CN112307996B - Fingertip electrocardio identity recognition device and method - Google Patents
Fingertip electrocardio identity recognition device and method Download PDFInfo
- Publication number
- CN112307996B CN112307996B CN202011226668.XA CN202011226668A CN112307996B CN 112307996 B CN112307996 B CN 112307996B CN 202011226668 A CN202011226668 A CN 202011226668A CN 112307996 B CN112307996 B CN 112307996B
- Authority
- CN
- China
- Prior art keywords
- layer
- fingertip
- neural network
- convolutional neural
- deep convolutional
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 104
- 238000007781 pre-processing Methods 0.000 claims abstract description 18
- 238000013508 migration Methods 0.000 claims abstract description 13
- 230000005012 migration Effects 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims description 50
- 230000006870 function Effects 0.000 claims description 35
- 238000012549 training Methods 0.000 claims description 21
- 238000013519 translation Methods 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000005516 engineering process Methods 0.000 claims description 12
- 230000006978 adaptation Effects 0.000 claims description 11
- 238000013526 transfer learning Methods 0.000 claims description 11
- 238000010586 diagram Methods 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 9
- 125000004122 cyclic group Chemical group 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000013139 quantization Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 238000011176 pooling Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000008014 freezing Effects 0.000 claims description 3
- 238000007710 freezing Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 18
- 230000003750 conditioning effect Effects 0.000 abstract description 13
- 238000004891 communication Methods 0.000 abstract description 10
- 230000007246 mechanism Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000010355 oscillation Effects 0.000 description 4
- 230000003321 amplification Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000033001 locomotion Effects 0.000 description 3
- 238000003199 nucleic acid amplification method Methods 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 239000000853 adhesive Substances 0.000 description 1
- 230000001070 adhesive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/148—Wavelet transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
Abstract
The invention discloses a fingertip electrocardio identity recognition device and method. The device comprises a fingertip electrocardiosignal acquisition module, a signal conditioning circuit, a controller and a Bluetooth communication module, wherein the output end of the fingertip electrocardiosignal acquisition module is electrically connected with the input end of the signal conditioning circuit, the output end of the signal conditioning circuit is electrically connected with the controller, and the controller is also electrically connected with the Bluetooth communication module. The method comprises the steps of collecting fingertip ECG signals of human fingertips, preprocessing the fingertip ECG signals, inputting the fingertip ECG signals to a fingertip electrocardio identity recognition model, carrying out identity recognition on the fingertip electrocardio identity recognition model, outputting corresponding identity information, and carrying out two-time deep model migration on the fingertip electrocardio identity recognition model based on a deep convolutional neural network GoogLeNet. The invention collects the fingertip ECG signal for identity recognition, which not only ensures the stability of the electrocardio collection process, but also realizes convenient and comfortable recognition under small samples.
Description
Technical Field
The invention relates to the technical field of modern identity recognition, in particular to a fingertip electrocardio identity recognition device and method.
Background
With the progressive and deep development of internet of things (Internet of Things, ioT) and artificial intelligence (artificial intelligence, AI), many lawbreakers apply them to modern information security systems, resulting in increasingly more intelligent counterfeiting and theft technologies such as face-imitation, fake fingerprint, voice imitation, etc., which undoubtedly increase the potential hazards of the information security systems. The development of sensor technology promotes the gradual application of small-volume, easy-to-integrate and low-power-consumption electrocardio acquisition chips without conductive adhesive, provides technical support for developing a wearable or fingertip electrocardio signal (ECG) acquisition system, and provides wide application prospect for an ECG-based identification system. Because the living body acquisition requirement and hidden internal characteristics of the ECG lead the ECG to have high resistance to external attack, the finger-tip electrocardio-based identification technology has wider application prospect. Technical challenges currently exist, however, principally in the following respects:
the first challenge comes from the physical implementation of the fingertip electrocardiographic acquisition system. As a time-varying signal, ECG is extremely unstable during acquisition, causing a significant amount of noise interference.
The second challenge comes from the accurate identification of the fingertip electrocardiographic identification system. On the one hand, algorithm applicability challenges: compared with the ECG acquired by the chest standard lead, the amplitude of the ECG component of the fingertip acquisition signal is weaker, the interference such as motion artifact is more frequent and even takes the dominant role, and most of the existing ECG identification methods mostly adopt the chest acquisition ECG signal with rich data volume as a data set for model training, and the identification system acquired by the chest is transferred to the identification system acquired by the fingertip from the identification system acquired by the chest, so that the practicability and the adaptability still face challenges; on the other hand, the absence of a fingertip data sample: the vigorous development of Deep Learning (DL) makes more and more domestic and foreign experts see the strong advantage of the Deep Learning in feature self-Learning, and provides strong technical support for finger tip electrocardiographic identification, but the strong and sufficient feature self-Learning still depends on big data samples, and how to solve the problem that the Deep neural network is very difficult to rely on the lazy nature of the big computing power of big data in feature self-Learning under the condition of insufficient sample data (the existing network database has less ECG data acquired through wearable and fingertip acquisition equipment).
In view of the above, the identity recognition device and the method based on fingertip electrocardio acquisition are provided, and the high-efficiency and accurate recognition of the identity of the individual is realized while the fingertip electrocardio is stably acquired, so that the device and the method have wide practical application value.
Disclosure of Invention
The invention provides a fingertip electrocardio identity recognition device and method, which are used for acquiring fingertip ECG signals for identity recognition and are convenient and comfortable in acquisition process, and the technical problems that the acquisition process is complicated and inconvenient in acquisition of chest ECG signals in the existing ECG identity recognition system are solved.
In order to solve the problems, the invention is realized by adopting the following technical scheme:
the invention discloses a fingertip electrocardio identity recognition device which comprises a fingertip electrocardio signal acquisition module, a signal conditioning circuit, a controller and a Bluetooth communication module, wherein the output end of the fingertip electrocardio signal acquisition module is electrically connected with the input end of the signal conditioning circuit, the output end of the signal conditioning circuit is electrically connected with the controller, and the controller is also electrically connected with the Bluetooth communication module.
