Professional Documents
Culture Documents
Received May 19, 2019, accepted June 3, 2019, date of publication June 7, 2019, date of current version July 16, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2921568
ABSTRACT In the field of human health monitoring, intelligent diagnostic methods have drawn much
attention recently to tackle the health problems and challenges faced by patients. In this paper, an efficient and
flexible diagnostic method is proposed, which enables the simultaneous use of a machine learning method
and sparsity-based representation technique. Specifically, the proposed method is based on a convolutional
neural network (CNN) and generalized minimax-concave (GMC) method. First, measured potential signals,
for instance, electroencephalogram (EEG) and electrocardiogram (ECG) signals are directly inputted into
the designed network based on CNN for health conditions classification. The designed network adopts small
convolution kernels to enhance the performance of feature extraction. In the training process, small batch
samples are applied to improve the generalization of the model. Meanwhile, the ‘‘Dropout’’ strategy is
applied to overcome the overfitting problem in fully connected layers. Then, for a record of the interested
EEG or ECG signal, the sparse representation of useful time-frequency features can be estimated via the
GMC method. Case studies of seizure detection and arrhythmia signal analysis are adopted to verify the
performance of the proposed method. The experimental results demonstrate that the proposed method
can effectively identify different health conditions and maximally enhance the sparsity of time-frequency
features.
INDEX TERMS Feature extraction, convolutional neural network, deep learning, sparse representation,
health monitoring.
I. INTRODUCTION tumors [7], [8], adopting ECG signals to identify and diag-
With the development of society and the improvement of nose arrhythmia, heart disease and myocardial infarction
economy, humans have been paying more and more atten- (MI) [9], [10]. These methods use complex medical equip-
tion to their health conditions [1]–[3]. Meanwhile, data- ment to obtain potential signals of human organs, so that it
driven intelligent technologies have achieved great success could analyze and recognize various lesion [11]. Therefore,
in tackling the health problems and challenges faced by it is of great significance to extract representative information
patients [4]–[6], such as introducing EEG signals to mon- from these complicated signals to monitor human health
itor and prevent epilepsy, encephalitis and intracranial conditions [12].
In recent years, much work has been devoted to develop-
The associate editor coordinating the review of this manuscript and ing health monitoring techniques [13]–[15], such as spec-
approving it for publication was Weihong Huang. trum analysis, wavelet transform. Among them, artificial
2169-3536
2019 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 7, 2019 Personal use is also permitted, but republication/redistribution requires IEEE permission. 85985
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
W. He et al.: Simultaneous Human Health Monitoring and Time-Frequency Sparse Representation Using EEG and ECG Signals
in a convolutional layer can be defined as follows: parameter optimization is to minimize the loss function actu-
X ally and the primary way to optimize parameters is stochastic
Xjl = f ( Xil−1 ∗ wlij + blj ) (1) gradient descent (SGD). The method updates the parameters
i∈Mj by calculating the gradient of the loss function, which has a
where Xjl denotes the j th neuron of the l th layer, Mj is the faster convergence speed than other methods. The algorithm
is described as
j th convolution area of the l − 1 th layer, Xjl−1 represents a
∂E
neuron of Mi , wlij refers to the weight matrix of the convolu- θ0 = θ − η (4)
∂θ
tion kernel, bli denotes a trainable bias, ∗ is the computation
where E denotes the loss function in (3), θ is the weight and
of convolution and f (·) represents the activation function.
bias before updating, θ 0 represents the weight and bias after
updating, η refers to the learning rate.
2) POOLING LAYER
After feature extraction in the convolutional layer, the signals
6) ACTIVATION FUNCTION
are input into the pooling layer for feature selection and
In the neural network, extracted features are input into the
information filtering. There are two commonly used pool-
following layer through an activation function. If there is no
ing operations, maximum pooling and average pooling [30],
activation function, the output of each layer is a linear func-
which merely reduce dimensions. A pooling layer has neither
tion of the input signal of this layer. Therefore, the activation
parameters nor weights updating. In the pooling layer, each
function is adopted to add nonlinear factors into the network.
non-overlapping area of the input signals with n ∗ n size is
It allows the neural network to approximate any nonlinear
pooled. The maximum or average value is selected as the
function and make it available in many nonlinear modelS.
result. Finally, the original signal is reduced by n times in
Generally, the commonly adopted activation functions are
both dimensions.
sigmoid functions, RELu functions.
