Intelligent drowning alarm method and system based on electromyographic signal

Abstract

The present invention discloses an intelligent drowning alarm method and system based on electromyographic signals, including: a data acquisition module, an electromyographic signal preprocessing module, an electromyographic signal feature extraction module, an electromyographic signal classification module, an electromyographic signal state analysis module, an alarm module, an airbag ejection module and a central rescue cloud platform module; according to the characteristics of electromyographic signals that can predict the user's intention and muscle fatigue to a certain extent, the present invention collects multi-channel electromyographic signals from different muscle parts, processes and analyzes the multi-channel electromyographic signals to obtain the drowning state, and realizes high-accuracy and good real-time drowning state judgment. Based on the automatic inflatable ejection airbag, the problem of untimely rescue and low success rate of existing drowning rescue products is solved. The wearable waterproof electromyographic signal acquisition device and self-rescue device of the present invention are not restricted by scenarios, solve the problem of blind spots in open water rescue, and improve the rescue success rate.

Description

Technical Field

[0001] The present invention belongs to the field of drowning safety rescue, and in particular relates to an intelligent drowning alarm method and system based on electromyographic signals. Background Art

[0002] In recent years, swimming has become more and more popular among modern people. Swimming is the most popular form of exercise for modern people, and more and more people are choosing swimming as a sport. However, during swimming, drowning accidents often occur due to human or various unexpected factors, especially for beginners or teenagers, the chance of drowning is greatly increased. In reality, when a drowning accident occurs, it is often not that there is no guardian or lifeguard, but that it is difficult for lifeguards to discover and identify it in the first time. It is crucial to complete the rescue in the golden time. The existing lifesaving products on the market are divided into the following two categories:

[0003] (1) The first category is life-saving products with structural design

[0004] This type of product uses a manually triggered pop-up structure design. Compared with traditional life jackets and floating boards, this type of product is smaller, more portable, and has a better user experience. However, the disadvantage of this type of product is that due to its active trigger structure design, people are often unable to call for help in an emergency due to water entering the respiratory tract, and they are unable to trigger the switch to use the product to save themselves after rational thinking. This is especially true for children, who are not suitable for this type of product. In addition, the circumstances leading to drowning are more complicated. When users are injured or attacked, etc., the possibility of actively triggering the use of the product is low. In summary, this type of product has high limitations and low utilization rate.

[0005] Existing technologies, such as the technology of Chinese patent application 202011133592.6, solve the problems of small rescue protection range and low reliability of existing anti-drowning alarm bracelets: since protection relies on lifeguard patrols, the actual effective area is limited by the number of lifeguards and the range of lifeguard observation; when the drowning person sinks into the water or the drowning person does not send out an alarm signal in time, the lifeguard cannot find it in time and misses the best rescue opportunity. Its disadvantage is that on the one hand, manual control of alarm and self-rescue has high requirements for drowning people, requiring drowning people to have a certain degree of awareness and self-rescue ability. On the other hand, although it no longer relies on the number of lifeguards and the observation range of lifeguards, it requires the alarm signal to be seen by people nearby. When there is no one or few people near the drowning person, the alarm signal will not be received by people, resulting in untimely rescue.

[0006] (2) The second category is intelligent alarm system products

[0007] Most of these products are cameras that monitor swimmers for abnormal movements such as drowning, thereby determining whether they are drowning. However, because human behavior is complex and the boundaries between normal and drowning movements are unclear, misjudgments often occur. In addition, this method requires multiple cameras to be set up underwater, which is costly, and there are blind spots for monitoring such as mutual obstruction between swimmers. In summary, this intelligent alarm system based on video monitoring has obvious drawbacks.

[0008] Existing technologies such as the technology of Chinese patent application 201711240813.8, the drowning monitoring alarm system, uses an image acquisition device to acquire images in the monitored area in real time and transmits them to a processing unit. The processing unit acquires the behavior of all dynamic targets in the monitored area based on the images in the monitored area. When the behavior of the dynamic target exceeds the preset range of safe behavior, the processing unit generates an alarm message and transmits it to the wireless communication device, and at the same time controls the alarm device to play the alarm message.

[0009] However, existing vision-based technologies need to detect key points in the video and extract the features of key points to identify human behavior. Some of the key points are generated by background noise, which has a large environmental impact and poor robustness. In addition, the algorithm is highly complex and estimates the human posture through key point detection to determine the swimming state, which has poor real-time performance. This system requires a large number of cameras and has high installation costs. It is not suitable for monitoring in open waters and will cause blind spots in monitoring, resulting in delayed rescue. The above problems reduce the practicality of this method.