In the scheme, the fingertip electrocardiosignal of the user is rapidly acquired through the fingertip electrocardiosignal acquisition module, is subjected to differential amplification and filtering by the signal conditioning circuit, is sent to the ADC of the controller MCU for analog-to-digital acquisition conversion, is sent to the mobile phone end through the Bluetooth communication module in a UART mode, and the mobile phone end acquires fingertip electrocardiosignal data of a person to be identified. Meanwhile, the mobile phone terminal accesses the cloud platform through the wifi/4G wireless transmission module, uploads fingertip electrocardio data of the current user and downloads an identification result of an identification model based on fingertip electrocardio in the cloud platform. The invention only needs to collect fingertip electrocardiographic data, and is more convenient and comfortable compared with the collection of chest electrocardiographic data.
The invention discloses a fingertip electrocardio identity recognition method, which comprises the following steps of:
collecting fingertip ECG signals of human fingertips, preprocessing the fingertip ECG signals, inputting the fingertip ECG signals to a fingertip electrocardio identity recognition model, carrying out identity recognition on the fingertip electrocardio identity recognition model, and outputting corresponding identity information;
the fingertip electrocardio identity recognition model construction method comprises the following steps:
s1: establishing a deep convolutional neural network GoogLeNet;
s2: constructing a deep convolutional neural network GoogLeNet-I1 capable of identifying the identity of an individual according to chest ECG signals based on the deep convolutional neural network GoogLeNet, preprocessing large sample chest ECG signal data acquired by standard leads in a database (Physionet network reference library) by using a transfer learning finetune technology, and training the deep convolutional neural network GoogLeNet-I1;
s3: based on the trained deep convolutional neural network GoogLeNet-I1, a deep convolutional neural network GoogLeNet-I2 capable of identifying the identity of an individual according to fingertip ECG signals is constructed, a transfer learning finetune technology is used for preprocessing small sample fingertip ECG signal data in a database (Check Your Biosignals Here initiative database), then the deep convolutional neural network GoogLeNet-I2 is trained, and the trained deep convolutional neural network GoogLeNet-I2 is the fingertip electrocardio identity identification model.
Preferably, the method for constructing the deep convolutional neural network GoogLeNet-I1 in the step S2 comprises the following steps:
n1: modifying an acceptance structure of a deep convolutional neural network GoogLeNet to obtain an acceptance-I1 structure, wherein the acceptance-I1 structure comprises an input layer, a processing layer and a connecting layer, the processing layer comprises a first processing branch, a second processing branch, a third processing branch and a fourth processing branch which are connected in parallel between the input layer and the connecting layer, the first processing branch comprises a 1*1 convolution unit, the second processing branch comprises a 1*1 convolution unit and a 3*3 convolution unit which are all connected in sequence, the third processing branch comprises a 1*1 convolution unit, a 2 x 2 convolution unit and a 3*3 convolution unit which are all connected in sequence, and the fourth processing branch comprises a 3*3 maximum pooling unit and a 1*1 convolution unit which are all connected in sequence;
n2: migrating the first 10 layers of the deep convolutional neural network GoogLeNet, and freezing network layer parameters and optimizer states of the first 10 layers to remain unchanged;
and N3: batch Normalization layers are added between the 39 th layer and the 40 th layer, between the 110 th layer and the 111 th layer and between the 139 th layer and the 140 th layer of the deep convolutional neural network GoogLeNet;
n4: dropout layers and Fully connected layers are sequentially added between a 142 th layer and a 143 th layer of the deep convolutional neural network GoogLeNet, so that a new deep convolutional neural network GoogLeNet-I1 is constructed, and a last 6 layer of the new deep convolutional neural network GoogLeNet-I1 is sequentially a 144 th layer Dropout layer, a 145 th layer Fully connected layer, a 146 th layer Dropout layer, a 147 th layer Fully connected layer, a 148 th Softmax layer and a 149 th Output layer from front to back.
The traditional acceptance mechanism consists of 3 groups of convolution kernels and a pooling unit, and receives an input image from a previous layer together, processes the input image in parallel, and splices output results according to channels. The invention improves the acceptance mechanism on the basis of the prior art, and aims at solving the problem that the large convolution (5 multiplied by 5) limits the calculation capacity of the model; in addition, a small convolution of 1×1 is added before the original convolution of 3×3, so that the dimension and depth effects of the image space are reduced.
The deep convolutional neural network GoogLeNet is used as the champion of image Net challenge in 2014, has reasonable network depth, few parameters and high calculation performance, and is widely applied to classification/identification tasks. The deep convolutional neural network GoogLeNet also converts the full-connection layer and the convolutional layer into sparse connection, and uses a global average pool to replace the sparse connection layer and the sparse matrix to cluster into denser submatrices, so that the overfitting phenomenon is avoided, and the calculation performance is improved. The invention uses the deep convolutional neural network GoogleNet as a reference model to reconstruct and train the ECG identification model acquired by the chest.
Since the first layers of each network learn the general characteristics of the current task, the effect of direct migration is better, and the deep convolutional neural network GoogLeNet starts to increase the acceptance mechanism from layer 11 to perform multi-resolution learning. The first 10 layers of the original google net structure is therefore directly migrated and its network layer parameters and optimizer states are frozen and kept unchanged.
In model training, parameters and optimizer states of each layer are updated continuously along with iteration, and if a certain layer deviates, serious deviation of internal covariates is caused, so that continuous change of data characteristic distribution is caused. In order to solve the problem, the invention adds a Batch Normalization layer after the 39 th, 110 th and 139 th layers of the original deep convolutional neural network GoogLeNet so as to avoid the problem of gradient disappearance; dropout-fusion connected (Dropout-Fc) combinations are added after layer 142 of the original deep convolutional neural network google net to reduce the over/under fit problem.
The initial construction of the GoogLeNet-I1 is completed through the steps, then the current network structure is kept unchanged by utilizing the finetune technology of transfer learning, the model is trained by designing new training parameters, the weight parameters of the model and the state of an optimizer are adjusted, the identification accuracy and the equal error rate are used as measurement indexes, and training of the ECG identification model based on chest acquisition is carried out on the constructed GoogLeNet-I1 network.
Preferably, the Dropout layer of the 144 th layer of the control depth convolution neural network GoogLeNet-I1 has a characteristic rejection degree of 50% and the Dropout layer of the 146 th layer has a characteristic rejection degree of 10%.