The sigmoid function is widely used in the hidden layers of
3) FULLY CONNECTED LAYER
ANN and the fully connected layers of CNN. Its geometry is
A fully connected layer in CNN is equivalent to a hidden
an S-shaped curve. The RELu function is commonly adopted
layer in a traditional feed forward neural network. It is usually
in the convolutional layers of CNN, which can make the
located at the last part of the hidden layers of CNN and only
backward-transmission process more efficient and simplify
transfers signals to other fully connected layers. The feature
computing. The formula of the Sigmoid function and the
map is converted into a one-dimensional vector in fully con-
RELu function are described, respectively, as follows:
nected layer and then sent into the following layer through an
activation function. The output of a fully connected layer is 1
f (x) = (5)
obtained by: 1 + e−x
f (x) = max(0, x) (6)
yl = f (wl ∗ x l−1 + bl ) (2)
7) DROPOUT STRATEGY
where yl refers to the output of the l th layer, x l−1 denotes the
Overfitting is a widespread problem in training, which results
one-dimensional vector of the l − 1 th layer, wl is a trainable
in poor performance on the testing datasets. Hence, this paper
weight, bl represents a trainable bias and f (·) refers to the
applies the dropout strategy to overcome that problem as well
activation function.
as gradient vanishing in deep neural networks. In the learning
process of a neural network, some neurons of hidden layers
4) LOSS FUNCTION
are randomly ignored. Since the ignored neurons in each iter-
A loss function measures the error between the output of
ation of the training process are different, the importance of
the network and the correct result. Commonly adopted loss
each neuron is balanced. On the contrary, a few high-weight
functions are mean-square error function, cross-entropy func-
neurons will completely control the output if the dropout
tion, negative log likelihood function. In this paper, the cross-
strategy is not adopted. As a result, this method effectively
entropy function is chosen as the loss function, which can be
reduces the structural risk of the network.
defined as
n
1X B. TIME-FREQUENCY SPARSE REPRESENTATION VIA
E =− [yk lntk + (1 − yk )ln(1 − tk )] (3) GMC METHOD
n
k=1
This section introduces a GMC method for time-frequency
where n represents the number of samples, t refers to the sparse representation. The GMC penalty is a class of non-
predicted value and y denotes the true value. convex penalties for sparse-regularized linear least squares
that maintains the convexity of the cost function to be mini-
5) PARAMETER OPTIMIZATION METHOD mized and promotes more strongly sparsity than `1 norm [28].
After calculating by the abovementioned loss function, In time-frequency domain, features are extracted by estimat-
the error has to be transmitted back. In the neural network, ing the short-time Fourier transform (STFT) coefficients,
1) GMC PENALTIES
The scaled Huber function s : R → R is defined as
sb (x) := s(b2 x)/b2 , b 6= 0 (8)
where b is a scaled variable and
1 x 2,
|x| ≤ 1 FIGURE 3. The MC penalty for several values of b.
s(x) = 2 1 (9)
|x| = , |x| ≥ 1
2
For b = 0, the function is defined as s0 := 0. For b 6 = 0, The penalty is parameterized by a scalar b or a matrix B,
the scaled Huber function is given by and the convexity of F depends on b or B being suitably
prescribed. The MC penalty for several values of the scaling
1
b2 x 2 ,
|x| ≤ 1/b2 parameter is shown in Figure 3.