[0010] In summary, the existing technology has the problems of low judgment accuracy, small scope of application, and poor real-time performance. Studies have shown that regular swimming is one of the best ways to exercise muscles throughout the body. The greater the resistance during swimming, the more it can stimulate the cerebral cortex, and then reflexively adjust more muscle groups to move, so that the muscles of the whole body can be trained in an orderly and complete manner, especially for the pectoralis major, deltoids, triceps and upper back muscles. When a person drowns, the muscles will be tense, and all nerve cells will send out instructions, causing muscle contraction. The surface electromyography signal is a complex electrical signal generated by muscle contraction, which contains rich human movement information. Therefore, judging the drowning state by the physiological electrical signals generated by the muscles during drowning has the characteristics of low latency, high accuracy and wide application scenarios. Therefore, this paper proposes an intelligent drowning alarm device and intelligent drowning alarm method based on electromyography signals. Summary of the invention

[0011] In order to solve the problems existing in the prior art, the present invention provides the following intelligent drowning alarm method and system based on electromyographic signals, which can replace traditional drowning rescue products to achieve rapid and accurate rescue. The real-time monitoring of the user's swimming status can be achieved through the collection, processing and analysis technology of electromyographic signals of different muscle parts, so as to obtain the user's drowning status, and the rescue of drowning people can be achieved by the automatic pop-up airbag and cloud platform. The problems of low success rate of current drowning rescue products, low accuracy of drowning status judgment, untimely rescue in open waters, and rescue blind spots can be solved.

[0012] According to a first aspect of the present invention, there is provided an intelligent drowning alarm system based on electromyographic signals, comprising: a data acquisition module, an electromyographic signal preprocessing module, an electromyographic signal feature extraction module, an electromyographic signal classification module, an electromyographic signal state analysis module, an alarm module, an airbag ejection module and a central rescue cloud platform module;

[0013] The data acquisition module is used to collect the electromyographic signals of the user when swimming in real time, collect signals from pre-set muscle sites through waterproof electromyographic electrodes, and obtain multiple electromyographic signals generated by the user when swimming in real time through wireless transmission; the multiple electromyographic signals are transmitted to the electromyographic signal preprocessing module.

[0014] The electromyographic signal preprocessing module is used to eliminate the clutter interference in the transmission process of the multi-channel electromyographic signal, and preprocess the multi-channel electromyographic signal to remove noise through a preprocessing algorithm; and transmit the multi-channel electromyographic signal with noise removed to the electromyographic signal feature extraction module.

[0015] The electromyographic signal feature extraction module is used to obtain the time domain and frequency domain features of the electromyographic signal action segment, obtain the position of the electromyographic signal action segment by the signal strength of the electromyographic signal action segment being greater than the signal strength of the rest segment, obtain the time domain features of the electromyographic signal action segment using the time domain feature extraction algorithm, obtain the frequency domain features of the electromyographic signal action segment using the frequency domain feature extraction algorithm, and fuse the time domain features and frequency domain features of the electromyographic signal action segment to obtain the multi-channel electromyographic signal fusion features.

[0016] The electromyographic signal classification module is used to calculate the classification results of the multiple electromyographic signals according to the fusion features of the multiple electromyographic signals, and the classification results are normal or abnormal. The fusion features of the multiple electromyographic signals are respectively input into a pre-trained classification network, and the classification network processes and identifies the fusion features of the multiple electromyographic signals to obtain the classification results of the multiple electromyographic signals.

[0017] The electromyographic signal state analysis module is used to combine and analyze the classification results of multiple electromyographic signal features, and to judge the user's swimming state according to the output classification results. The electromyographic signal state analysis module is connected to the alarm module, and the alarm module is respectively connected to the airbag ejection module and the central rescue cloud platform module.

[0018] Furthermore, the alarm module is used to sound an alarm. The alarm module includes a manual shutdown function. If the alarm is manually turned off within a preset time, it means that the user is not drowning and it is a system misjudgment, so the alarm stops sounding.

[0019] Furthermore, the control trigger device of the airbag ejection module opens the carbon dioxide cylinder, and the carbon dioxide gas fills the airbag, enabling the drowning person to save himself.