Preferably, the method for constructing the deep convolutional neural network GoogLeNet-I2 in the step S3 comprises the following steps:
the network structures of two Dropout layers and two Fully connected layers in the last 6 layers of the deep convolutional neural network GoogLeNet-I1 are modified into self-adaptive layer structures, and the method is as follows: the Dropout layers of the 144 th layer and the 146 th layer of the GoogleNet-I1 are modified into two parallel connected Dropout structures and an adaptive loss function, and the Fully connected layers of the 145 th layer and the 147 th layer of the GoogleNet-I1 are modified into two parallel connected Fully connected structures and an adaptive loss function;
the self-adaptive layer structure performs countermeasure learning from a source domain to a target domain, and builds a self-adaptive loss function to measure self-adaptive errors from the source domain to the target domain, so that deep countermeasure migration learning of a new task is realized, the source domain is a chest ECG signal, the target domain is a fingertip ECG signal, and the self-adaptive loss function corresponding to the self-adaptive layer structure is as shown in formula (1):
since the conventional loss function is a conventional classification loss on the source domain, the difference between the predicted result and the true result is measured, and the conventional loss function is as in formula (2):
the loss function of the whole network is calculated based on the formulas (1) and (2) as follows:
where Φ (·) represents mapping the source/target domain to the regenerated kernel hilbert space, Θ represents all weights and bias parameters of the deep convolutional neural network, J (·) represents cross entropy, K represents the total network layer number,representing all annotated data sets in the source domain, the source domain being chest ECG signal, +.>Representing all annotated data sets in a target domain, the target domain being a fingertip ECG signal, n1 representing an adaptation layer of a source domain, n2 representing an adaptation layer of the target domain, D s Data set representing source domain, D t Data set representing target domain, lambda represents the degree of network adaptation (the invention adjusts lambda e 0,1]Performing feature learning);
a new deep convolutional neural network google net-I2 was constructed.
Although the advantage of "finishing" is evident, it does not require the reconstruction of the network structure, it also has certain drawbacks: the situation that the distribution of the training data and the test data is different cannot be handled. Specifically, in the model migration of the invention, the image classification task of the original deep convolutional neural network GoogLeNet and the signal classification of the current deep convolutional neural network GoogLeNet-I1 do not follow the same data distribution, so the invention provides the concept of a self-adaptive layer of the deep convolutional neural network. The invention aims to realize the self-adaption from a source domain (based on chest ECG signals acquired by the chest) to a target domain (based on fingertip ECG signals acquired by the fingertip) by adding a network self-adaption layer, so that the data distribution of the source domain and the target domain is more similar, and better learning of key features is realized. Since the first layers of the network learn the general characteristics of the current task and the second layers are new training using training data of the new task, the migration and adaptation tasks of the second layers of the model should be mainly considered. From the theoretical analysis of deep learning, the previous layer of classifier, namely the feature, adding self-adaption to the feature is also the work to be done by transfer learning.
Preferably, the method for preprocessing the chest ECG signal is the same as the method for preprocessing the fingertip ECG signal, and the preprocessing method is as follows:
m1: processing the original ECG signal by adopting a cyclic translation denoising algorithm based on a wavelet hard threshold value to obtain a clean ECG signal;
m2: the fixed window length is 3f, the ECG signal with the fixed window length is randomly intercepted to be used as a short period ECG signal, and f is the sampling frequency of the ECG signal;
m3: and carrying out generalized S transformation on the short-period ECG signal, constructing a corresponding frequency domain instantaneous locus structure characteristic diagram group, and taking the frequency domain instantaneous locus structure characteristic diagram group as the input of the deep convolutional neural network.
As a non-stationary time-varying weak signal, the electrocardiosignals are extremely easily interfered by various aspects such as instruments, human body movement and the like in the acquisition process, so that the components of the electrocardiosignals are weak and are mixed with noise frequency bands, the subsequent characteristic learning, capturing and characterization of the time sequence signals are seriously influenced, and the invention acquires a clean ECG signal by adopting a cyclic translation denoising algorithm based on a wavelet hard threshold value.
On the basis of the denoising, a sliding window with the length of 3f is adopted to process the denoised ECG signal by random sliding window, and a short-period clean electrocardiosignal with the window length is intercepted. And analyzing the time-frequency joint characteristics of the ECG time sequence signals, and obtaining a sufficient frequency domain instantaneous track structure characteristic diagram which is used as the input of a deep convolutional neural network model to provide sufficient two-dimensional image characteristics for identity recognition.
Preferably, the step M1 includes the steps of:
m11: performing 8-cycle translation processing on the original ECG signal, and changing the position of a singular point in the original ECG signal;
m12: decomposing the original ECG signal by discrete wavelet transformation;
m13: performing threshold quantization processing on the wavelet coefficient in a wavelet domain through a hard threshold function, and performing inverse discrete wavelet transformation according to the estimated wavelet coefficient obtained after the quantization processing to obtain a reconstructed ECG signal;
m14: and 8 times of inverse cyclic translation processing is carried out on the reconstructed ECG signal, so that a clean ECG signal is obtained.
The translation changes the position of the signal singular point in the whole signal section, the inverse translation operation keeps the translation invariance of the signal, the whole operation reduces the process of waveform oscillation, weakens or even eliminates Gibbs oscillation, and better approximates to the original clean signal.
Preferably, the step M3 includes the steps of:
m31: the short period ECG signal is generalized S-transformed as follows:
the window function of the generalized S transform is:
wherein z (t) is a short period ECG electrocardiosignal, t is time, tau is a time shift factor, f is ECG signal sampling frequency, h is window width parameter of window function, and w is amplitude parameter of window function;
through the above processing, a complex matrix is obtained
The complex matrix comprises a real part and an imaginary part,
wherein n=1.5f, m=3f;
m32: and drawing a frequency domain instantaneous locus structural feature map corresponding to each column of data of the complex matrix, and obtaining M pieces of frequency domain instantaneous locus structural feature maps altogether, wherein M Zhang Pinyu instantaneous locus structural feature maps form a frequency domain instantaneous locus structural feature map group corresponding to short-period ECG signals and serve as input of a deep convolutional neural network.
Each column of the complex matrix reflects the "instantaneous frequency characteristic" of the current point in time.
Preferably, in the step M32, the method for drawing the frequency domain instantaneous locus structure feature map corresponding to a certain column data is as follows: and taking the real part alpha as an X axis and the imaginary part beta as a Y axis, establishing a rectangular coordinate system, marking the position of the row of data on the rectangular coordinate system, drawing a track, and reducing the image pixels to 224 multiplied by 224.
