CN112818773A - Heart rate detection method and device and storage medium - Google Patents
Heart rate detection method and device and storage medium Download PDFInfo
- Publication number
- CN112818773A CN112818773A CN202110072811.2A CN202110072811A CN112818773A CN 112818773 A CN112818773 A CN 112818773A CN 202110072811 A CN202110072811 A CN 202110072811A CN 112818773 A CN112818773 A CN 112818773A
- Authority
- CN
- China
- Prior art keywords
- heart rate
- network model
- preset
- parameters
- user information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 59
- 230000000306 recurrent effect Effects 0.000 claims abstract description 53
- 238000003062 neural network model Methods 0.000 claims abstract description 49
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 21
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 210000002569 neuron Anatomy 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 11
- 125000004122 cyclic group Chemical group 0.000 description 10
- 238000004590 computer program Methods 0.000 description 7
- 230000033764 rhythmic process Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000001133 acceleration Effects 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 210000000707 wrist Anatomy 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 230000003862 health status Effects 0.000 description 2
- 238000010030 laminating Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012536 packaging technology Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a heart rate detection method, a device and a storage medium, wherein the method comprises the following steps: acquiring user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are detected by a heart rate sensor in the wearable device; and determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate parameters collected by a heart rate sensor as input, and the heart rate value collected by a heart rate belt is obtained as output training. The technical problem of low heart rate detection accuracy is solved, and the heart rate detection accuracy is improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a heart rate detection method, heart rate detection equipment and a storage medium.
Background
The intelligent wearable product or the consumer electronics product has certain limitations when detecting the heart rate of a human body, the traditional dynamic heart rate detection algorithm has the defects of low measurement and calculation precision, low robustness and poor anti-interference performance, and especially the dynamic heart rate of the human body under the motion condition is influenced by multiple factors, such as: the inaccurate problem of rhythm of the heart resolving can all be caused to the speed of motion, wrist inclination, bracelet and wrist laminating etc. not compact, and the different people of unable accurate measurement are the rhythm of the heart under different environment to make people to self rhythm of the heart and health status appear judging the deviation.
Disclosure of Invention
The embodiment of the application provides a heart rate detection method, a device and a medium, and aims to solve the technical problem that in the prior art, the heart rate detection accuracy is low.
In order to achieve the above object, an embodiment of the present application provides a heart rate detection method, including the following steps:
acquiring user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are detected by a heart rate sensor in the wearable device;
and determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate value acquired by the heart rate sensor as input, and the heart rate acquired by the heart rate belt is obtained as output training.
Optionally, the training process of the pre-stored recurrent neural network model includes:
acquiring user information, motion parameters, heart rate parameters and heart rate values acquired by a heart rate band from an acquisition sample library;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate band as output values of the preset network model so as to train the preset network model;
acquiring the accuracy of a preset network model in the training process;
stopping training when the accuracy reaches a preset accuracy;
and taking the preset network model obtained by current training as the recurrent neural network model.
Optionally, the step of taking the user information, the exercise parameter, and the heart rate parameter as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of the preset network model, so as to train the preset network model includes:
normalizing the user information, the motion parameters and the heart rate parameters in the sample library;
and taking the user information, the motion parameters and the heart rate parameters after the normalization processing as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of a recurrent neural network so as to train the preset network model.
Optionally, the step of taking the user information, the motion parameter, and the heart rate parameter as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of a recurrent neural network, so as to train the preset network model includes:
acquiring preset parameters and preset algorithms corresponding to a preset network model;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate zone as output values of the preset network model;
training the preset network model after the input value and the output value are set according to the preset parameters and a preset algorithm, wherein the preset parameters comprise the number of preset neuron nodes, the number of preset network layers, a preset initial value, a preset learning rate, a preset neuron discarding value and weight and a preset training period of the preset network model, and the preset algorithm is an adaptive moment estimation algorithm.
Optionally, the step of obtaining the preset parameter and the preset algorithm corresponding to the preset network model further includes:
acquiring a preset regular item corresponding to a preset network model;
and adding the preset regular term to a loss function of the preset network model.
Optionally, the input layer, the hidden layer, and the output layer of the preset network model are all connected or partially connected, each node in the layer of the preset network model is connected to each other, the input of the hidden layer includes an input value of the input layer and an output value of the hidden layer at the previous time, and weight values between neurons in the preset network model are shared.