sb (x) = 2 1 (10)
|x| −
, |x| ≥ 1/b 2
2) CONVEXITY CONDITION OF THE OBJECT FUNCTION
2b2
The scaled Huber function can be written as Using GMC penalty in (15), the cost function F can be
n o expressed as
sb (x) = min |v| + 12 b2 (x − v)2 (11)
v∈R F(x)
λ
The minimax-concave (MC) penalty function is defined as 1
= ky − Axk22 +λ kxk1 − min λ kvk1 + kB(x − v)k22
2 2
1 v∈RN
|x| − x 2 , |x| ≤ 1
λ
φ(x) = |x| − s(x) = 1 2 (12) 1 2 2
= max ky − Axk2 +λ kxk1 −λ kvk1 − kB(x − v)k2
,
|x| = 1 v∈RN 2 2
2 1 T T
Therefore, the scalar scaled penalty can be defined as = x (A A−λB B)x + λ kxk1 + max h(x, v)
T
(17)
2 v∈RN
φb (x) = |x| − sb (x) (13) where the function h is affine in x. The last term of (17) is
Note that φ0 (x) = |x|. For b 6 = 0, guaranteed to be convex as it is the pointwise maximum of a
set of convex functions. Thus, F is convex if AT A − λBT B is
1
|x| − b2 x 2 , |x| ≤ 1/b2
positive semidefinite. In other words, the cost function F is
φb (x) = 1
2 (14) convex if
, |x| ≥ 1/b2
2b2 1 T
BT B A A (18)
Let B ∈ RM ×N , the generalized Huber function sB (x) can be λ
defined as B can be set
n o
sB (x) = min kvk1 + 21 kB(x − v)k22 (15) γ /λA, 0≤γ ≤1
p
B= (19)
v∈RN
A multivariate generalization of the minimax-concave The parameter γ controls the non-convexity of the GMC
(GMC) penalty can be defined as penalty. If γ = 0, then B = 0 and the penalty reduces to
the `1 norm. If γ = 1, then BT B = (1/λ)AT A, and the
ψB (x) = kxk1 − sB (x) (16)
non-convexity of the penalty reaches maximum. In practice,
The GMC penalty function is illustrated in Figure 2. the range of 0.5 ≤ γ ≤ 0.8 is used.
TABLE 1. The iterative algorithm for gmc method. Step 3: Feed the training datasets into the formulated net-
work. Calculate the error between the output of the last layer
and the target values. In this process, small batch samples
are applied for training to improve the generalization of the
network.
Step 4: Transmit the error backward and update the weights
along with biases layer by layer. Transmit forward and back-
ward iteratively until the output error meets the predefined
termination criterion.
Step 5: Feed the testing datasets into the well trained
network to classify the health conditions.
Step 6: Construct the time-frequency representation oper-
ator and specify the related parameters in the optimization
problem formulated by the GMC method. Note that the STFT
is adopted and implemented as a normalized tight frame in
this paper.
3) OPTIMIZATION ALGORITHM Step 7: For a record of interested EEG or ECG signal,
An effective approach for the global minimizer of the sparse- the sparse representation of useful time-frequency features
regularized cost function is to formulate the optimization can be extracted via the presented GMC method. Notice that
problem: the GMC method uses a non-convex penalty to maintain the
cost function is strictly convex and induce stronger sparsity.
1 2
(x̃, ṽ) = arg min max F(x, v) = ky − Axk2 +λψB (x, v) IV. EXPERIMENTAL DEMONSTRATIONS
x∈RN v∈RN 2
(20) In this section, the effectiveness of the proposed intelligent
health monitoring method based on CNN and GMC method
The solution can be obtained using the forward-backward is verified on the experimental datasets. The program of the
(FB) algorithm, and the iterative algorithm is shown formulated model is developed in python 3.6.7 with Tensor-
in Table 1. flow deep learning library and run on Windows 10 with a
RTX 2080 Ti GPU.