The frequency domain instantaneous locus structure feature map reflects the frequency feature information of the time sequence signal at each time point and the change trend of the time points before and after the time point, and is used as the input of the deep convolutional neural network.
Preferably, the method comprises
The setting of w and h directly affects the shape of g (t), thereby affecting the capture and recognition accuracy of key features. The invention uses the recognition precision Acc as an objective function to adjust the parameters w and h to obtain the optimal parameter group table
The beneficial effects of the invention are as follows: (1) An electrocardio acquisition system architecture for acquiring fingertip electrocardiosignals of a person to be identified is constructed based on the internet of things technology, so that stability and signal quality of an electrocardio acquisition process are guaranteed, and convenience and comfort of the acquisition process are also considered. (2) The transfer learning holds a deep convolutional neural network to construct and train a fingertip electrocardio identity recognition model, the deep convolutional neural network GoogLeNet is used for carrying out model transfer twice, model improvement and optimization are carried out from multiple layers such as a network structure, an optimization algorithm, a loss function and the like, the electrocardio identity recognition model based on chest collection is obtained through primary transfer, namely the deep convolutional neural network GoogLeNet-I1, the electrocardio identity recognition model based on fingertip collection is determined through secondary transfer, namely the deep convolutional neural network GoogLeNet-I2, manual feature extraction is not needed, and recognition under a small sample is realized.
Drawings
FIG. 1 is a schematic structural view of an embodiment;
FIG. 2 is a flow chart of an embodiment;
FIG. 3 is a flow chart of the construction of a finger tip electrocardiographic identification model;
FIG. 4 is a schematic diagram of the structure of the acceptance-I1;
FIG. 5 is a schematic diagram of the structure of a deep convolutional neural network GoogLeNet-I2.
In the figure: 1. the system comprises a fingertip electrocardiosignal acquisition module, a signal conditioning circuit, a controller, a Bluetooth communication module and a fingertip electrocardiosignal acquisition module.
Detailed Description
The technical scheme of the invention is further specifically described below through examples and with reference to the accompanying drawings.
Examples: the fingertip electrocardio identity recognition device in this embodiment, as shown in fig. 1, comprises a fingertip electrocardio signal acquisition module 1, a signal conditioning circuit 2, a controller 3 and a bluetooth communication module 4, wherein the output end of the fingertip electrocardio signal acquisition module 1 is electrically connected with the input end of the signal conditioning circuit 2, the output end of the signal conditioning circuit 2 is electrically connected with the controller 3, and the controller 3 is also electrically connected with the bluetooth communication module 4.
In the scheme, the fingertip electrocardiosignal of the user is rapidly acquired through the fingertip electrocardiosignal acquisition module, is subjected to differential amplification and filtering by the signal conditioning circuit, is sent to the ADC of the controller MCU for analog-to-digital acquisition conversion, is sent to the mobile phone end through the Bluetooth communication module in a UART mode, and the mobile phone end acquires fingertip electrocardiosignal data of a person to be identified. Meanwhile, the mobile phone terminal accesses the cloud platform through the wifi/4G wireless transmission module, uploads fingertip electrocardio data of the current user and downloads an identification result of an identification model based on fingertip electrocardio in the cloud platform. The invention only needs to collect fingertip electrocardiographic data, and is more convenient and comfortable compared with the collection of chest electrocardiographic data.
The PS25253 sensor in the EPIC sensor of the plessey company is used as a fingertip electrocardiosignal acquisition module to acquire fingertip ECG signals, and the fingertip ECG signals are converted into electric signals. PS25253 has ultra-high impedance solid state, can be used as a dry contact type ECG sensor, is convenient and quick to measure, has less noise interference and is stable to collect. In addition, two metal dry electrodes are used as measuring electrodes, and gel or matrix is not needed to enhance signal transmission. The whole process ensures the stability of the electrocardio acquisition process and the accuracy of signal acquisition, and also gives consideration to the comfort of the acquisition process.
Because the electric signals obtained by the direct difference of the sensors are very weak, and the acquisition process is interfered by various aspects of individuals, instruments and the like, a large amount of noise signals are mixed in the acquired original signals, and therefore differential amplification and shaping filtering processing are adopted on the signals acquired by the fingertip electrocardiosignal acquisition module in the last step, and the noise interference is primarily eliminated.
And sending the filtered electric signals into an ADC (analog-to-digital conversion) of the controller MCU to perform digital-to-analog conversion data, thereby completing the whole signal conditioning process. The controller MCU not only controls the normal operation of the peripheral circuit, but also performs A/D conversion on the denoised signal, and simultaneously, the design of a power supply module of the system, the design processing of related circuits such as a Bluetooth communication module and the like are finished, and the signals are transmitted to the mobile phone end in real time.
The fingertip electrocardiographic identity recognition method of the embodiment, as shown in fig. 2, comprises the following steps:
collecting fingertip ECG signals of human fingertips, preprocessing the fingertip ECG signals, inputting the fingertip ECG signals to a fingertip electrocardio identity recognition model, carrying out identity recognition on the fingertip electrocardio identity recognition model, and outputting corresponding identity information;
as shown in fig. 3, the method for constructing the fingertip electrocardiographic identification model is as follows:
s1: establishing a deep convolutional neural network GoogLeNet;
s2: constructing a deep convolutional neural network GoogLeNet-I1 capable of identifying the identity of an individual according to chest ECG signals based on the deep convolutional neural network GoogLeNet, preprocessing large sample chest ECG signal data acquired by standard leads in a Physionet network reference library by using a transfer learning finetune technology, training the deep convolutional neural network GoogLeNet-I1 (namely, keeping the structure of the deep convolutional neural network GoogLeNet-I1 unchanged, training the deep convolutional neural network GoogLeNet-I1 by designing new training parameters, thereby adjusting the weight parameters of the deep convolutional neural network GoogLeNet-I1 and the state of an optimizer, taking the identification accuracy and the error rate as measurement indexes, and training the constructed deep convolutional neural network GoogLeNet-I1);
s3: based on the trained deep convolutional neural network GoogLeNet-I1, a deep convolutional neural network GoogLeNet-I2 capable of identifying the identity of an individual according to fingertip ECG signals is constructed, a transfer learning finetune technology is used, 63 small sample fingertip ECG signal data in a Check Your Biosignals Here initiative database are preprocessed, then the deep convolutional neural network GoogLeNet-I2 is trained (namely, the structure of the deep convolutional neural network GoogLeNet-I2 is kept unchanged, training parameters are set to train the deep convolutional neural network GoogLeNet-I2, and therefore the weight parameters of the deep convolutional neural network GoogLeNet-I2 and the state of an optimizer are adjusted, the identification accuracy and the error rate are combined to serve as measurement indexes, and the constructed deep convolutional neural network GoogLeNet-I2 is trained, namely, the fingertip electrocardio identity identification model.