Optionally, the step of using the user information, the exercise parameter, and the heart rate parameter as input values of a preset network model further includes:
acquiring an output value of current training of a hidden layer in the preset network model;
and taking the output value of the current training of the hidden layer, the user information, the motion parameter and the heart rate parameter as the input values of the preset network model.
Optionally, after the step of obtaining the user information, the exercise parameter, and the heart rate parameter of the user wearing the wearable device, the method further includes:
adding user information, motion parameters, and heart rate parameters of a user of the wearable device to the sample library.
In order to achieve the above object, an embodiment of the present application provides a heart rate detection apparatus, which includes a memory, a processor, and a heart rate detection program stored in the memory and executable on the processor, and when executed by the processor, the heart rate detection program implements the heart rate detection method as described above.
To achieve the above object, an embodiment of the present application provides a computer-readable storage medium storing a heart rate detection program, which when executed by a processor implements the heart rate detection method as described above.
According to the heart rate detection method, the heart rate detection device and the heart rate detection medium provided by the embodiment of the invention, the heart rate detection device acquires user information, motion parameters and heart rate parameters of a user wearing the heart rate detection device, inputs the user information, the motion parameters and the heart rate parameters into a pre-stored trained recurrent neural network model, and outputs a heart rate value of the user through the recurrent neural network model, the recurrent neural network takes the user information, the motion parameters and the heart rate parameters acquired by a heart rate sensor as input, and the heart rate value acquired by a heart rate belt is obtained by output training. Thus, parameters collected by the heart rate detection equipment are input into the trained recurrent neural network model to obtain the heart rate value of the user, and the heart rate detection accuracy is improved.
Drawings
FIG. 1 is a schematic diagram of an apparatus architecture of a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a first embodiment of a heart rate detection method according to the invention;
FIG. 3 is a flowchart illustrating a second embodiment of a heart rate detection method according to the present invention;
fig. 4 is a schematic structural diagram of the preset network model after being expanded according to time.
Detailed Description
The intelligent wearable product or the consumer electronics product has certain limitations when detecting the heart rate of a human body, the traditional dynamic heart rate detection algorithm has the defects of low measurement and calculation precision, low robustness and poor anti-interference performance, and especially the dynamic heart rate of the human body under the motion condition is influenced by multiple factors, such as: the inaccurate problem of rhythm of the heart resolving can all be caused to the speed of motion, wrist inclination, bracelet and wrist laminating etc. not compact, and the different people of unable accurate measurement are the rhythm of the heart under different environment to make people to self rhythm of the heart and health status appear judging the deviation. In order to solve the above problems, the present invention provides a heart rate detection method, which comprises the following steps: acquiring user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are detected by a heart rate sensor in the wearable device; and determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate parameters collected by a heart rate sensor as input, and the heart rate value collected by a heart rate belt is obtained as output training. The accuracy rate of heart rate detection is improved.
For a better understanding of the above technical solutions, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
As an implementation, the heart rate detection device may be as shown in fig. 1.
The embodiment of the invention relates to heart rate detection equipment, which comprises: a processor 101, e.g. a CPU, a memory 102, a communication bus 103. Wherein a communication bus 103 is used for enabling the connection communication between these components.
The memory 102 may be a high-speed RAM memory or a non-volatile memory (e.g., a disk memory). As shown in fig. 1, a heart rate detection program may be included in the memory 102, which is a type of computer storage medium; and the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
acquiring user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are detected by a heart rate sensor in the wearable device;
and determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate value acquired by the heart rate sensor as input, and the heart rate acquired by the heart rate belt is obtained as output training.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
acquiring user information, motion parameters, heart rate parameters and heart rate values acquired by a heart rate band from an acquisition sample library;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate band as output values of the preset network model so as to train the preset network model;
acquiring the accuracy of a preset network model in the training process;
stopping training when the accuracy reaches a preset accuracy;
and taking the preset network model obtained by current training as the recurrent neural network model.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
normalizing the user information, the motion parameters and the heart rate parameters in the sample library;
and taking the user information, the motion parameters and the heart rate parameters after the normalization processing as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of a recurrent neural network so as to train the preset network model.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
acquiring preset parameters and preset algorithms corresponding to a preset network model;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate zone as output values of the preset network model;
training the preset network model after the input value and the output value are set according to the preset parameters and a preset algorithm, wherein the preset parameters comprise the number of preset neuron nodes, the number of preset network layers, a preset initial value, a preset learning rate, a preset neuron discarding value and weight and a preset training period of the preset network model, and the preset algorithm is an adaptive moment estimation algorithm.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
acquiring a preset regular item corresponding to a preset network model;
and adding the preset regular term to a loss function of the preset network model.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
acquiring an output value of current training of a hidden layer in the preset network model;
and taking the output value of the current training of the hidden layer, the user information, the motion parameter and the heart rate parameter as the input values of the preset network model.