III. THE PROPOSED METHOD BASED ON CNN AND GMC
A. CLASSIFICATION OF EPILEPTIC CONDITIONS USING
METHOD
EEG SIGNALS
In this section, an efficient and flexible diagnostic method
based on CNN and the GMC method is presented for human 1) DATA PREPARATION
health monitoring. The proposed method enables the simul- EEG signals used in this experiment are collected by Andrze-
taneous use of machine learning method and sparsity-based jak et al. at Bonn University, Germany [29]. These segments
representation technique. More specifically, potential signals were selected and cut out from continuous multi-channel
(e.g., EEG and ECG) of human organs are directly input into EEG recordings after visual inspection for artifacts. The
the proposed model for processing. Firstly, a network based dataset consists of five sets, and each set containing 100 sin-
on CNN is designed for human health conditions classifica- gle channel EEG segments with a duration of 23.6s. The
tion. Then, for a record of interested EEG or ECG signal, sampling rate is set at 173.61 Hz. The health conditions of the
the sparse representation of useful time-frequency features five sets are listed in Table 2 respectively. Typical samples of
can be obtained and enhanced by the GMC method. The the measured EEG signals corresponding to the five types of
overall framework of the proposed health monitoring method health conditions.
is demonstrated in Figure 4. The main steps are summarized In the five conditions above, 500 samples need to be
as follows: divided into training datasets and testing datasets. In order
Step 1: Collect EEG or ECG signals from human organs. to ensure the accuracy of the experiment, these samples were
Segment the signals and construct the training datasets and shuffled and 400 of them were randomly selected as training
testing datasets. samples. Then, the remainder of 100 samples were introduced
Step 2: Design the network based on CNN for health as testing samples. Table 3 shows the sample size and labels
conditions classification. Initialize the parameters of weights, for different health conditions. The labels are applied to com-
biases and learning rates. In the formulated network, small pare with the output of the formulated network to calculate
one-dimensional convolution kernels are adopted in the con- error and optimize parameters.
volutional layers to improve the performance of feature
extraction. In full connection layers, the dropout strategy is 2) MODEL CONSTRUCTION
adopted to invalidate the weights of some neurons, so that it The proposed network based on CNN in this paper is
could overcome the overfitting problem. depicted in Figure 6, which has three convolutional layers,
three polling layers followed by two fully connected layers TABLE 3. Health conditions samples distribution.
and a softmax layer in the end. The convolution layers and
pooling layers aim to extract representative features from the
input signals. The fully connected layers and a softmax layer
is to identify the machinery condition.
In the proposed model, the input one-dimensional signal is
4096 ∗ 1. After three convolutional layers with multiple 7 ∗ 1
small convolution kernels and three pooling layers, the ouput
signals are sixty-four signals of 512 ∗ 1. Then, the feature
maps are flattened and connected with a fully connected layer.
After two fully connected layers, 128 ∗ 1 signal is output to
the next layer. Finally, the original data is divided into five
patterns through the softmax classification layer.
RELu function performs activation in the convolutional
layer. The activation function of the fully connected layer 3) EXPERIMENTAL RESULT
is the Sigmoid function. The pooling layers introduces the As mentioned above, there are 100 testing datasets in total.
maximum pooling method for feature selection and informa- 20, 13, 25, 25 and 17 datasets are in the five condition
tion filtering. The ‘‘Dropout’’ strategy is adopted to over- respectively. The testing datasets are used to verify the clas-
come the overfitting problem in the fully connected layer. sification accuracy of the proposed method. After repeated
The loss function taken in the model is the cross- entropy training, the classification accuracy of the testing datasets
function. Back-propagation is employed for the updates of will be output by the last layer. Figure 7 demonstrates the
model parameters where the SGD method is used. learning classification accuracy of the proposed method based on a
rate is set to 0.0001 and batch size is set to 1. confusion matrix. The row and the column of the confusion
V. CONCLUSION
In this paper, an efficient and flexible diagnostic method
based on CNN and the GMC method is proposed for human
health monitoring. The proposed method enables the simul-
taneous use of machine learning method and sparsity-based
representation technique. It extracts features directly from
potential signals (e.g., EEG and ECG) of human organs.
