free movement and avoid the sliding error of the sensor. While ensuring the monitoring efficiency, the cost of the material is also lower, and it is easy to promote and apply to ordinary patients' families.

[0269] The optical fiber sensor used in this embodiment has high flexibility and bendability, and is suitable for use in environments that require bending or flexing. It is very suitable for the sports scenes in the medical rehabilitation process, is not easy to break or damage, can be reused, and has anti-electromagnetic interference ability, better biocompatibility and chemical stability, and does not need to directly contact the skin, making the monitoring process safe and stable. It is suitable for various patients, and the cost is very low, which is easy to promote;

[0270] This embodiment also utilizes the characteristics of the optical fiber material itself, and realizes an audio-visual reminder method by flashing a flash light at the optical fiber gap and combining it with a mobile phone ringtone. This can take care of elderly patients with poor eyesight or hearing or some patients with hearing or visual disabilities, and is more suitable for medical monitoring scenarios.

[0271] The medical rehabilitation process motion fatigue monitoring system of this embodiment realizes the integration of respiratory heart rate monitoring and gait monitoring. Its innovation lies in providing a highly accurate and convenient integrated motion fatigue monitoring device for the medical monitoring field. In particular, it can accurately distinguish and identify mental fatigue and muscle fatigue states, and can use related equipment for multi-terminal reminders. This is of great significance for helping patients in this field to achieve full-process fatigue monitoring and helping doctors to achieve personalized rehabilitation diagnosis.

[0272] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A wearable sports fatigue monitoring and reminder system based on multi-channel optical signals, characterized in that: include: Data acquisition module, data transmission module, data processing module and feedback interaction module; The data collection module is used to collect multiple types of physiological information data of the user; wherein the multiple types of physiological information data include: respiratory physiological information, heart rate physiological information and gait information; The data transmission module is used to transmit the collected multiple types of physiological information data of the user to the data processing module; The data processing module is used to separate the multiple types of physiological information data into channels using a threshold segmentation method, and to identify the separated physiological information data using a preset mental fatigue model and a muscle fatigue model to determine the patient's fatigue state; wherein the fatigue state includes: mental fatigue state and muscle fatigue state; The data processing module includes: a path separation unit, a data preprocessing unit and a fatigue determination unit; The pathway separation unit is used to perform pathway separation on the multiple types of physiological information data, extract the optical signal and convert it into physiological information; The path separation unit includes: a grayscale processing subunit, a threshold segmentation subunit, a mask operation subunit, a contour extraction subunit and a signal segmentation subunit; The grayscale processing subunit is used to divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images; The threshold segmentation subunit is used to perform threshold segmentation on the grayscale processed image using the Sauvola adaptive local binarization algorithm to obtain the threshold of each signal pathway; The mask operation subunit is used to obtain a Boolean matrix corresponding to the image through mask operation based on the threshold of the signal path; The contour extraction subunit is used to extract the contour of each signal pathway by marking the connected areas in the Boolean matrix and using the label function in the skimage library to obtain the edge coordinate point set; The signal segmentation subunit is used to determine the positions of the images corresponding to different signal channels according to the edge coordinate point set, and segment the overall image signals of different channels by using ROI operation to realize the division of a single image signal path into different independent signal paths; The data preprocessing unit is used to preprocess the separated physiological information data; The fatigue determination unit is used to analyze the network characteristics of the pre-processed physiological information data to determine the fatigue state of the user; The fatigue determination unit includes: a data set construction subunit, a model construction subunit, a training subunit, a mental fatigue state recognition subunit and a muscle fatigue state recognition subunit; The data set construction subunit is used to construct a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and the corresponding fatigue levels; The model building subunit is used to build a neural network algorithm model; The training subunit is used to train the neural network algorithm model using the data set to obtain a mental fatigue model and a muscle fatigue model; The mental fatigue state recognition subunit is used to recognize the respiratory physiological information and the heart rate physiological information by using the mental fatigue model to obtain the corresponding mental fatigue state; The muscle fatigue state recognition subunit is used to recognize the gait information using the muscle fatigue model to obtain the corresponding muscle fatigue state; The feedback interaction module is used to remind the user of the fatigue state in real time and provide status feedback.

2. The wearable sports fatigue monitoring and reminder system based on multi-channel optical signals according to claim 1 is characterized in that: The data acquisition module includes: a respiratory heart rate monitoring sensor, a gait monitoring sensor and a movable portable device; The respiration and heart rate monitoring sensor is used to collect the periodic physiological information of the user's respiration and heart rate; The gait monitoring sensor is used to collect gait information of the user; The movable portable device is used to collect periodic physiological information and gait information.