In one embodiment, the processor 101 may be configured to invoke a heart rate detection program stored in the memory 102 and perform the following operations:
adding user information, motion parameters, and heart rate parameters of a user of the wearable device to the sample library.
According to the scheme, the heart rate detection device acquires user information, motion parameters and heart rate parameters of a user wearing the heart rate detection device, the user information, the motion parameters and the heart rate parameters are input into a pre-stored trained recurrent neural network model, the heart rate value of the user is output through the recurrent neural network model, the recurrent neural network takes the user information, the motion parameters and the heart rate parameters collected by a heart rate sensor as input, and the heart rate value collected by a heart rate belt is obtained by output training. Thus, parameters collected by the heart rate detection equipment are input into the trained recurrent neural network model to obtain the heart rate value of the user, and the heart rate detection accuracy is improved.
Based on the hardware architecture of the heart rate detection device, the embodiment of the heart rate detection method is provided.
Referring to fig. 2, fig. 2 is a first embodiment of the heart rate detection method of the present invention, which includes the following steps:
step S10, obtaining user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are obtained through detection of a heart rate sensor in the wearable device;
optionally, user information of a user of the wearable device, the motion parameters and the heart rate parameters are added to a sample library for use as input values for a preset network model.
The center rate detection device in this embodiment may be an intelligent wearable device, a user registration/login interface, an acceleration sensor, a gyroscope and a heart rate sensor are arranged in the wearable device, and user information is acquired through user registration information, and includes user height, weight, age and the like; the motion parameters are acquired through an acceleration sensor and a gyroscope and comprise an acceleration value and an angular velocity value in the motion process of a user; the heart rate parameters are acquired by a heart rate sensor and comprise an electrocardiogram and the like. The acceleration sensor, the gyroscope and the heart rate sensor are embedded into the intelligent wearable device through the Sip system packaging technology.
And step S20, determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate parameters collected by the heart rate sensor as input, and the heart rate value collected by the heart rate belt is obtained as output training.
Optionally, the user information, the motion parameter, and the heart rate parameter are input into the pre-stored recurrent neural network model, and a target heart rate value is output through the recurrent neural network model, where the pre-stored recurrent neural network model is a model trained in advance, and the target heart rate value is the heart rate value of the user. The recurrent neural network model takes user information, motion parameters and heart rate parameters collected by a heart rate sensor as input, heart rate values collected by a heart rate band are obtained as output training, the accuracy of the recurrent neural network meets a preset accuracy, and the preset accuracy is set according to actual conditions. Because the heart rate belt conforms to the user's heart position when in use, the user's actual heart rate value can be more accurately measured. When the cyclic neural network model is obtained through training, user information, motion parameters and heart rate parameters in a sample base are used as input values of a preset network model, the preset network model is an untrained network model, a user heart rate value obtained through heart rate band detection is used as an output value of the preset network model, the preset network model is trained to obtain the cyclic neural network model, and an algorithm corresponding to the cyclic neural network model is added into wearable equipment so as to optimize a dynamic heart rate detection method, and therefore the heart rate value is calculated more accurately.
According to the scheme, the heart rate detection device acquires user information, motion parameters and heart rate parameters of a user wearing the heart rate detection device, the user information, the motion parameters and the heart rate parameters are input into a pre-stored trained recurrent neural network model, the heart rate value of the user is output through the recurrent neural network model, the recurrent neural network takes the user information, the motion parameters and the heart rate parameters collected by a heart rate sensor as input, and the heart rate value collected by a heart rate belt is obtained by output training. Thus, parameters collected by the heart rate detection equipment are input into the trained recurrent neural network model to obtain the heart rate value of the user, and the heart rate detection accuracy is improved.