The method for constructing the deep convolutional neural network GoogLeNet-I1 in the step S2 comprises the following steps:
n1: modifying an acceptance structure of a deep convolutional neural network GoogLeNet to obtain an acceptance-I1 structure, wherein the acceptance-I1 structure comprises an input layer, a processing layer and a connecting layer, the processing layer comprises a first processing branch, a second processing branch, a third processing branch and a fourth processing branch which are connected between the input layer and the connecting layer in parallel, the first processing branch comprises a 1*1 convolution unit, the second processing branch comprises a 1*1 convolution unit and a 3*3 convolution unit which are all connected in sequence, the third processing branch comprises a 1*1 convolution unit, a 2 x 2 convolution unit and a 3*3 convolution unit which are all connected in sequence, and the fourth processing branch comprises a 3*3 maximum pooling unit and a 1*1 convolution unit which are all connected in sequence;
n2: migrating the first 10 layers of the deep convolutional neural network GoogLeNet, and freezing network layer parameters and optimizer states of the first 10 layers to remain unchanged;
and N3: batch Normalization layers are added between the 39 th layer and the 40 th layer, between the 110 th layer and the 111 th layer and between the 139 th layer and the 140 th layer of the deep convolutional neural network GoogLeNet;
n4: adding a Dropout layer and a Fully connected layer between a 142 th layer and a 143 th layer of the deep convolutional neural network GoogLeNet in sequence to construct a new deep convolutional neural network GoogLeNet-I1, wherein the last 6 layers of the new deep convolutional neural network GoogLeNet-I1 are a 144 th layer of Dropout layer, a 145 th layer of Fully connected layer, a 146 th layer of Dropout layer, a 147 th layer of Fully connected layer, a 148 th Softmax layer and a 149 th layer of Output layer in sequence from front to back;
the Dropout layer of the 144 th layer of the control deep convolutional neural network GoogLeNet-I1 has a characteristic rejection degree of 50% and the Dropout layer of the 146 th layer has a characteristic rejection degree of 10%.
The traditional acceptance mechanism consists of 3 groups of convolution kernels and a pooling unit, and receives an input image from a previous layer together, processes the input image in parallel, and splices output results according to channels. The invention improves the acceptance mechanism on the basis of the prior art, and aims at solving the problem that the large convolution (5 multiplied by 5) limits the calculation capacity of the model; in addition, a small convolution of 1×1 is added before the original convolution of 3×3, so that the dimension and depth effects of the image space are reduced.
The deep convolutional neural network GoogLeNet is used as the champion of image Net challenge in 2014, has reasonable network depth, few parameters and high calculation performance, and is widely applied to classification/identification tasks. The deep convolutional neural network GoogLeNet also converts the full-connection layer and the convolutional layer into sparse connection, and uses a global average pool to replace the sparse connection layer and the sparse matrix to cluster into denser submatrices, so that the overfitting phenomenon is avoided, and the calculation performance is improved. The invention uses the deep convolutional neural network GoogleNet as a reference model to reconstruct and train the ECG identification model acquired by the chest.
Since the first layers of each network learn the general characteristics of the current task, the effect of direct migration is better, and the deep convolutional neural network GoogLeNet starts to increase the acceptance mechanism from layer 11 to perform multi-resolution learning. The first 10 layers of the original google net structure is therefore directly migrated and its network layer parameters and optimizer states are frozen and kept unchanged.
In model training, parameters and optimizer states of each layer are updated continuously along with iteration, and if a certain layer deviates, serious deviation of internal covariates is caused, so that continuous change of data characteristic distribution is caused. In order to solve the problem, the invention adds a Batch Normalization layer after the 39 th, 110 th and 139 th layers of the original deep convolutional neural network GoogLeNet so as to avoid the problem of gradient disappearance; dropout-fusion connected (Dropout-Fc) combinations are added after layer 142 of the original deep convolutional neural network google net to reduce the over/under fit problem.
The initial construction of the GoogLeNet-I1 is completed through the steps, then the current network structure is kept unchanged by utilizing the finetune technology of transfer learning, the model is trained by designing new training parameters, the weight parameters of the model and the state of an optimizer are adjusted, the identification accuracy and the equal error rate are used as measurement indexes, and training of the ECG identification model based on chest acquisition is carried out on the constructed GoogLeNet-I1 network.
The method for constructing the deep convolutional neural network GoogLeNet-I2 in the step S3 comprises the following steps:
although the advantage of "finishing" is evident, it does not require the reconstruction of the network structure, it also has certain drawbacks: the situation that the distribution of the training data and the test data is different cannot be handled. Specifically, in the model migration of the invention, the image classification task of the original deep convolutional neural network GoogLeNet and the signal classification of the current deep convolutional neural network GoogLeNet-I1 do not follow the same data distribution, so the invention provides the concept of a self-adaptive layer of the deep convolutional neural network. The invention aims to realize the self-adaption from a source domain (based on chest ECG signals acquired by the chest) to a target domain (based on fingertip ECG signals acquired by the fingertip) by adding a network self-adaption layer, so that the data distribution of the source domain and the target domain is more similar, and better learning of key features is realized. Since the first layers of the network learn the general characteristics of the current task and the second layers are new training using training data of the new task, the migration and adaptation tasks of the second layers of the model should be mainly considered. From the theoretical analysis of deep learning, the previous layer of classifier, namely the feature, adding self-adaption to the feature is also the work to be done by transfer learning.