[0020] Furthermore, the central rescue cloud platform module is used to analyze the alarm information (including GPS positioning information and comprehensive personal information) and dispatch a rescue team according to the alarm information sent by the embedded processor of the smart wrist rescue device, so as to achieve accurate rescue of the drowning person.

[0021] The second aspect of the present invention provides an intelligent drowning alarm method based on electromyographic signals, comprising: real-time acquisition of electromyographic signals of swimmers by electromyographic signal acquisition equipment to obtain physiological electrical signals of people when swimming, obtaining the swimming status of users by electromyographic signal processing and analysis technology, and controlling the alarm and airbag according to the category of abnormal swimming status, and transmitting the alarm information to the central rescue cloud platform at the same time to realize the rescue of drowning people, solving the problems of low success rate of drowning rescue products, low accuracy of drowning status judgment, untimely rescue in open waters, and rescue blind spots.

[0022] Compared with the prior art, the intelligent drowning alarm method and system based on electromyographic signals provided by the present invention has the following beneficial effects:

[0023] Based on the characteristics of electromyographic signals that can predict the user's intention and muscle fatigue to a certain extent, by collecting multi-channel electromyographic signals from different muscle parts, the multi-channel electromyographic signals are processed and analyzed to obtain the drowning status, and a high-accuracy and real-time drowning status judgment is achieved. Based on the automatic inflatable pop-up airbag, the problem of untimely rescue and low success rate of existing drowning rescue products is solved. The wearable waterproof electromyographic signal acquisition device and self-rescue device of the present invention are not limited by the scene, solve the problem of blind spots in open water rescue, and improve the rescue success rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A representative diagram of an intelligent drowning alarm system based on electromyographic signals provided by the present invention.

[0025] Figure 2 A flowchart of an electromyographic signal preprocessing module provided by an embodiment of the present invention.

[0026] Figure 3 A flowchart of an electromyographic signal feature extraction module provided by an embodiment of the present invention.

[0027] Figure 4 A flowchart of an electromyographic signal classification module provided by an embodiment of the present invention.

[0028] Figure 5 A flowchart of an electromyographic signal state analysis module provided by an embodiment of the present invention.

[0029] Figure 6 A flowchart of an alarm information processing module for a wearable device provided by an embodiment of the present invention.

[0030] Figure 7 A flow chart of an airbag ejection module provided by an embodiment of the present invention.

[0031] Figure 8 A flow chart of a central rescue cloud platform module provided by an embodiment of the present invention.

[0032] Fig. 9 A classification network architecture diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0034] The method provided by the present invention can be implemented in the following terminal environment, and the terminal can include one or more of the following components: an electromyography acquisition device and an intelligent wrist rescue device.

[0035] The myoelectric acquisition device consists of an electromyographic sensor and a wireless transmission module. The electromyographic sensor is used to collect human muscle electrical signals by attaching it to the skin surface, and the wireless transmission module is used to transmit the collected human muscle electrical signals to the smart wrist rescue device in real time for subsequent processing.

[0036] The smart wrist rescue device consists of an alarm, an airbag, an embedded processor, and a wireless transmission module. The smart wrist rescue device is fixed on the user's wrist. The embedded processor is used to execute various functions of the terminal and process data; the wireless transmission module is used to receive the human muscle electrical signals collected by the electromyography acquisition device and transmit them to the embedded processor, which is used to transmit the drowning user alarm information to the rescue cloud platform. The airbag has a reliable water buoyancy function and is used for self-rescue from drowning; the alarm can be used to send a drowning signal to prompt rescuers of the drowning person's location information.

[0037] In addition, those skilled in the art can understand that the structure of the above terminal does not constitute a limitation on the terminal, and the terminal may include more or fewer components, or combine certain components, or arrange the components differently. For example, the terminal also includes an input unit, a radio frequency circuit, a sensor, a power supply, and other components, which will not be described in detail here.

[0038] In order to make the purpose, technical scheme and advantages of the embodiment of the present invention clearer, the technical scheme in the embodiment of the present invention will be clearly and completely described in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The present invention is described in detail below through specific embodiments.

[0039] In a first aspect of this embodiment, an intelligent drowning alarm method based on electromyographic signals is provided, as shown in the schematic diagram Figure 1 As shown, the following steps are included:

[0040] Step S10: multi-channel electromyographic signal acquisition module;

[0041] Step S20: EMG signal preprocessing module;

[0042] Step S30: myoelectrical features extraction module;

[0043] Step S40: EMG signal classification module;

[0044] Step S50: EMG signal state analysis module;

[0045] Step S60: alarm module;

[0046] Step S70: airbag ejection module;

[0047] Step S80: Central rescue cloud platform module;

[0048] The relevant signal processing in steps S20-S70 of this embodiment is completed by the embedded processor in the smart wrist rescue device, and the embedded processor is integrated with the airbag and the alarm. The embedded processor can be a hardware device such as FPGA, Raspberry Pi, or Jetson series edge device.