Referring to fig. 3, fig. 3 is a second embodiment of the heart rate detection method of the present invention, wherein the training process of the pre-stored recurrent neural network model includes the following steps:
step S30, obtaining user information, motion parameters, heart rate parameters and heart rate values collected by a heart rate belt from an obtaining sample library;
optionally, before the step S30, the method further includes:
normalizing the user information, the motion parameters and the heart rate parameters in the sample library;
and taking the user information, the motion parameters and the heart rate parameters after the normalization processing as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of a recurrent neural network so as to train the preset network model.
It can be understood that, before inputting the user information, the exercise parameter, the heart rate parameter, and the heart rate value acquired by the heart rate zone into the preset network model, the parameters need to be preprocessed, in this embodiment, a dispersion normalization method may be used to normalize the parameters, and the formula of the dispersion normalization method is:
wherein, X*Is the value of the parameter after normalization, XiFor the original sample parameter, XmaxIs the maximum value, X, in the original sample parameterminIs the minimum value in the original sample data. After normalization processing is carried out along with the parameters, useless information in the parameters can be filtered out, the range of the parameters is unified, and the training speed of the neural network is accelerated.
Step S40, taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate zone as output values of the preset network model so as to train the preset network model;
optionally, the step S40 includes:
acquiring preset parameters and preset algorithms corresponding to a preset network model;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate zone as output values of the preset network model;
training the preset network model after the input value and the output value are set according to the preset parameters and a preset algorithm, wherein the preset parameters comprise the number of preset neuron nodes of the preset network model, the number of preset network layers, a preset initial value, a preset learning rate, a preset neuron discarding value (Dropout value and weight) and a preset training period, and the preset algorithm is an adaptive moment estimation algorithm (Adam algorithm).
The preset parameters can be adjusted according to the training condition of the preset network model, and corresponding target parameters after the training of the preset network model is finished, namely parameters corresponding to the recurrent neural network model, are used as a final whole set of algorithm model to be added into the wearable device.
Further, acquiring a preset regular term corresponding to the preset network model;
and adding the preset regular term into a loss function of the preset network model to optimize the training of the preset network model.
It should be noted that the preset network model generally includes a loss function, and the larger the loss function is, the worse the performance of the preset network model is. When the loss function approaches positive infinity, it indicates that the training is diverging and the learning rate needs to be adjusted down. When the loss function is always about 6.9, no sign of training convergence is shown, and the learning rate is tried to be increased or the weight initialization mode is modified. Over-fitting (over-fitting) often results when there is insufficient training data, or over-training (over-training). Regularization methods are a general term for a class of methods that introduce additional information (regularization terms) to the original model at this time in order to prevent overfitting and improve the generalization performance of the model. The best fit model (in the sense of minimizing the generalization error) is a large model that is properly regularized. The bias unit (bias unit), also called bias term (bias term) or intercept term (intercept term), is the intercept of the function, consistent with the meaning of b in the linear equation y — wx + b. In y-wx + b, b represents the intercept of the function on the y-axis, and the offset of the function from the origin is controlled, so that the bias unit in the neural network can accelerate the convergence of the neural network and improve the accuracy. In this embodiment, assuming that the original loss function is J (θ), the loss function added with the regularization term is J (θ) + λ R (ω), where R (ω) is an index characterizing the complexity of the model and is generally determined by a weight ω, λ represents a proportion of the loss function of the complexity of the model in the new loss function after the regularization term is added, and θ represents all parameters to be optimized in the neural network, including the weight ω and the bias term b, so that the loss function J (θ) of the optimized network model is to find a suitable weight ω and a suitable bias term b, so that J (θ) is the minimum.
Step S50, acquiring the accuracy of the preset network model in the training process;
step S60, stopping training when the accuracy reaches a preset accuracy;
and step S70, taking the preset network model obtained by current training as the recurrent neural network model.
The preset accuracy can be set to 99.5% according to actual needs, for example, it can be understood that the preset network model outputs a heart rate value in each training process, and the training accuracy of the preset network model can be obtained by comparing the output heart rate value of each training with the heart rate value acquired through the heart rate band. Stopping training when the training accuracy reaches a preset accuracy, taking the current preset network model as a cyclic neural network model, and adding all parameters corresponding to the current cyclic neural network model as target parameters and an algorithm corresponding to the cyclic neural network model into the wearable device.