Firstly, a flexible CNN model is designed for human health
conditions classification. In the formulated network, small
one-dimensional convolution kernels are adopted in the con-
FIGURE 8. Classification result visualization. volutional layers to effectively improve network perfor-
mance. In the training process, small batch samples are
applied for training to improve the generalization of the
model. Besides, the ‘‘Dropout’’ strategy is applied to over-
come the overfitting problem in fully connected layers. Then,
for a record of interested EEG or ECG signal, the sparse rep-
resentation of useful time-frequency features can be achieved
and enhanced by the GMC method.
The analyzed results of two experimental cases with dif-
FIGURE 9. The time domain waveform of the ECG signal.
ferent potential signals, namely seizure detection using EEG
signals and arrhythmia analysis using ECG signals, demon-
strate the superiority of the proposed method. The proposed
method achieves a satisfactory accuracy and can jointly
extracts useful time-frequency sparse features from interested
EEG or ECG signals, which is practical and convenient in
human health monitoring.
Future work will focus on developing automated machine
learning techniques to adaptively meet the requirements of
different diagnostic applications. Acceleration algorithms
will also be adopted to guarantee the intelligent classification
FIGURE 10. The time-frequency spectrogram of the interested ECG signal.
network satisfying real-time applications. Moreover, data-
driven sparse represented technique will be studied which can
avoid the problem of manually specifying parameters.
It can be observed from Figure 11 that the GMC solu-
tion extracted sparse time-frequency spectrogram via the
REFERENCES
GMC method has fewer extraneous noise artifacts. More-
[1] Q. Zhang, D. Zhou, and X. Zeng, ‘‘HeartID: A multiresolution convolu-
over, the characteristic features are clearly revealed in the tional neural network for ECG-based biometric human identification in
time-frequency domain without the interferences of noisy smart health applications,’’ IEEE Access, vol. 5, pp. 11805–11816, 2017.
components. Notice that in this subsection, the results [2] I. Megiddo, A. Colson, D. Chisholm, T. Dua, A. Nandi, and
R. Laxminarayan, ‘‘Health and economic benefits of public financing of
of an interested signal are extracted for the purpose of epilepsy treatment in India: An agent-based simulation model,’’ Epilepsia,
exhibiting useful sparse features in time-frequency domain. vol. 57, no. 3, pp. 464–474, 2016.
[3] P. Swami, T. K. Gandhi, B. K. Panigrahi, M. Tripathi, and S. Anand, [24] J. Wang, M. She, S. Nahavandi, and A. Kouzani, ‘‘Human iden-
‘‘A novel robust diagnostic model to detect seizures in electroencephalog- tification from ECG signals via sparse representation of local seg-
raphy,’’ Expert Syst. Appl., vol. 56, pp. 116–130, Sep. 2016. ments,’’ IEEE Signal Process. Lett., vol. 20, no. 10, pp. 937–940,
[4] C. Venkatesan, P. Karthigaikumar, A. Paul, S. Satheeskumaran, and Oct. 2013.
R. Kumar, ‘‘ECG signal preprocessing and SVM classifier-based abnor- [25] Q. She, K. Chen, Y. Ma, T. Nguyen, and Y. Zhang, ‘‘Sparse representation-
mality detection in remote healthcare applications,’’ IEEE Access, vol. 6, based extreme learning machine for motor imagery EEG classification,’’
pp. 9767–9773, 2018. Comput. Intell. Neurosci., vol. 2018, Oct. 2018, Art. no. 9593682.
[5] U. R. Acharya, H. Fujita, M. Adam, O. S. Lih, V. K. Sudarshan, [26] S. U. Amin, M. Alsulaiman, G. Muhammad, M. A. Bencherif, and
T. J. Hong, J. E. W. Koh, Y. Hagiwara, C. K. Chua, C. K. Poo, and M. S. Hossain, ‘‘Multilevel weighted feature fusion using convolutional
T. R. San, ‘‘Automated characterization and classification of coronary neural networks for EEG motor imagery classification,’’ IEEE Access,
artery disease and myocardial infarction by decomposition of ECG signals: vol. 7, pp. 18940–18950, 2019.
A comparative study,’’ Inf. Sci., vol. 337, pp. 17–29, Jan. 2017. [27] R. M. Mehmood, R. Du, and H. J. Lee, ‘‘Optimal feature selection and deep
[6] B. Chen, Y. Li, and N. Zeng, ‘‘Centralized wavelet multiresolution for exact learning ensembles method for emotion recognition from human brain
translation invariant processing of ECG signals,’’ IEEE Access, vol. 7, EEG sensors,’’ IEEE Access, vol. 5, pp. 14797–14806, 2017.
pp. 42322–42330, 2019. [28] I. Selesnick, ‘‘Sparse regularization via convex analysis,’’ IEEE Trans.
[7] A. Sharmila and P. Geethanjali, ‘‘DWT based detection of epileptic seizure Signal Process., vol. 65, no. 17, pp. 4481–4494, Sep. 2017.
[29] R. G. Andrzejak, K. Lehnertz, C. Rieke, F. Mormann, P. David, and
from EEG signals using naive Bayes and K-NN classifiers,’’ IEEE Access,
C. E. Elger, ‘‘Indications of nonlinear deterministic and finite dimensional
vol. 4, pp. 7716–7727, 2016.
structures in time series of brain electrical activity: Dependence on record-
[8] J. Liu, K. Narasimhan, V. Elamaran, N. Arunkumar, M. Solarte, and
ing region and brain state,’’ Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat.
G. Ramirez-Gonzalez, ‘‘Clinical decision support system for alcoholism
Interdiscip. Top., vol. 64, p. 061907, 2001.
detection using the analysis of EEG signals,’’ IEEE Access, vol. 6,
[30] MIT-BIH Arrhythmia Database. Accessed: May 17, 2019. [Online]. Avail-
pp. 61457–61461, 2018.
able: https://physionet.org/physiobank/database/mitdb/
[9] D. Craven, B. McGinley, L. Kilmartin, M. Glavin, and E. Jones, ‘‘Adap-
tive dictionary reconstruction for compressed sensing of ECG sig-
nals,’’ IEEE J. Biomed. Health Inform., vol. 21, no. 3, pp. 645–654,
May 2017. WANGPENG HE received the B.S. and Ph.D.
[10] S. Raj and K. C. Ray, ‘‘Sparse representation of ECG signals for auto- degrees in mechanical engineering from Xi’an
mated recognition of cardiac arrhythmias,’’ Expert Syst. Appl., vol. 105, Jiaotong University, Xi’an, China, in 2007 and
pp. 49–64, Sep. 2018. 2016, respectively. In 2014, he was appointed as
[11] M. Rizwan, B. M. Whitaker, and D. V. Anderson, ‘‘AF detection from ECG a Visiting Scholar with the New York Univer-
recordings using feature selection, sparse coding, and ensemble learning,’’ sity, New York, NY, USA. He joined the School
Physiol. Meas., vol. 39, no. 12, Dec. 2018, Art. no. 124007. of Aerospace Science and Technology, Xidian
[12] J. Li, G. Deng, W. Wei, H. Wang, and Z. Ming, ‘‘Design of a real-time University, where he is currently an Assistant
ECG filter for portable mobile medical systems,’’ IEEE Access, vol. 5,
Professor. His research interests include signal
pp. 696–704, 2017.
processing, machine vision, sparsity-based signal
[13] A. Bhattacharyya, R. B. Pachori, A. Upadhyay, and U. R. Acharya,
processing, machine learning, and machinery fault diagnosis.
‘‘Tunable-Q wavelet transform based multiscale entropy measure for auto-
mated classification of epileptic EEG signals,’’ Appl. Sci., vol. 7, no. 4,
p. 385, 2017.
[14] B. van Ginneken, ‘‘Fifty years of computer analysis in chest imaging: Rule- GENG WANG received the B.S. degree in mea-
based, machine learning, deep learning,’’ Radiological Phys. Technol., surement and control technology and instrument
vol. 10, no. 1, pp. 23–32, 2017. major from the School of Electro-mechanical
[15] Y. Yu and Z. Sun, ‘‘Sparse coding extreme learning machine for classifi- Engineering, Xidian University, in 2018, where
cation,’’ Neurocomputing, vol. 261, pp. 50–56, Oct. 2017. he is currently pursuing the M.S. degree with the
[16] H. Greenspan, B. van Ginneken, and R. M. Summers, ‘‘Deep learning School of Aerospace Science and Technology. His
in medical imaging: Overview and future promise of an exciting new main research interests include deep learning and
technique,’’ IEEE Trans. Med. Imag., vol. 35, no. 5, pp. 1153–1159, neural networks.
May 2016.
[17] I. Ullah, M. Hussain, E.-U.-H. Qazi, and H. Aboalsamh, ‘‘An auto-
mated system for epilepsy detection using EEG brain signals based
on deep learning approach,’’ Expert Syst. Appl., vol. 107, pp. 61–71,
Oct. 2018. JIE HU received the B.S. degree in automa-
[18] U. R. Acharya, H. Fujita, S. L. Oh, Y. Hagiwara, J. H. Tan, and tion engineering major from the Xi’an Univer-
M. Adam, ‘‘Application of deep convolutional neural network for auto- sity of Architecture and Technology, Xi’an, China,
mated detection of myocardial infarction using ecg signals,’’ Inf. Sci., in 2018. She is currently pursuing the M.S. degree
vol. 415, pp. 190–198, Nov. 2017. with the School of Aerospace Science and Tech-
[19] R. San-Segundo, M. Gil-Martín, L. F. D’Haro-Enríquez, and nology, Xidian University. Her main research
J. M. Pardo, ‘‘Classification of epileptic EEG recordings using signal interests include sparsity-based signal processing
transforms and convolutional neural networks,’’ Comput. Biol. Med., and machinery fault diagnosis.
vol. 109, pp. 148–158, Jun. 2019.
[20] M. Luo and K. Zhang, ‘‘A hybrid approach combining extreme learning
machine and sparse representation for image classification,’’ Eng. Appl.
Artif. Intell., vol. 27, pp. 228–235, Jan. 2014.
[21] M. Yang, L. Zhang, X. Feng, and D. Zhang, ‘‘Sparse representation based CHENG LI received the B.S. degree from the
Fisher discrimination dictionary learning for image classification,’’ Int. J. School of Electro-mechanical Engineering, Xidian
Comput. Vis., vol. 109, no. 3, pp. 209–232, Sep. 2014. University, Xi’an, China, in 2013, and the M.S.
[22] W. He, B. Chen, N. Zeng, and Y. Zi, ‘‘Sparsity-based signal extraction degree from the School of Aerospace Science and
using dual Q-factors for gearbox fault detection,’’ ISA Trans., vol. 79, Technology, Xidian University, in 2017, where he
pp. 147–160, Aug. 2018. is currently pursuing the Ph.D. degree with the
[23] Z. Zhang, D. Luo, Y. Rasim, Y. Li, G. Meng, J. Xu, and C. Wang, ‘‘A vehi- Intelligent Control and Image Engineering Insti-
cle active safety model: Vehicle speed control based on driver vigilance tute. He focuses on researching in multi-scale
detection using wearable EEG and sparse representation,’’ Sensors, vol. 16, image representation.
no. 2, p. 242, Feb. 2016.
BAOLONG GUO received the B.S., M.S., and FENGPING LI is currently a Distinguished
Ph.D. degrees from Xidian University, in 1984, Professor with the Institute of Laser and Opto-
1988, and 1995, respectively, all in communication electronics Intelligent Manufacturing, Wenzhou
and electronic systems. University, Wenzhou, China. His research fields
From 1998 to 1999, he was a Visiting Scholar mainly include adaptive control algorithm and
with Doshisha University. He is currently a Full machine vision, laser processing application, and
Professor with the School of Aerospace Sci- industrial control systems.
ence and Technology, Xidian University, Xi’an,
China. His current research interests include pat-
tern recognition, intelligent information process-
ing, image processing, and video communication. He has published over
100 academic papers and organized several international conferences as the
Co-Chair.