Abstract: As biometric systems become widely deployed, the security of identification systems can be easily attacked by various spoof materials. This paper contributes to finding a reliable and practical anti-spoofing method using Convolutional Neural Networks (CNNs) based on the types of loss functions and optimizers. The types of CNNs used in this paper include AlexNet, VGGNet, and ResNet. By using various loss functions including Cross-Entropy, Center Loss, Cosine Proximity, and Hinge Loss, and various loss optimizers which include Adam, SGD, RMSProp, Adadelta, Adagrad, and Nadam, we obtained significant performance changes. We realize that choosing the correct loss function for each model is crucial since different loss functions lead to different errors on the same evaluation. By using a subset of the Livdet 2017 database, we validate our approach to compare the generalization power. It is important to note that we use a subset of LiveDet and the database is the same across all training and testing for each model. This way, we can compare the performance, in terms of generalization, for the unseen data across all different models. The best CNN (AlexNet) with the appropriate loss function and optimizers result in more than 3% of performance gain over the other CNN models with the default loss function and optimizer. In addition to the highest generalization performance, this paper also contains the models with high accuracy associated with parameters and mean average error rates to find the model that consumes the least memory and computation time for training and testing. Although AlexNet has less complexity over other CNN models, it is proven to be very efficient. For practical anti-spoofing systems, the deployed version should use a small amount of memory and should run very fast with high anti-spoofing performance. For our deployed version on smartphones, additional processing steps, such as quantization and pruning algorithms, have been applied in our final model.
Abstract: The neural network quantization is highly desired
procedure to perform before running neural networks on mobile
devices. Quantization without fine-tuning leads to accuracy drop of
the model, whereas commonly used training with quantization is done
on the full set of the labeled data and therefore is both time- and
resource-consuming. Real life applications require simplification and
acceleration of quantization procedure that will maintain accuracy of
full-precision neural network, especially for modern mobile neural
network architectures like Mobilenet-v1, MobileNet-v2 and MNAS. Here we present a method to significantly optimize training with
quantization procedure by introducing the trained scale factors for
discretization thresholds that are separate for each filter. Using the
proposed technique, we quantize the modern mobile architectures of
neural networks with the set of train data of only ∼ 10% of the
total ImageNet 2012 sample. Such reduction of train dataset size and
small number of trainable parameters allow to fine-tune the network
for several hours while maintaining the high accuracy of quantized
model (accuracy drop was less than 0.5%). Ready-for-use models and
code are available in the GitHub repository.
Abstract: The control problem of underactuated spacecrafts has
attracted a considerable amount of interest. The control method for
a spacecraft equipped with less than three control torques is useful
when one of the three control torques had failed. On the other hand,
the quantized control of systems is one of the important research
topics in recent years. The random dither quantization method that
transforms a given continuous signal to a discrete signal by adding
artificial random noise to the continuous signal before quantization
has also attracted a considerable amount of interest. The objective of
this study is to develop the control method based on random dither
quantization method for stabilizing the rotational motion of a rigid
spacecraft with two control inputs. In this paper, the effectiveness of
random dither quantization control method for the stabilization of
rotational motion of spacecrafts with two torque inputs is verified
by numerical simulations.
Abstract: This paper presents a road vehicle detection approach for the intelligent transportation system. This approach mainly uses low-cost magnetic sensor and associated data collection system to collect magnetic signals. This system can measure the magnetic field changing, and it also can detect and count vehicles. We extend Mel Frequency Cepstral Coefficients to analyze vehicle magnetic signals. Vehicle type features are extracted using representation of cepstrum, frame energy, and gap cepstrum of magnetic signals. We design a 2-dimensional map algorithm using Vector Quantization to classify vehicle magnetic features to four typical types of vehicles in Australian suburbs: sedan, VAN, truck, and bus. Experiments results show that our approach achieves a high level of accuracy for vehicle detection and classification.
Abstract: Hammerstein–Wiener model is a block-oriented model
where a linear dynamic system is surrounded by two static
nonlinearities at its input and output and could be used to model
various processes. This paper contains an optimization approach
method for analysing the problem of Hammerstein–Wiener systems
identification. The method relies on reformulate the identification
problem; solve it as constraint quadratic problem and analysing its
solutions. During the formulation of the problem, effects of adding
noise to both input and output signals of nonlinear blocks and
disturbance to linear block, in the emerged equations are discussed.