According to the scheme, the user information, the motion parameters, the heart rate parameters and the heart rate values collected by the heart rate belt are obtained through the sample library, the user information, the motion parameters and the heart rate parameters serve as input values of the preset network model, the heart rate values collected by the heart rate belt serve as output values of the preset network model, the preset network model is trained according to the preset parameters and the preset algorithm, when the training accuracy reaches the preset accuracy, the training is stopped, and the current preset network model serves as the recurrent neural network model and is added to the wearable device to serve as the heart rate detection algorithm. The trained recurrent neural network model is used as a heart rate detection algorithm in the wearable device to calculate the heart rate value, and the accuracy of calculating the heart rate value is improved.
Referring to fig. 4, fig. 4 is a schematic structural diagram of the preset network model of the present invention after being expanded according to time, and in this embodiment, the preset network model may be a recurrent neural network model. The nodes between the bin shifting layers of the recurrent neural network model are not independent but connected, so that the recurrent neural network modelThe cyclic neural network model can process the time context of input values, from the network structure, the cyclic neural network model can memorize the information of a front-layer network and utilize the information of the front-layer node to influence the output of a rear-layer node, and the input of the hidden-layer node not only comprises the input values of the input layer, but also comprises the output of the hidden layer at the last moment. In the recurrent neural network, each node has an input Xt at each time t, and then the state h of the node at the previous time is determined according to the recurrent neural network modelt-1Calculate the new state htAnd output OtThat is, the current state of the recurrent neural network model is based on the state h at the previous timet-1Obtained by co-action with the input Xt at the current moment, and expressed by the formula:
ht＝tanh(ω*Xt+ω*ht-1+b)
Ot＝g(ω*ht)
wherein ω represents the weight in the neurons, g is an activation function, and it can be seen with reference to fig. 4 that the weight values between the neurons in the recurrent neural network model are all shared, and this feature of the recurrent neural network model can greatly reduce the number of parameters that the network needs to learn, and only different input values are needed in each execution process.
The input layer, the hidden layer and the output layer of the preset network model are all connected or partially connected, all nodes in the layer of the preset network model are connected with one another, the input of the hidden layer comprises an input value of the input layer and an output value of the hidden layer at the last moment, and weight values among neurons in the preset network model are shared.
The step of using the user information, the motion parameter and the heart rate parameter as input values of a preset network model further comprises:
acquiring an output value of current training of a hidden layer in the preset network model;
and taking the output value of the current training of the hidden layer, the user information, the motion parameter and the heart rate parameter as the input values of the preset network model.
According to the scheme, the effect of detecting the heart rate value in real time can be more accurately realized by the cyclic neural network due to the characteristic that the cyclic neural network can memorize the information of the front-layer network.
An embodiment of the present invention further provides a computer-readable storage medium, where a heart rate detection program is stored, and when being executed by a processor, the heart rate detection program implements the heart rate detection method as described above.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the invention without departing from the invention
With clear spirit and scope. Thus, if such 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 such modifications and variations.
Claims (10)
1. A heart rate detection method is applied to wearable equipment and comprises the following steps:
acquiring user information, motion parameters and heart rate parameters of a user wearing the wearable device, wherein the heart rate parameters are detected by a heart rate sensor in the wearable device;
and determining a target heart rate value according to the user information, the motion parameters, the heart rate parameters and a prestored recurrent neural network model, wherein the recurrent neural network model takes the user information, the motion parameters and the heart rate parameters collected by a heart rate sensor as input, and the heart rate value collected by a heart rate belt is obtained as output training.
2. The heart rate detection method of claim 1, wherein the training process of the pre-stored recurrent neural network model comprises:
acquiring user information, motion parameters, heart rate parameters and heart rate values acquired by a heart rate band from an acquisition sample library;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate band as output values of the preset network model so as to train the preset network model;
acquiring the accuracy of a preset network model in the training process;
stopping training when the accuracy reaches a preset accuracy;
and taking the preset network model obtained by current training as the recurrent neural network model.
3. The heart rate detection method according to claim 2, wherein the step of using the user information, the exercise parameter and the heart rate parameter as input values of a preset network model and using the heart rate value collected by the heart rate band as an output value of the preset network model to train the preset network model comprises:
normalizing the user information, the motion parameters and the heart rate parameters in the sample library;
and taking the user information, the motion parameters and the heart rate parameters after the normalization processing as input values of a preset network model, and taking a heart rate value acquired by a heart rate zone as an output value of a recurrent neural network so as to train the preset network model.
4. The heart rate detection method according to claim 2, wherein the step of training the preset network model by using the user information, the exercise parameters and the heart rate parameters as input values of the preset network model and using a heart rate value acquired by a heart rate band as an output value of the recurrent neural network comprises:
acquiring preset parameters and preset algorithms corresponding to a preset network model;
taking the user information, the motion parameters and the heart rate parameters as input values of a preset network model, and taking heart rate values acquired by a heart rate zone as output values of the preset network model;
training the preset network model after the input value and the output value are set according to the preset parameters and a preset algorithm, wherein the preset parameters comprise the number of preset neuron nodes, the number of preset network layers, a preset initial value, a preset learning rate, a preset neuron discarding value and weight and a preset training period of the preset network model, and the preset algorithm is an adaptive moment estimation algorithm.
5. The heart rate detection method according to claim 4, wherein the step of obtaining the preset parameters and the preset algorithm corresponding to the preset network model further comprises:
acquiring a preset regular item corresponding to a preset network model;
and adding the preset regular term to a loss function of the preset network model.
6. The method for detecting heart rate according to claim 2, wherein the input layer, the hidden layer, and the output layer of the preset network model are all connected or partially connected, nodes in the layers of the preset network model are connected to each other, the input of the hidden layer includes an input value of the input layer and an output value of the hidden layer at a previous time, and weight values between neurons in the preset network model are shared.
7. The heart rate detection method of claim 6, wherein the step of using the user information, the exercise parameter, and the heart rate parameter as input values of a preset network model further comprises:
acquiring an output value of current training of a hidden layer in the preset network model;
and taking the output value of the current training of the hidden layer, the user information, the motion parameter and the heart rate parameter as the input values of the preset network model.
8. The heart rate detection method of claim 1, wherein the step of obtaining user information, motion parameters, and heart rate parameters of a user wearing the wearable device is followed by further comprising:
adding user information, motion parameters, and heart rate parameters of a user of the wearable device to the sample library.
9. A heart rate detection device, characterized in that the device comprises a memory, a processor and a heart rate detection program stored in the memory and executable on the processor, the heart rate detection program, when executed by the processor, implementing the heart rate detection method according to any one of claims 1-8.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a heart rate detection program, which when executed by a processor implements the heart rate detection method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110072811.2A CN112818773A (en) | 2021-01-19 | 2021-01-19 | Heart rate detection method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110072811.2A CN112818773A (en) | 2021-01-19 | 2021-01-19 | Heart rate detection method and device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112818773A true CN112818773A (en) | 2021-05-18 |
Family
ID=75858502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110072811.2A Pending CN112818773A (en) | 2021-01-19 | 2021-01-19 | Heart rate detection method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112818773A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115120217A (en) * | 2022-08-25 | 2022-09-30 | 首都医科大学附属北京同仁医院 | Motion artifact removing method and device, storage medium and electronic equipment |
| CN117243584A (en) * | 2023-09-11 | 2023-12-19 | 深圳市熠华智能科技有限公司 | Heart rate monitoring method and device based on bracelet type electrocardiograph equipment and storage medium |
| CN117243584B (en) * | 2023-09-11 | 2024-06-21 | 深圳市熠华智能科技有限公司 | Heart rate monitoring method and device based on bracelet type electrocardiograph equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190038148A1 (en) * | 2013-12-12 | 2019-02-07 | Alivecor, Inc. | Health with a mobile device |
| CN110420019A (en) * | 2019-07-29 | 2019-11-08 | 西安电子科技大学 | A kind of depth recurrence heart rate estimation method of ballistocardiography signal |
| WO2020047750A1 (en) * | 2018-09-04 | 2020-03-12 | 深圳先进技术研究院 | Arrhythmia detection method and apparatus, electronic device, and computer storage medium |
| CN111160139A (en) * | 2019-12-13 | 2020-05-15 | 中国科学院深圳先进技术研究院 | Electrocardiosignal processing method and device and terminal equipment |