The network structures of two Dropout layers and two Fully connected layers in the last 6 layers of the deep convolutional neural network GoogLeNet-I1 are modified into self-adaptive layer structures, and the method is as follows: the Dropout layers of the 144 th layer and the 146 th layer of the GoogleNet-I1 are modified into two parallel connected Dropout structures and an adaptive loss function, and the Fully connected layers of the 145 th layer and the 147 th layer of the GoogleNet-I1 are modified into two parallel connected Fully connected structures and an adaptive loss function;
the self-adaptive layer structure performs countermeasure learning from a source domain to a target domain, and builds a self-adaptive loss function to measure self-adaptive errors from the source domain to the target domain, so that deep countermeasure migration learning of a new task is realized, the source domain is a chest ECG signal, the target domain is a fingertip ECG signal, and the self-adaptive loss function corresponding to the self-adaptive layer structure is as shown in formula (1):
since the conventional loss function is a conventional classification loss on the source domain, the difference between the predicted result and the true result is measured, and the conventional loss function is as in formula (2):
the loss function of the whole network is calculated based on the formulas (1) and (2) as follows:
where Φ (·) represents mapping of source/target domains to a Regenerated Kernel Hilbert Space (RKHS), Θ represents all weights and bias parameters of the deep convolutional neural network (which are targets for learning), J (·) represents cross entropy, K represents total network layer number,representing all annotated data sets in the source domain, the source domain being chest ECG signal, +.>Representing all annotated data sets in a target domain, the target domain being a fingertip ECG signal, n1 representing an adaptation layer of a source domain, n2 representing an adaptation layer of the target domain, D s Data set representing source domain, D t Data set representing target domain, lambda represents the degree of network adaptation (the invention adjusts lambda e 0,1]Performing feature learning);
a new deep convolutional neural network google net-I2 was constructed as shown in fig. 5.
The method for preprocessing the chest ECG signal is the same as the method for preprocessing the fingertip ECG signal, and the preprocessing method is as follows:
m1: processing the original ECG signal by adopting a cyclic translation denoising algorithm based on a wavelet hard threshold value to obtain a clean ECG signal;
m2: the fixed window length is 3f, the ECG signal with the fixed window length is randomly intercepted to be used as a short period ECG signal, and f is the sampling frequency of the ECG signal;
m3: and carrying out generalized S transformation on the short-period ECG signal, constructing a corresponding frequency domain instantaneous locus structure characteristic diagram group, and taking the frequency domain instantaneous locus structure characteristic diagram group as the input of the deep convolutional neural network.
As a non-stationary time-varying weak signal, the electrocardiosignals are extremely easily interfered by various aspects such as instruments, human body movement and the like in the acquisition process, so that the components of the electrocardiosignals are weak and are mixed with noise frequency bands, the subsequent characteristic learning, capturing and characterization of the time sequence signals are seriously influenced, and the invention acquires a clean ECG signal by adopting a cyclic translation denoising algorithm based on a wavelet hard threshold value.
On the basis of the denoising, a sliding window with the length of 3f is adopted to process the denoised ECG signal by random sliding window, and a short-period clean electrocardiosignal with the window length is intercepted. And analyzing the time-frequency joint characteristics of the ECG time sequence signals, and obtaining a sufficient frequency domain instantaneous track structure characteristic diagram which is used as the input of a deep convolutional neural network model to provide sufficient two-dimensional image characteristics for identity recognition.
Step M1 comprises the steps of:
m11: performing 8-cycle translation processing on the original ECG signal, and changing the position of a singular point in the original ECG signal;
m12: decomposing the original ECG signal by discrete wavelet transformation;
m13: performing threshold quantization processing on the wavelet coefficient in a wavelet domain through a hard threshold function, and performing inverse discrete wavelet transformation according to the estimated wavelet coefficient obtained after the quantization processing to obtain a reconstructed ECG signal;
m14: and 8 times of inverse cyclic translation processing is carried out on the reconstructed ECG signal, so that a clean ECG signal is obtained.
The translation changes the position of the signal singular point in the whole signal section, the inverse translation operation keeps the translation invariance of the signal, the whole operation reduces the process of waveform oscillation, weakens or even eliminates Gibbs oscillation, and better approximates to the original clean signal.
Step M3 comprises the steps of:
m31: the short period ECG signal is generalized S-transformed as follows:
the window function of the generalized S transform is:
wherein z (t) is a short period ECG electrocardiosignal, t is time, tau is a time shift factor, f is ECG signal sampling frequency, h is window width parameter of window function, and w is amplitude parameter of window function;
through the above processing, a complex matrix is obtained
The complex matrix comprises a real part and an imaginary part,
wherein n=1.5f, m=3f;
m32: and drawing a frequency domain instantaneous locus structural feature map corresponding to each column of data of the complex matrix, and obtaining M pieces of frequency domain instantaneous locus structural feature maps altogether, wherein M Zhang Pinyu instantaneous locus structural feature maps form a frequency domain instantaneous locus structural feature map group corresponding to short-period ECG signals and serve as input of a deep convolutional neural network.
Each column of the complex matrix reflects the "instantaneous frequency characteristic" of the current point in time.
In the step M32, the method for drawing the frequency domain instantaneous locus structure feature map corresponding to a certain column data is as follows: and taking the real part alpha as an X axis and the imaginary part beta as a Y axis, establishing a rectangular coordinate system, marking the position of the row of data on the rectangular coordinate system, drawing a track, and reducing the image pixels to 224 multiplied by 224.
The frequency domain instantaneous locus structure feature map reflects the frequency feature information of the time sequence signal at each time point and the change trend of the time points before and after the time point, and is used as the input of the deep convolutional neural network.
The setting of w and h directly affects the shape of g (t), thereby affecting the capture and recognition accuracy of key features. The invention uses the recognition precision Acc as an objective function, adjusts the parameters w and h to obtain the optimal parameter combination +.>
The embodiment designs an acceptance-I1 mechanism, and performs multi-scale and multi-resolution fusion on the image; based on a deep convolutional neural network GoogLeNet, combining with migration learning finetune, adding Dropout and other over-fitting/under-fitting measures to realize one-time migration of the model, and creating an ECG identity recognition model trained by acquiring ECG data through a chest of a large sample, wherein the ECG identity recognition model is marked as a deep convolutional neural network GoogLeNet-I1; and then migrating the deep convolutional neural network GoogLeNet-I1 to fingertip electrocardio, constructing a deep network self-adaptive layer, designing a self-adaptive measurement mode, realizing secondary migration of a model, creating the deep convolutional neural network GoogLeNet-12, realizing self-adaptation from a source domain (ECG acquired by a chest) to a target domain (ECG acquired by a fingertip), resolving the requirement of a deep learning algorithm on strong computing capacity of big data under small sample data, and storing the trained deep convolutional neural network GoogLeNet-I2 for subsequent identification requirement based on fingertip electrocardio so as to quickly output an individual identification result.