Additionally, the possible parametric form of matrix operations
to reduce the equation size is presented. To analyse the possible
solutions to the mentioned system of equations, a method to reduce
the difference between the number of equations and number of
unknown variables by formulate and importing existing knowledge
about nonlinear functions is presented. Obtained equations are applied
to an instance H–W system to validate the results and illustrate the
proposed method.
Abstract: The random dither quantization method enables us
to achieve much better performance than the simple uniform
quantization method for the design of quantized control systems.
Motivated by this fact, the stochastic model predictive control
method in which a performance index is minimized subject to
probabilistic constraints imposed on the state variables of systems
has been proposed for linear feedback control systems with random
dither quantization. In other words, a method for solving optimal
control problems subject to probabilistic state constraints for linear
discrete-time control systems with random dither quantization has
been already established. To our best knowledge, however, the
feasibility of such a kind of optimal control problems has not
yet been studied. Our objective in this paper is to investigate the
feasibility of stochastic model predictive control problems for linear
discrete-time control systems with random dither quantization. To
this end, we provide the results of numerical simulations that verify
the feasibility of stochastic model predictive control problems for
linear discrete-time control systems with random dither quantization.
Abstract: Recently, the effectiveness of random dither
quantization method for linear feedback control systems has
been shown in several papers. However, the random dither
quantization method has not yet been applied to nonlinear feedback
control systems. The objective of this paper is to verify the
effectiveness of random dither quantization method for nonlinear
feedback control systems. For this purpose, we consider the attitude
stabilization problem of satellites using discrete-level actuators.
Namely, this paper provides a control method based on the random
dither quantization method for stabilizing the attitude of satellites
using discrete-level actuators.
Abstract: Motivated by recent experimental and theoretical developments, we investigate the influence of embedded quantum dot (EQD) of different geometries (lens, ring and pyramidal) in a double barrier heterostructure (DBH). We work with a general theory of quantum transport that accounts the tight-binding model for the spin dependent resonant tunneling in a semiconductor nanostructure, and Rashba spin orbital to study the spin orbit coupling. In this context, we use the second quantization theory for Rashba effect and the standard Green functions method. We calculate the current density as a function of the voltage without and in the presence of quantum dots. In the second case, we considered the size and shape of the quantum dot, and in the two cases, we worked considering the spin polarization affected by external electric fields. We found that the EQD generates significant changes in current when we consider different morphologies of EQD, as those described above. The first thing shown is that the current decreases significantly, such as the geometry of EQD is changed, prevailing the geometrical confinement. Likewise, we see that the current density decreases when the voltage is increased, showing that the quantum system studied here is more efficient when the morphology of the quantum dot changes.
Abstract: Recently, feedback control systems using random dither
quantizers have been proposed for linear discrete-time systems.
However, the constraints imposed on state and control variables
have not yet been taken into account for the design of feedback
control systems with random dither quantization. Model predictive
control is a kind of optimal feedback control in which control
performance over a finite future is optimized with a performance
index that has a moving initial and terminal time. An important
advantage of model predictive control is its ability to handle
constraints imposed on state and control variables. Based on the
model predictive control approach, the objective of this paper is to
present a control method that satisfies probabilistic state constraints
for linear discrete-time feedback control systems with random dither
quantization. In other words, this paper provides a method for
solving the optimal control problems subject to probabilistic state
constraints for linear discrete-time feedback control systems with
random dither quantization.
Abstract: Performances analysis of remote sensing sensor is required to pursue a range of scientific research and application objectives. Laboratory analysis of any remote sensing instrument is essential, but not sufficient to establish a valid inflight one. In this study, with the aid of the in situ measurements and corresponding image of three-gray scale permanent artificial target, the in-flight radiometric performances analyses (in-flight radiometric calibration, dynamic range and response linearity, signal-noise-ratio (SNR), radiometric resolution) of self-developed short-wave infrared (SWIR) camera are performed. To acquire the inflight calibration coefficients of the SWIR camera, the at-sensor radiances (Li) for the artificial targets are firstly simulated with in situ measurements (atmosphere parameter and spectral reflectance of the target) and viewing geometries using MODTRAN model. With these radiances and the corresponding digital numbers (DN) in the image, a straight line with a formulation of L = G × DN + B is fitted by a minimization regression method, and the fitted coefficients, G and B, are inflight calibration coefficients. And then the high point (LH) and the low point (LL) of dynamic range can be described as LH= (G × DNH + B) and LL= B, respectively, where DNH is equal to 2n − 1 (n is the quantization number of the payload). Meanwhile, the sensor’s response linearity (δ) is described as the correlation coefficient of the regressed line. The results show that the calibration coefficients (G and B) are 0.0083 W·sr−1m−2µm−1 and −3.5 W·sr−1m−2µm−1; the low point of dynamic range is −3.5 W·sr−1m−2µm−1 and the high point is 30.5 W·sr−1m−2µm−1; the response linearity is approximately 99%. Furthermore, a SNR normalization method is used to assess the sensor’s SNR, and the normalized SNR is about 59.6 when the mean value of radiance is equal to 11.0 W·sr−1m−2µm−1; subsequently, the radiometric resolution is calculated about 0.1845 W•sr-1m-2μm-1. Moreover, in order to validate the result, a comparison of the measured radiance with a radiative-transfer-code-predicted over four portable artificial targets with reflectance of 20%, 30%, 40%, 50% respectively, is performed. It is noted that relative error for the calibration is within 6.6%.