| CN111798980A (en) * | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN111887858A (en) * | 2020-08-04 | 2020-11-06 | 西安电子科技大学 | Ballistocardiogram signal heart rate estimation method based on cross-modal mapping |
| CN112057066A (en) * | 2020-09-03 | 2020-12-11 | 青岛歌尔智能传感器有限公司 | Heart rate detection method, wearable device and computer storage medium |
-
2021
- 2021-01-19 CN CN202110072811.2A patent/CN112818773A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190038148A1 (en) * | 2013-12-12 | 2019-02-07 | Alivecor, Inc. | Health with a mobile device |
| WO2020047750A1 (en) * | 2018-09-04 | 2020-03-12 | 深圳先进技术研究院 | Arrhythmia detection method and apparatus, electronic device, and computer storage medium |
| CN110420019A (en) * | 2019-07-29 | 2019-11-08 | 西安电子科技大学 | A kind of depth recurrence heart rate estimation method of ballistocardiography signal |
| CN111160139A (en) * | 2019-12-13 | 2020-05-15 | 中国科学院深圳先进技术研究院 | Electrocardiosignal processing method and device and terminal equipment |
| CN111798980A (en) * | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN111887858A (en) * | 2020-08-04 | 2020-11-06 | 西安电子科技大学 | Ballistocardiogram signal heart rate estimation method based on cross-modal mapping |
| CN112057066A (en) * | 2020-09-03 | 2020-12-11 | 青岛歌尔智能传感器有限公司 | Heart rate detection method, wearable device and computer storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 张德丰: "《TENSORFLOW深度学习 从入门到进阶》", 30 April 2020 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115120217A (en) * | 2022-08-25 | 2022-09-30 | 首都医科大学附属北京同仁医院 | Motion artifact removing method and device, storage medium and electronic equipment |
| CN117243584A (en) * | 2023-09-11 | 2023-12-19 | 深圳市熠华智能科技有限公司 | Heart rate monitoring method and device based on bracelet type electrocardiograph equipment and storage medium |
| CN117243584B (en) * | 2023-09-11 | 2024-06-21 | 深圳市熠华智能科技有限公司 | Heart rate monitoring method and device based on bracelet type electrocardiograph equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gholamiangonabadi et al. | Deep neural networks for human activity recognition with wearable sensors: Leave-one-subject-out cross-validation for model selection | |
| CN108926338B (en) | Heart rate prediction technique and device based on deep learning | |
| CN108703760A (en) | Human motion gesture recognition system and method based on nine axle sensors | |
| CN106407889A (en) | Video human body interaction motion identification method based on optical flow graph depth learning model | |
| AU2020103811A4 (en) | Method for recognizing human knee motion postures based on extreme learning machine | |
| US10610109B2 (en) | Emotion representative image to derive health rating | |
| JP2018147474A (en) | Learning device, learning result utilization device, learning method, and learning program | |
| Duro et al. | Discrete-time backpropagation for training synaptic delay-based artificial neural networks | |
| Cai et al. | GBDT‐Based Fall Detection with Comprehensive Data from Posture Sensor and Human Skeleton Extraction | |
| CN109693234A (en) | Robot tumble prediction technique, device, terminal device and computer storage medium | |
| Ramachandran et al. | Evaluation of feature engineering on wearable sensor-based fall detection | |
| CN112818773A (en) | Heart rate detection method and device and storage medium | |
| KR20210054349A (en) | Method for predicting clinical functional assessment scale using feature values derived by upper limb movement of patients | |
| Urresty Sanchez et al. | Fall detection using accelerometer on the user’s wrist and artificial neural networks | |
| Bezobrazov et al. | Artificial intelligence for sport activitity recognition | |
| Imran et al. | Harresnext: An efficient resnext inspired network for human activity recognition with inertial sensors | |
| CN110598789A (en) | Human fatigue state prediction method and system based on fuzzy perceptron | |
| Bevilacqua et al. | Rehabilitation exercise segmentation for autonomous biofeedback systems with ConvFSM | |
| CN112801283B (en) | Neural network model, action recognition method, device and storage medium | |
| CN112861679B (en) | Transfer learning method and system for behavior recognition | |
| CN111743668B (en) | Prosthesis control method, device, electronic equipment and storage medium | |
| Al-Khamees et al. | Enhancing the stability of the deep neural network using a non-constant learning rate for data stream. | |
| US20230086261A1 (en) | Clustering device, clustering method, and clustering program | |
| Petlenkov et al. | Application of self organizing Kohonen map to detection of surgeon motions during endoscopic surgery | |
| Rivera et al. | A novel method for automatic detection and classification of movement patterns in short duration playing activities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210518 |
|
| RJ01 | Rejection of invention patent application after publication |