[0049] In the data acquisition module step S10 of the present embodiment, a multi-channel electromyographic signal of the user while swimming is acquired by using an electromyographic acquisition device, and the multi-channel electromyographic signal is transmitted to step S20 for processing by wireless transmission.

[0050] The type of myoelectric acquisition equipment in step S10 of the data acquisition module of this embodiment can be one or more of a dry electrode myoelectric sensor, a wet-electrode myoelectric sensor, a stick-on myoelectric sensor, a strap-on myoelectric sensor, etc. The acquisition locations of multiple myoelectric signals can be the wrist, upper arm, lower arm, thigh, calf, front chest, back, etc., and the waterproof myoelectric electrode is integrated into the swimsuit. The wireless transmission method includes Bluetooth, 4G, 5G, Wifi, etc. This embodiment is only illustrated by collecting arm electromyographic signals and respiratory electromyographic signals and 5G transmission methods as an example. Collect arm electromyographic signals and respiratory electromyographic signals during swimming, and transmit the collected electromyographic signals to step S20 via 5G wireless.

[0051] In this embodiment, the electromyographic signal preprocessing module step S20: receives multiple electromyographic signals from step S10, and performs preprocessing operations on the multiple electromyographic signals to reduce noise, so as to eliminate noise in the signal acquisition and transmission process.

[0052] The preprocessing algorithm in the electromyographic signal preprocessing module step S20 of this embodiment can adopt wavelet transform, filtering, blind signal separation (such as ICA), etc. This embodiment performs ICA preprocessing operation on the multi-channel electromyographic signals. Specifically includes the following steps:

[0053] Step S210: receiving the multiple electromyographic signals collected by the waterproof electromyographic electrodes in step S10.

[0054] Step S220: Use ICA independent component analysis algorithm to pre-process the multi-channel electromyographic signal to remove noise. First, the multi-channel electromyographic signal data is centered, and then whitening and decorrelation operations are performed on it to obtain independent multi-channel electromyographic signals. The calculation method is shown in formula (1):

[0055]

[0056] Eigenvalue decomposition is used to obtain E (eigenvector matrix) and D (eigenvalue diagonal matrix).

[0057] Finally, according to formula (2), the denoised electromyographic signal is obtained.

[0058]

[0059] Where w represents the unmixing matrix for blind source signal separation.

[0060] Step S230: Obtain the preprocessed multi-channel electromyographic data from step S220 and input it into step S30.

[0061] In this embodiment, the electromyographic signal feature extraction module step S30 receives the multi-channel electromyographic signal data preprocessed in the step S20, and uses a feature extraction algorithm to perform time domain and frequency domain feature analysis on the preprocessed multi-channel electromyographic signal data.

[0062] The electromyographic signal feature extraction module step S30 of this embodiment can extract time domain features such as peak value, mean value, variance, number of zero crossing points, etc., and can extract frequency domain features such as median frequency, mean frequency, etc. This embodiment extracts time domain amplitude and frequency domain spectrum for analysis. Specifically, it includes the following steps:

[0063] Step S310: Obtain the pre-processed multi-channel electromyographic signal data from step S230.

[0064] Step S320: According to formula (3), the electromyographic data of each channel in step S310 are summed to form a new channel.

[0065]

[0066] Among them, E sum (t) represents the result after multi-channel superposition, and n represents the number of channels. The screening window windows_1 is used to slide on the new channel with a certain step length to extract the local maximum of the electromyographic signal. sum (t) satisfies both formula (4) and formula (5), then the effective maximum value of the signal appears near time t.

[0067]

[0068]

[0069] According to formula (6), the sorting window windows_2 is used to determine the specific time when the signal maximum value appears, so as to extract the action segment of the electromyographic signal.

[0070] max{[E sum (t-windows_2),E sum (t+windows_2)]} (6)

[0071] Step S330: Using feature extraction window windows_3, extract the time domain features of the action segment in step S320, that is, the signal amplitude E in the time domain i (t).