Abstract: Bio-based carbon nanotubes (CNTs) have received considerable research attention due to their comparative advantages of high level stability, simplistic use, low toxicity and overall environmental friendliness. New potentials for improvement in heat transfer applications are presented due to their high aspect ratio, high thermal conductivity and special surface area. Phonons have been identified as being responsible for thermal conductivities in carbon nanotubes. Therefore, understanding the mechanism of heat conduction in CNTs involves investigating the difference between the varieties of phonon modes and knowing the kinds of phonon modes that play the dominant role. In this review, a reference to a different number of studies is made and in addition, the role of phonon relaxation rate mainly controlled by boundary scattering and three-phonon Umklapp scattering process was investigated. Results show that the phonon modes are sensitive to a number of nanotube conditions such as: diameter, length, temperature, defects and axial strain. At a low temperature (
Abstract: In this paper, we propose an improved face recognition algorithm using histogram-based features in spatial and frequency domains. For adding spatial information of the face to improve recognition performance, a region-division (RD) method is utilized. The facial area is firstly divided into several regions, then feature vectors of each facial part are generated by Binary Vector Quantization (BVQ) histogram using DCT coefficients in low frequency domains, as well as Local Binary Pattern (LBP) histogram in spatial domain. Recognition results with different regions are first obtained separately and then fused by weighted averaging. Publicly available ORL database is used for the evaluation of our proposed algorithm, which is consisted of 40 subjects with 10 images per subject containing variations in lighting, posing, and expressions. It is demonstrated that face recognition using RD method can achieve much higher recognition rate.
Abstract: Detecting changes in multiple images of the same
scene has recently seen increased interest due to the many
contemporary applications including smart security systems, smart
homes, remote sensing, surveillance, medical diagnosis, weather
forecasting, speed and distance measurement, post-disaster forensics
and much more. These applications differ in the scale, nature, and
speed of change. This paper presents an application of image
processing techniques to implement a real-time change detection
system. Change is identified by comparing the RGB representation of
two consecutive frames captured in real-time. The detection threshold
can be controlled to account for various luminance levels. The
comparison result is passed through a filter before decision making to
reduce false positives, especially at lower luminance conditions. The
system is implemented with a MATLAB Graphical User interface
with several controls to manage its operation and performance.
Abstract: Digital cameras to reduce cost, use an image sensor to
capture color images. Color Filter Array (CFA) in digital cameras
permits only one of the three primary (red-green-blue) colors to be
sensed in a pixel and interpolates the two missing components
through a method named demosaicking. Captured data is interpolated
into a full color image and compressed in applications. Color
interpolation before compression leads to data redundancy. This
paper proposes a new Vector Quantization (VQ) technique to
construct a VQ codebook with Differential Evolution (DE)
Algorithm. The new technique is compared to conventional Linde-
Buzo-Gray (LBG) method.
Abstract: Speaker Identification (SI) is the task of establishing
identity of an individual based on his/her voice characteristics. The SI
task is typically achieved by two-stage signal processing: training and
testing. The training process calculates speaker specific feature
parameters from the speech and generates speaker models
accordingly. In the testing phase, speech samples from unknown
speakers are compared with the models and classified. Even though
performance of speaker identification systems has improved due to
recent advances in speech processing techniques, there is still need of
improvement. In this paper, a Closed-Set Tex-Independent Speaker
Identification System (CISI) based on a Multiple Classifier System
(MCS) is proposed, using Mel Frequency Cepstrum Coefficient
(MFCC) as feature extraction and suitable combination of vector
quantization (VQ) and Gaussian Mixture Model (GMM) together
with Expectation Maximization algorithm (EM) for speaker
modeling. The use of Voice Activity Detector (VAD) with a hybrid
approach based on Short Time Energy (STE) and Statistical
Modeling of Background Noise in the pre-processing step of the
feature extraction yields a better and more robust automatic speaker
identification system. Also investigation of Linde-Buzo-Gray (LBG)
clustering algorithm for initialization of GMM, for estimating the
underlying parameters, in the EM step improved the convergence rate
and systems performance. It also uses relative index as confidence
measures in case of contradiction in identification process by GMM
and VQ as well. Simulation results carried out on voxforge.org
speech database using MATLAB highlight the efficacy of the
proposed method compared to earlier work.