3. The wearable sports fatigue monitoring and reminder system based on multi-channel optical signals according to claim 1 is characterized in that: The feedback interaction module includes: a real-time reminder unit and a recording feedback unit; The real-time reminder unit is used to specify a preset classical Chinese to remind the user based on the user's fatigue state, and to remind the user by controlling the flash of the smartphone's flashlight and the vibration ringtone; The recording and feedback unit is used to record the exercise duration and recovery time of each user, and provide feedback to the staff after the rehabilitation exercise is completed.

4. A wearable sports fatigue monitoring and reminder method based on multi-channel optical signals, characterized in that: A wearable sports fatigue monitoring and reminder system based on multi-channel optical signals as described in any one of claims 1 to 3, the method comprising: Collect multiple types of physiological information data of the user, and use a threshold segmentation method to separate information channels of the multiple types of physiological information data; wherein the multiple types of physiological information data include: respiratory physiological information, heart rate physiological information and gait information; Separating the multiple types of physiological information data into information channels includes: Divide the mobile phone video signal into images frame by frame and perform grayscale processing on the images; The Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale image to obtain the threshold of each signal pathway. Based on the threshold of the signal pathway, the Boolean matrix corresponding to the image is obtained through mask operation; By marking the connected regions in the Boolean matrix, the label function in the skimage library was used to extract the outline of each signal pathway and obtain the edge coordinate point set; Determine the position of the image corresponding to different signal channels according to the edge coordinate point set, and use ROI operation to segment the overall image signals of different channels, so as to divide a single image signal path into different independent signal paths; Preprocessing the separated physiological information data; Based on deep learning, the signal network change characteristics of the preprocessed physiological information data are extracted to evaluate and classify the fatigue state; wherein the fatigue state includes: mental fatigue state and muscle fatigue state; Provide real-time reminders and record feedback on multiple terminals based on fatigue status determination.

5. The wearable sports fatigue monitoring and reminder method based on multi-channel optical signals according to claim 4 is characterized in that: Preprocessing of the separated physiological information data includes: Performing baseline correction on the separated physiological information data; Perform signal denoising on the corrected data; Perform sliding window segmentation processing on the denoised data; The segmented data were subjected to signal normalization.

6. The wearable sports fatigue monitoring and reminder method based on multi-channel optical signals according to claim 4 is characterized in that: Based on deep learning, the signal network change characteristics of preprocessed physiological information data are extracted to evaluate and classify fatigue status, including: Constructing a data set; wherein the data set includes: mental fatigue data, muscle fatigue data and corresponding fatigue levels; Build a neural network algorithm model; Using the data set to train the neural network algorithm model to obtain a mental fatigue model and a muscle fatigue model; Using the mental fatigue model to identify the respiratory physiological information and the heart rate physiological information, and obtaining the corresponding mental fatigue state; The muscle fatigue model is used to identify the gait information to obtain the corresponding muscle fatigue state.

7. The wearable sports fatigue monitoring and reminder method based on multi-channel optical signals according to claim 6 is characterized in that: Building a neural network algorithm model includes: Modify the input layer in the MobileNetV3 algorithm model so that the MobileNetV3 algorithm model can process one-dimensional time domain data to form the neural network algorithm model; Using the data set to train the neural network algorithm model includes: Processing the data set into a one-dimensional time-domain series; The input one-dimensional time domain sequence is amplified and sampled through convolution, and the sampled data is processed through several blocks; Overlap the feature matrices output in the blocks, then perform convolution and dimensionality reduction, use the activation function to act on the feature vector after dimensionality reduction, and complete data feature extraction; A multi-layer perception structure is introduced in the fully connected layer to complete the classification.

8. The wearable sports fatigue monitoring and reminder method based on multi-channel optical signals according to claim 4 is characterized in that: The Sauvola adaptive local binarization algorithm is used to perform threshold segmentation on the grayscale image, and the thresholds of each signal pathway are obtained, including: For each pixel in the grayscale image, the local mean is calculated in the defined neighborhood according to the pre-formula; in, represents the total number of pixels in the neighborhood, Represents the light intensity at the pixel of the corresponding coordinate in the local neighborhood, i represents the i-th pixel point processed, Represents the local pixel intensity mean in the neighborhood; Based on the local mean, obtaining the local standard deviation in the same neighborhood; in, represents the local standard deviation within the same neighborhood; Obtain the adaptive threshold of each pixel based on the pixel neighborhood mean and local standard deviation; in, Represents the adaptive threshold of the pixel (x, y), Indicates custom correction parameters, 0 <k<1, Represents the dynamic range of pixel intensity; The original intensity value of the pixel is compared with the corresponding adaptive threshold. If the original intensity value of the pixel is higher than the adaptive threshold, it is recorded as 1, otherwise it is recorded as 0. The same channel signal will be recorded as the same threshold.