[0072] Step S340: Using the feature extraction window windows_3, perform a fast Fourier transform operation on the signal amplitude in the time domain to obtain frequency domain features, and extract the frequency domain features of the action segment in step S320, that is, the signal spectrum value F(E i(t)).

[0073] Step S350: The time domain and frequency domain features in step S330 and step S340 are merged to form a feature matrix. First, a two-dimensional matrix of n*(4*windows_3+2) is created to store the features of a segment of action, and then the signal time domain and frequency domain features in the interval [t0-windows_3, t0+windows_3] of each channel are stored in the two-dimensional matrix to form the feature matrix.

[0074] Step S360: Obtain the time domain and frequency domain fusion features from step S350 and input them into step S40.

[0075] In this embodiment, the electromyographic signal classification module step S40: first, receives the electromyographic signal fusion feature in step S360; then processes and identifies it using a classification network; and finally obtains a multi-channel electromyographic signal feature classification result.

[0076] The electromyographic signal classification module step S40 of this embodiment can use the K nearest neighbor algorithm, linear discriminator, recurrent neural network, convolutional neural network and other classification networks to process and identify the electromyographic signal fusion features. This embodiment specifically uses a classification network composed of a shallow convolutional neural network fused with a gated recurrent unit. The network has both the ability of a convolutional neural network to effectively extract the features of the data and the strong memory of the gated recurrent unit for the timing information, which effectively ensures the accuracy of the electromyographic signal classification. Specifically includes the following steps:

[0077] Step S410: Receive the electromyographic signal fusion feature in step S360 and input it to step S420.

[0078] Step S420: Input the electromyographic signal fusion feature into the classification network composed of the shallow convolutional neural network fusion gated recurrent unit, process and identify it, and obtain the electromyographic signal feature classification result. There are two types of classification results, namely, normal swimming state and abnormal swimming state. The preferred classification network in this embodiment is as shown in the attached figure. Fig. 9 shown.

[0079] Step S430: outputting the multi-channel electromyographic signal feature classification result in step S420 to step S50.

[0080] In this embodiment, the electromyographic signal classification module step S50: receives the classification result in step S440, performs combined analysis on it, and thus determines the swimming state of the user. Specifically, it includes the following steps:

[0081] Step S510: receiving the classification result in step S440.

[0082] Step S520: performing combined analysis on the feature classification results of the multiple myoelectric signals, and determining whether each myoelectric signal is in an abnormal state.

[0083] Step S530: If the result is not a complete abnormal category, it means that the user's swimming is normal, and the process ends at this time.

[0084] Step S540: If the results are all abnormal categories, it means that the user is in an abnormal state of drowning, and an alarm signal is transmitted to step S60.

[0085] In this embodiment, the alarm module step S60 receives the alarm signal of step S530, controls the start of the alarm mode, and the alarm emits an alarm sound. If the alarm is not turned off manually within the preset time, a pop-up instruction is issued to step S70, and an alarm message is output to step S80. If the alarm is turned off manually within the preset time, the system makes a wrong judgment and the process ends.

[0086] Step S610: receiving the alarm signal of step S530.

[0087] Step S620: Process the alarm signal and activate the alarm mode.

[0088] Step S630: If the user turns off the alarm button within the system preset time, the process proceeds to step S640; if the user does not turn off the alarm button within the system preset time, the process proceeds to step S650.

[0089] Step S640: If the user turns off the alarm, it means that the user is not drowning, and the alarm stops sounding.

[0090] Step S650: If the user does not turn off the alarm, it means that the user is drowning. At this time, the airbag ejection instruction is output to step S70, and the alarm information (including GPS positioning information and personal information, wherein the personal information includes name, age, address, relatives and friends' phone numbers, etc.) is output to step S80.

[0091] In the present embodiment, the airbag pop-up module step S70: first, receives the airbag pop-up instruction outputted from the step S650, then controls the switch control mainboard in the airbag pop-up device to send an electrical signal, and finally, starts the trigger device to open the carbon dioxide cylinder, at which time the carbon dioxide gas is rapidly released to fill the airbag, and the airbag pops out successfully.

[0092] Step S710: Receive the airbag ejection instruction issued in step S650.

[0093] Step S720: Control and trigger the airbag to pop out.

[0094] Step S730: The trigger device opens the carbon dioxide cylinder

[0095] Step S740: The carbon dioxide gas is released to quickly fill the airbag, causing the airbag to float to the surface

[0096] In this embodiment, the central rescue cloud platform module step S80: according to the alarm information (including GPS positioning information and comprehensive personal information), the central rescue cloud platform is notified that a drowning person has appeared, and a rescue team is dispatched to the drowning area for rescue. The specific steps are as follows:

[0097] Step S810: Obtain alarm information from step S680, including GPS positioning information and comprehensive information of the drowning person, the comprehensive information including name, age, gender, occupation, place of origin, etc.

[0098] Step S820: The cloud platform processes the GPS positioning information and the comprehensive information of the drowning person, converts them into alarm control commands for controlling the alarm of the rescue platform, and displays the detailed positioning information of the drowning person.

[0099] Step S830: The rescue platform alarm sounds to remind the rescue platform that someone is drowning.

[0100] Step S840: According to the alarm information of the drowning person, a rescue team is dispatched to carry out on-site rescue.

[0101] A second aspect of this embodiment provides an intelligent drowning alarm system based on electromyographic signals, comprising:

[0102] The data acquisition module is used to collect the electromyographic signals of the user when swimming in real time. The waterproof electromyographic electrodes are used to collect signals from pre-set muscle sites, and the multi-channel electromyographic signals generated by the user when swimming are obtained in real time by wireless transmission.

[0103] The electromyographic signal preprocessing module is used to eliminate the clutter interference in the transmission process of the multi-channel electromyographic signal. The multi-channel electromyographic signal is preprocessed by a preprocessing algorithm to remove noise.

[0104] The EMG signal feature extraction module is used to obtain the time domain and frequency domain features of the EMG signal action segment. The position of the EMG signal action segment is obtained by the signal strength of the EMG signal action segment being greater than the signal strength of the rest segment. The time domain features of the EMG signal action segment are obtained by using the time domain feature extraction algorithm. The frequency domain features of the EMG signal action segment are obtained by using the frequency domain feature extraction algorithm. The time domain features and frequency domain features of the EMG signal action segment are fused to obtain the fusion features of the multi-channel EMG signals.

[0105] The electromyographic signal classification module is used to calculate the classification results of the multiple electromyographic signals according to the fusion features of the multiple electromyographic signals, and the classification results are normal or abnormal. The fusion features of the multiple electromyographic signals are respectively input into a pre-trained classification network, and the classification network processes and identifies the fusion features of the multiple electromyographic signals to obtain the classification results of the multiple electromyographic signals.

[0106] The electromyographic signal state analysis module is used to perform combined analysis on the classification results of multiple electromyographic signal features, and to determine the user's swimming state based on the output classification results.

[0107] The alarm module is used to sound an alarm. The module includes a manual shutdown function. If the alarm is turned off manually within a preset time, it means that the user is not drowning and it is a system misjudgment, then the alarm stops sounding.

[0108] The airbag ejection module controls the trigger device to open the carbon dioxide cylinder, and the carbon dioxide gas fills the airbag, enabling the drowning person to save himself.

[0109] The central rescue cloud platform module is used to analyze the alarm information (including GPS positioning information and comprehensive personal information) and dispatch a rescue team according to the alarm information sent by the embedded processor of the smart wrist rescue device, so as to achieve accurate rescue of drowning people.

[0110] This embodiment collects electromyographic signals of swimmers in real time through electromyographic signal collection equipment to obtain physiological electrical signals of people when swimming. The swimming status of users is obtained through electromyographic signal processing and analysis technology, and the alarm and airbag are controlled according to the category of abnormal swimming status. At the same time, the alarm information is transmitted to the central rescue cloud platform to realize the rescue of drowning people, solve the problems of low success rate of drowning rescue products, low accuracy of drowning status judgment, untimely rescue in open waters, and rescue blind spots.