Abstract: In this paper a novel color image compression
technique for efficient storage and delivery of data is proposed. The
proposed compression technique started by RGB to YCbCr color
transformation process. Secondly, the canny edge detection method is
used to classify the blocks into the edge and non-edge blocks. Each
color component Y, Cb, and Cr compressed by discrete cosine
transform (DCT) process, quantizing and coding step by step using
adaptive arithmetic coding. Our technique is concerned with the
compression ratio, bits per pixel and peak signal to noise ratio, and
produce better results than JPEG and more recent published schemes
(like CBDCT-CABS and MHC). The provided experimental results
illustrate the proposed technique that is efficient and feasible in terms
of compression ratio, bits per pixel and peak signal to noise ratio.
Abstract: With the growing of computer and network, digital
data can be spread to anywhere in the world quickly. In addition,
digital data can also be copied or tampered easily so that the security
issue becomes an important topic in the protection of digital data.
Digital watermark is a method to protect the ownership of digital data.
Embedding the watermark will influence the quality certainly. In this
paper, Vector Quantization (VQ) is used to embed the watermark into
the image to fulfill the goal of data hiding. This kind of watermarking
is invisible which means that the users will not conscious the existing
of embedded watermark even though the embedded image has tiny
difference compared to the original image. Meanwhile, VQ needs a lot
of computation burden so that we adopt a fast VQ encoding scheme by
partial distortion searching (PDS) and mean approximation scheme to
speed up the data hiding process.
The watermarks we hide to the image could be gray, bi-level and
color images. Texts are also can be regarded as watermark to embed.
In order to test the robustness of the system, we adopt Photoshop to
fulfill sharpen, cropping and altering to check if the extracted
watermark is still recognizable. Experimental results demonstrate that
the proposed system can resist the above three kinds of tampering in
general cases.
Abstract: This paper presents the voltage problem location
classification using performance of Least Squares Support Vector
Machine (LS-SVM) and Learning Vector Quantization (LVQ) in
electrical power system for proper voltage problem location
implemented by IEEE 39 bus New- England. The data was collected
from the time domain simulation by using Power System Analysis
Toolbox (PSAT). Outputs from simulation data such as voltage, phase
angle, real power and reactive power were taken as input to estimate
voltage stability at particular buses based on Power Transfer Stability
Index (PTSI).The simulation data was carried out on the IEEE 39 bus
test system by considering load bus increased on the system. To verify
of the proposed LS-SVM its performance was compared to Learning
Vector Quantization (LVQ). The results showed that LS-SVM is faster
and better as compared to LVQ. The results also demonstrated that the
LS-SVM was estimated by 0% misclassification whereas LVQ had
7.69% misclassification.
Abstract: A face recognition system is a computer application for automatically identifying or verifying a person from a digital image or a video frame. A lot of algorithms have been proposed for face recognition. Vector Quantization (VQ) based face recognition is a novel approach for face recognition. Here a new codebook generation for VQ based face recognition using Integrated Adaptive Fuzzy Clustering (IAFC) is proposed. IAFC is a fuzzy neural network which incorporates a fuzzy learning rule into a competitive neural network. The performance of proposed algorithm is demonstrated by using publicly available AT&T database, Yale database, Indian Face database and a small face database, DCSKU database created in our lab. In all the databases the proposed approach got a higher recognition rate than most of the existing methods. In terms of Equal Error Rate (ERR) also the proposed codebook is better than the existing methods.
Abstract: This paper presents a new Quality-Controlled, wavelet based, compression method for electrocardiogram (ECG) signals. Initially, an ECG signal is decomposed using the wavelet transform. Then, the resulting coefficients are iteratively thresholded to guarantee that a predefined goal percent root mean square difference (GPRD) is matched within tolerable boundaries. The quantization strategy of extracted non-zero wavelet coefficients (NZWC), according to the combination of RLE, HUFFMAN and arithmetic encoding of the NZWC and a resulting look up table, allow the accomplishment of high compression ratios with good quality reconstructed signals.