Claims (7)
1. The fingertip electrocardio identity recognition method is characterized by comprising the following steps of:
collecting fingertip ECG signals of human fingertips, preprocessing the fingertip ECG signals, inputting the fingertip ECG signals to a fingertip electrocardio identity recognition model, carrying out identity recognition on the fingertip electrocardio identity recognition model, and outputting corresponding identity information;
the fingertip electrocardio identity recognition model construction method comprises the following steps:
s1: establishing a deep convolutional neural network GoogLeNet;
s2: constructing a deep convolutional neural network GoogLeNet-I1 capable of identifying the identity of an individual according to chest ECG signals based on the deep convolutional neural network GoogLeNet, preprocessing large sample chest ECG signal data acquired by standard leads in a database by using a transfer learning finetune technology, and training the deep convolutional neural network GoogLeNet-I1;
s3: constructing a deep convolutional neural network GoogLeNet-I2 capable of identifying the identity of an individual according to fingertip ECG signals based on the trained deep convolutional neural network GoogLeNet-I1, preprocessing small sample fingertip ECG signal data in a database by using a transfer learning finetune technology, and training the deep convolutional neural network GoogLeNet-I2, wherein the trained deep convolutional neural network GoogLeNet-I2 is a fingertip electrocardio identity identification model;
the method for constructing the deep convolutional neural network GoogLeNet-I1 in the step S2 comprises the following steps:
n1: modifying an acceptance structure of a deep convolutional neural network GoogLeNet to obtain an acceptance-I1 structure, wherein the acceptance-I1 structure comprises an input layer, a processing layer and a connecting layer, the processing layer comprises a first processing branch, a second processing branch, a third processing branch and a fourth processing branch which are connected in parallel between the input layer and the connecting layer, the first processing branch comprises a 1*1 convolution unit, the second processing branch comprises a 1*1 convolution unit and a 3*3 convolution unit which are all connected in sequence, the third processing branch comprises a 1*1 convolution unit, a 2 x 2 convolution unit and a 3*3 convolution unit which are all connected in sequence, and the fourth processing branch comprises a 3*3 maximum pooling unit and a 1*1 convolution unit which are all connected in sequence;
n2: migrating the first 10 layers of the deep convolutional neural network GoogLeNet, and freezing network layer parameters and optimizer states of the first 10 layers to remain unchanged;
and N3: batch Normalization layers are added between the 39 th layer and the 40 th layer, between the 11O layer and the 111 th layer and between the 139 th layer and the 140 th layer of the deep convolutional neural network GoogLeNet;
n4: adding a Dropout layer and a Fully connected layer between a 142 th layer and a 143 th layer of the deep convolutional neural network GoogLeNet in sequence to construct a new deep convolutional neural network GoogLeNet-I1, wherein the last 6 layers of the new deep convolutional neural network GoogLeNet-I1 are a 144 th layer of Dropout layer, a 145 th layer of Fully connected layer, a 146 th layer of Dropout layer, a 147 th layer of Fully connected layer, a 148 th Softmax layer and a 149 th layer of Output layer in sequence from front to back;
the method for constructing the deep convolutional neural network GoogLeNet-I2 in the step S3 comprises the following steps:
the network structures of two Dropout layers and two Fully connected layers in the last 6 layers of the deep convolutional neural network GoogLeNet-I1 are modified into self-adaptive layer structures, and the method is as follows: the Dropout layers of the 144 th layer and the 146 th layer of the GoogleNet-I1 are modified into two parallel connected Dropout structures and an adaptive loss function, and the Fully connected layers of the 145 th layer and the 147 th layer of the GoogleNet-I1 are modified into two parallel connected Fully connected structures and an adaptive loss function;
the self-adaptive layer structure performs countermeasure learning from a source domain to a target domain, and builds a self-adaptive loss function to measure self-adaptive errors from the source domain to the target domain, so that deep countermeasure migration learning of a new task is realized, the source domain is a chest ECG signal, the target domain is a fingertip ECG signal, and the self-adaptive loss function corresponding to the self-adaptive layer structure is as shown in formula (1):
since the conventional loss function is as in equation (2):
the loss function of the whole network is calculated based on the formulas (1) and (2) as follows:
where Φ (·) represents mapping the source/target domain to the regenerated kernel hilbert space, Θ represents all weights and bias parameters of the deep convolutional neural network, J (·) represents cross entropy, K represents the total network layer number,representing all annotated data sets in a source domain, the source domain being the chestECG signal, & gt>Representing all annotated data sets in a target domain, the target domain being a fingertip ECG signal, n1 representing an adaptation layer of a source domain, n2 representing an adaptation layer of the target domain, D s Data set representing source domain, D t A data set representing a target domain, lambda representing the degree of network adaptation;
a new deep convolutional neural network google net-I2 was constructed.
2. The fingertip electrocardiographic identity recognition method according to claim 1 is characterized in that the characteristic rejection degree of a Dropout layer of a 144 th layer of the control depth convolution neural network GoogLeNet-I1 is 50%, and the characteristic rejection degree of a Dropout layer of a 146 th layer is 10%.
3. The fingertip electrocardiographic identification method according to claim 1 or 2, wherein the method of preprocessing the chest ECG signal is the same as the method of preprocessing the fingertip ECG signal, and the preprocessing method is as follows:
m1: processing the original ECG signal by adopting a cyclic translation denoising algorithm based on a wavelet hard threshold value to obtain a clean ECG signal;
m2: the fixed window length is 3f, the ECG signal with the fixed window length is randomly intercepted to be used as a short period ECG signal, and f is the sampling frequency of the ECG signal;
m3: and carrying out generalized S transformation on the short-period ECG signal, constructing a corresponding frequency domain instantaneous locus structure characteristic diagram group, and taking the frequency domain instantaneous locus structure characteristic diagram group as the input of the deep convolutional neural network.
4. A method for identifying a finger tip electrocardiographic identity according to claim 3, wherein said step M1 comprises the steps of:
m11: performing 8-cycle translation processing on the original ECG signal, and changing the position of a singular point in the original ECG signal;
m12: decomposing the original ECG signal by discrete wavelet transformation;
m13: performing threshold quantization processing on the wavelet coefficient in a wavelet domain through a hard threshold function, and performing inverse discrete wavelet transformation according to the estimated wavelet coefficient obtained after the quantization processing to obtain a reconstructed ECG signal;
m14: and 8 times of inverse cyclic translation processing is carried out on the reconstructed ECG signal, so that a clean ECG signal is obtained.