[0111] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent drowning alarm system based on electromyographic signals, characterized in that: include: Data acquisition module, EMG signal preprocessing module, EMG signal feature extraction module, EMG signal classification module, EMG signal status analysis module, alarm module, airbag ejection module and central rescue cloud platform module; The data acquisition module is used to collect the electromyographic signals of the user when swimming in real time, collect signals from pre-set muscle sites through waterproof electromyographic electrodes, and obtain multiple electromyographic signals generated by the user when swimming in real time through wireless transmission; Multiple channels of electromyographic signals are transmitted to the electromyographic signal preprocessing module; The electromyographic signal preprocessing module is used to eliminate the clutter interference in the transmission process of the multi-channel electromyographic signal, and preprocess the multi-channel electromyographic signal to remove noise through a preprocessing algorithm; and transmitting the multi-channel electromyographic signals with noise removed to the electromyographic signal feature extraction module; The electromyographic signal feature extraction module is used to obtain the time domain and frequency domain features of the electromyographic signal action segment, obtain the position of the electromyographic signal action segment by the signal strength of the electromyographic signal action segment being greater than the signal strength of the rest segment, obtain the time domain features of the electromyographic signal action segment using the time domain feature extraction algorithm, obtain the frequency domain features of the electromyographic signal action segment using the frequency domain feature extraction algorithm, and fuse the time domain features and frequency domain features of the electromyographic signal action segment to obtain the multi-channel electromyographic signal fusion features; The electromyographic signal classification module is used to calculate the classification results of the multiple electromyographic signals according to the fusion features of the multiple electromyographic signals, and the classification results are normal or abnormal; The fusion features of the multiple electromyographic signals are respectively input into a pre-trained classification network, and the classification network processes and identifies the fusion features of the multiple electromyographic signals to obtain classification results of the multiple electromyographic signals; The electromyographic signal state analysis module is used to perform combined analysis on the classification results of multiple electromyographic signal features, and to judge the user's swimming state based on the output classification results; The electromyographic signal state analysis module is connected to the alarm module, and the alarm module is connected to the airbag ejection module and the central rescue cloud platform module respectively; The EMG signal preprocessing algorithm uses wavelet transform, filtering or blind signal separation; Performing ICA preprocessing operation on the multiple electromyographic signals; The following steps are involved: Step S210: receiving multiple electromyographic signals collected by the waterproof electromyographic electrodes; Step S220: preprocessing the multi-channel electromyographic signals to remove noise using an ICA independent component analysis algorithm; First, the multi-channel electromyographic signal data is centered, and then a whitening and decorrelation operation is performed on it to obtain independent multi-channel electromyographic signals. The calculation method is shown in formula (1): The eigenvalue decomposition is used to obtain the E eigenvector matrix and the D eigenvalue diagonal matrix; Finally, according to formula (2), the denoised electromyographic signal is obtained; Where w represents the unmixing matrix for blind source signal separation; Step S230: Obtain the pre-processed multi-channel electromyographic data from the step S220 and input it into step S30; Step S30 extracts the time domain amplitude and frequency domain spectrum for analysis; specifically includes the following steps: Step S310: Acquire the pre-processed multi-channel electromyographic signal data from the step S230; Step S320: According to formula (3), summing the electromyographic data of each channel in step S310 to form a new channel; Among them, E sum (t) represents the result after multi-channel superposition, n represents the number of channels; the screening window windows_1 is used to slide on the new channel with a certain step length to extract the local maximum of the electromyographic signal; if E sum (t) satisfies both formula (4) and formula (5), then the effective maximum value of the signal appears near time t; According to formula (6), the sorting window windows_2 is used to determine the specific time when the signal maximum value appears, so as to extract the action fragment of the electromyographic signal; max{[E sum (t-windows_2),E sum (t+windows_2)]} (6) Step S330: Using feature extraction window windows_3, extract the time domain features of the action segment in step S320, that is, the signal amplitude E in the time domain i (t); Step S340: Using the feature extraction window windows_3, perform a fast Fourier transform operation on the signal amplitude in the time domain to obtain frequency domain features, and extract the frequency domain features of the action segment in step S320, that is, the signal spectrum value F(E i (t)); Step S350: Fusing the time domain and frequency domain features in step S330 and step S340 to form a feature matrix; first, creating a two-dimensional matrix of n*(4*windows_3+2) to store the features of a segment of action, and then storing the time domain and frequency domain features of the signal in the interval [t0-windows_3, t0+windows_3] of each channel into the two-dimensional matrix to form the feature matrix; Step S360: Obtain the time domain and frequency domain fusion features from step S350, and input them into step S40; Step S40 uses a K-nearest neighbor algorithm, a linear discriminator, a recurrent neural network, and a convolutional neural network classification network to process and identify the fusion features of the electromyographic signal; specifically, the following steps are included: Step S410: receiving the electromyographic signal fusion feature in step S360, and inputting it into step S420; Step S420: inputting the electromyographic signal fusion feature into a classification network composed of a shallow convolutional neural network fused with a gated recurrent unit, processing and identifying it, and obtaining the electromyographic signal feature classification result; the classification result has two categories, namely, normal swimming state and abnormal swimming state; Step S430: outputting the multi-channel electromyographic signal feature classification result in step S420 to step S50; Step S50 receives the classification results in step S430 and performs combined analysis on them to determine the swimming state of the user; specifically, it includes the following steps: Step S510: receiving the classification result in step S430; Step S520: performing combined analysis on the feature classification results of the multiple myoelectric signals, and determining whether each myoelectric signal is in an abnormal state individually; Step S530: If the result is not a complete abnormal category, it means that the user's swimming is normal, and the process ends at this time; Step S540: If the results are all abnormal categories, it means that the user is in an abnormal state of drowning, and an alarm signal is transmitted to step S60.

2. The intelligent drowning alarm system based on electromyographic signals according to claim 1 is characterized in that: The alarm module is used to sound an alarm. The alarm module includes a manual shutdown function. If the alarm is manually turned off within the preset time, it means that the user is not drowning and it is a system misjudgment, then the alarm stops sounding.

3. The intelligent drowning alarm system based on electromyographic signals according to claim 1 is characterized in that: The control trigger device of the airbag ejection module opens the carbon dioxide cylinder, and the carbon dioxide gas fills the airbag, enabling the drowning person to save himself.

4. The intelligent drowning alarm system based on electromyographic signals according to claim 1 is characterized in that: The central rescue cloud platform module is used to analyze the alarm information and dispatch rescue teams according to the alarm information sent by the embedded processor to achieve accurate rescue of drowning people.

5. An intelligent drowning alarm method based on electromyographic signals using the system according to any one of claims 1 to 4, characterized in that: The steps include: Step S10: collecting multiple channels of electromyographic signals; Step S20: preprocessing of electromyographic signals; Step S30: extracting electromyographic features; Step S40: classifying electromyographic signals; Step S50: EMG signal status analysis; Step S60: the alarm module is started; Step S70: the airbag ejection module is activated; Step S80: The central rescue cloud platform module is started.

6. The intelligent drowning alarm method based on electromyographic signals according to claim 5, characterized in that: Step S60: receiving the alarm signal of step S530, controlling the alarm mode to be started, the alarm emits an alarm sound, and if the alarm is not manually turned off within the preset time, a pop-up instruction is issued to step S70, and an alarm message is output to step S80 at the same time. If the alarm is manually turned off within the preset time, the system makes a judgment error; Step S610: receiving the alarm signal of step S530; Step S620: Process the alarm signal and start the alarm mode; Step S630: If the user turns off the alarm button within the system preset time, the process proceeds to step S640; if the user does not turn off the alarm button within the system preset time, the process proceeds to step S650; Step S640: If the user turns off the alarm, it means that the user is not drowning, and the alarm stops sounding. Step S650: If the user does not turn off the alarm, it means that drowning has occurred; at this time, the airbag ejection instruction is output to step S70, and the alarm information is output to step S80; Step S70: First, receiving the airbag ejection instruction outputted in step S650, then controlling the switch control mainboard in the airbag ejection device to send an electrical signal, and finally starting the trigger device to open the carbon dioxide cylinder, at which point the carbon dioxide gas is rapidly released to fill the airbag, and the airbag is ejected successfully; Step S710: receiving the airbag ejection instruction issued in step S650; Step S720: controlling and triggering the airbag to pop out; Step S730: The trigger device opens the carbon dioxide cylinder Step S740: The carbon dioxide gas is released to quickly fill the airbag, causing the airbag to float to the surface Step S80: notify the central rescue cloud platform of the drowning person according to the alarm information, and send a rescue team to the drowning area for rescue; the specific steps are as follows: Step S810: Acquire alarm information from step S650, including GPS positioning information and comprehensive information of the drowning person; Step S820: The cloud platform processes the GPS positioning information and the comprehensive information of the drowning person, converts them into an alarm control command for controlling the alarm of the rescue platform, and displays the detailed positioning information of the drowning person; Step S830: The rescue platform alarm sounds an alarm to remind the rescue platform that someone is drowning; Step S840: According to the alarm information of the drowning person, a rescue team is dispatched to carry out on-site rescue.

7. The intelligent drowning alarm method based on electromyographic signals according to claim 5, characterized in that: The type of the electromyographic acquisition device in step S10 is one or more of a dry electrode electromyographic sensor, a wet electrode electromyographic sensor, an adhesive electromyographic sensor, and a strap-type electromyographic sensor. The multi-channel electromyographic signal is collected at the wrist, upper arm, forearm, thigh, calf, front chest, and back, and the waterproof electromyographic electrode is integrated into the swimsuit.

Images