5. A method for identifying a finger tip electrocardiographic identity according to claim 3, wherein said step M3 comprises the steps of:
m31: the short period ECG signal is generalized S-transformed as follows:
the window function of the generalized S transform is:
wherein z (t) is a short period ECG electrocardiosignal, t is time, tau is a time shift factor, f is ECG signal sampling frequency, h is window width parameter of window function, and w is amplitude parameter of window function;
through the above processing, a complex matrix is obtained
The complex matrix comprises a real part and an imaginary part,
wherein n=1.5f, m=3f;
m32: and drawing a frequency domain instantaneous locus structural feature map corresponding to each column of data of the complex matrix, and obtaining M pieces of frequency domain instantaneous locus structural feature maps altogether, wherein M Zhang Pinyu instantaneous locus structural feature maps form a frequency domain instantaneous locus structural feature map group corresponding to short-period ECG signals and serve as input of a deep convolutional neural network.
6. The method for identifying the electrocardiographic identity of the fingertip according to claim 5, wherein the method for drawing the frequency domain instantaneous locus structure feature map corresponding to a certain column data in the step M32 is as follows: and taking the real part alpha as an X axis and the imaginary part beta as a Y axis, establishing a rectangular coordinate system, marking the position of the row of data on the rectangular coordinate system, drawing a track, and reducing the image pixels to 224 multiplied by 224.
7. The method for identifying the electrocardiographic identity of a fingertip according to claim 5, wherein the steps of
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011226668.XA CN112307996B (en) | 2020-11-05 | 2020-11-05 | Fingertip electrocardio identity recognition device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011226668.XA CN112307996B (en) | 2020-11-05 | 2020-11-05 | Fingertip electrocardio identity recognition device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112307996A CN112307996A (en) | 2021-02-02 |
| CN112307996B true CN112307996B (en) | 2024-04-02 |
Family
ID=74326168
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011226668.XA Active CN112307996B (en) | 2020-11-05 | 2020-11-05 | Fingertip electrocardio identity recognition device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112307996B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113486917B (en) * | 2021-05-17 | 2023-06-02 | 西安电子科技大学 | Radar HRRP small sample target recognition method based on metric learning |
| CN113468988B (en) * | 2021-06-18 | 2024-04-05 | 南京润楠医疗电子研究院有限公司 | Identification method under multiple pressure states based on ECG (electronic pulse-taking) signals |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458197A (en) * | 2019-07-11 | 2019-11-15 | 启东市知微电子科技有限公司 | Personal identification method and its system based on photoplethysmographic |
| CN111709267A (en) * | 2020-03-27 | 2020-09-25 | 吉林大学 | Electroencephalogram signal emotion recognition method of deep convolutional neural network |
| CN111783558A (en) * | 2020-06-11 | 2020-10-16 | 上海交通大学 | Satellite navigation interference signal type intelligent identification method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3686772A1 (en) * | 2019-01-25 | 2020-07-29 | Tata Consultancy Services Limited | On-device classification of fingertip motion patterns into gestures in real-time |
-
2020
- 2020-11-05 CN CN202011226668.XA patent/CN112307996B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458197A (en) * | 2019-07-11 | 2019-11-15 | 启东市知微电子科技有限公司 | Personal identification method and its system based on photoplethysmographic |
| CN111709267A (en) * | 2020-03-27 | 2020-09-25 | 吉林大学 | Electroencephalogram signal emotion recognition method of deep convolutional neural network |
| CN111783558A (en) * | 2020-06-11 | 2020-10-16 | 上海交通大学 | Satellite navigation interference signal type intelligent identification method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112307996A (en) | 2021-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112307996B (en) | Fingertip electrocardio identity recognition device and method | |
| Shen et al. | Gesture recognition through sEMG with wearable device based on deep learning | |
| CN110141216B (en) | Identification method, training method and system for QRS (QRS) characteristic waves of electrocardiosignals | |
| CN102930284A (en) | Surface electromyogram signal pattern recognition method based on empirical mode decomposition and fractal | |
| CN111772669B (en) | Elbow joint contraction muscle force estimation method based on adaptive long-time and short-time memory network | |
| CN113673434B (en) | Electroencephalogram emotion recognition method based on efficient convolutional neural network and contrast learning | |
| CN113111831A (en) | Gesture recognition technology based on multi-mode information fusion | |
| CN106599903B (en) | Signal reconstruction method for weighted least square dictionary learning based on correlation | |
| CN105796091B (en) | A kind of intelligent terminal for removing electrocardiosignal vehicle movement noise | |
| CN112401905B (en) | Natural action electroencephalogram recognition method based on source localization and brain network | |
| CN113128384A (en) | Brain-computer interface software key technical method of stroke rehabilitation system based on deep learning | |
| CN110751060B (en) | Portable motion mode real-time identification system based on multi-source signals | |
| CN116910464A (en) | Myoelectric signal prosthetic hand control system and method | |
| Fu et al. | Finger gesture recognition using sensing and classification of surface electromyography signals with high-precision wireless surface electromyography sensors | |
| CN115329800B (en) | Convolutional neural network activity recognition method based on decoupling training | |
| Wang et al. | Research on sEMG–based gesture recognition using the Attention-based LSTM-CNN with Stationary Wavelet Packet Transform | |
| CN114569143A (en) | Myoelectric gesture recognition method based on attention mechanism and multi-feature fusion | |
| Chen et al. | Recognition of american sign language gestures based on electromyogram (emg) signal with xgboost machine learning | |
| CN114343679A (en) | Surface electromyogram signal upper limb action recognition method and system based on transfer learning | |
| CN113712516A (en) | Method for establishing traditional Chinese medicine pulse condition recognition model based on time sequence convolution network | |
| Qin et al. | Arm movements recognition by implementing cnn on microcontrollers | |
| Zhao et al. | Extreme Gradient Boosting for Surface Electromyography Classification on Time-Domain Features | |
| Chen et al. | MyoBit: A Public Dataset Based on An Armband with 16 sEMG Channels for Gesture Recognition under Non-ideal Conditions | |
| CN117150292A (en) | Incremental learning-based gesture recognition model training method | |
| Qiuyue et al. | EEG Signal Processing Based on Incremental Sparse Nonnegative Matrix |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |