Abstract:The traditional SVM parameter optimization use a wide range of traversal algorithm or some intelligent iterative algorithm, generally need to consume great deal of time, it is not applicable to optimization parameters of big data sets .To get around this ,This paper presents a strategy of stepwise optimize parameters based on the contour plots of cross-validation accuracy. Generate 25 parameter combinations uniformly, output the contour plots of cross-validation accuracy, then narrowing the optimal region of the parameters, proceeding stepwise optimizing parameter, until the optimal parameters were found. Finally, use a 13910*128 data set to verify the algorithm, compare with the traditional grid search algorithm, the new method not only greatly shorten the time of SVM parameters optimization, and it can find the better parameter than the traditional methods. This paper provides an effective solution to optimize SVM parameters especially for large data.
Keywords: support vector machine, SVM , parameter optimization , big data
Abstract: This paper proposes a new method for learning English, by using English-subbed Japanese sounded cartoons (anime). In the proposed system, the English subtitles are recognized into text in real time and sent to the text-to-speech system. After that, the sound is generated and can be heard immediately, after Japanese sound. It means that, while the learner is watching anime, learner can hear the sound of English subtitles in immediately after he/she heard the sound of original Japanese. By hearing Japanese and English alternately a lot, finally the learner will be able to understand English. This system is suitable for everyone who wants to learn English while watching anime, including the child.
Keywords: English learning system; anime; subtitle recognition; text-to-speech
Abstract: For traditional arterial road coordinated control with predetermined cycle, split, and offset
cannot adapt for dynamic real-time traffic flow. In addition traditional vehicle actuated coordinated
control will fail in oversaturated traffic status and other limitations. To overcome these limitations, this paper adopts active electronic tags to get the update ratio of the vehicles in the test range, changes time interval between vehicles as control basis in traditional vehicle actuated control theory, puts forward putting vehicle density as whether green light extension control basis by using RFID electronic tag in vehicle actuated control. When the vehicle density is higher than the setting threshold value, green light continues, on the contrary, switching the phase. The method not only can overcome the drawback that traditional vehicle actuated coordinated control will fail in oversaturated traffic status, but also can effectively improve the utilization of green light and decrease the average delay in the intersection, which has been verified by simulated experiment through VISSIM
Keywords: RFID; Coordinated control; Vehicle actuated control; Average delay
School of Electrical and Electronic Engineering, Shanghai Institute of Technology, Shanghai, China & School of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai, China & Department of Computer Science and Technology, Tsinghua University, Beijing, China
Abstract: This paper designs and realizes the hardware of a PMSM control system, in which DSP is the main controller and CPLD is the auxiliary logic controller. The proposed power driver consists of three independent H-bridge inverter drive circuits. The hardware are designed including the motor phase current detection, main voltage monitoring, signal acquisition of both rotor position and speed, overcurrent protection and other related circuits. The experimental results show that the motor controller is reliable and stable with strong practicability.
Keywords: Control System; CPLD; DSP; PMSM
Abstract: Scale Invariant Feature Transform (SIFT) is a method to detect and match invariant feature points on images, and is robust against contrast, rotation, and scale changes. However, SIFT cannot find many correct matching points between affine transformed images because this method employs Gaussian function for scale parameter which specifies a circle area on image planes. In this paper, we propose a method using Bi-dimensional Empirical Mode Decomposition (BEMD) for keypoint detection, where a given image is decomposed into Intrinsic Mode Functions (IMFs). Our method also employs Affine Moment Invariants (AMIs) instead of SIFT’s feature values. As a result, the proposed method detects more matching points than SIFT in a steep affine
Keywords: Empirical Mode Decomposition, Affine Moment Invariants.
Abstract: LED probes are essential for testing the quality of LEDs, gaining its attention among industrial applications. Disturbance factors such as dust or noise affections may occur during the manufacturing process of the LED probes, which leads angle error to increase. With the increasing demand for LED probes, higher precision and efficiency are expected by users. Efficient method for edge detection and the preciseness of angle is crucial in our study. The previous study presented a method using Scharr edge detection and adaptive reconstruction. In this paper, we add a new method based on various segmentation and the averaging of sub-pixels (VSAS). Experimental results indicate that this method provides higher precision, with and an average error less than 1% compared to the other previous methods.
Keywords: Angle accuracy, edge detection, led, probe, sub-pixels, various segmentation
Abstract: ECG-based based human recognition is increasingly becoming a popular modality for
biometric authentication. Two important features of ECG signals are liveliness and the robustness
against falsification. However, ECG features vary due to muscle flexure, baseline wander, and other sources of noise. This paper presents a new method which extracts multiscale geometric features from ECG signals and apply them for human identification. A non-linear filter is applied for preprocessing the ECG signal. The refined ECG signal is then divided into multiple segments and feature matrix is computed by multiscale pattern extraction technique. Feature matrix is finally applied to a simple minimum distance to mean classifier adopting leave-one-out procedure. An experiment with 60 ECG signals from 60 subjects shows promising performance of the proposed method compared to the conventional ECG features.
Keywords: Binary patterns, multiscale representation, supervised classification, and human
Abstract: With the invention of Microsoft Kinect sensor, human-computer interaction is gaining its
attention and becoming available for widespread use. The previous study presented a method of
Kinect-based mid-air handwritten digit recognition for Android smart phones with a recognition
accuracy of only about 94.6%. In this paper, we propose an improved method based on the
normalizing and scaling of path order coordinates. With that, the proposed method leads to accuracy elevation and executing time reduction. Experimental results show an average recognition accuracy rate of 96.8% was achieved for each number.
Keywords: Accuracy rate, handwritten digit recognition, human-computer interaction, Kinect sensor support vector machine
Abstract: The use of Brain-Computer Interface (BCI) has been increasing exponentially in the recent years due to the use of low-cost commercial Fast Fourier Transform (FFT) based EEG reading devices with nonclinical accuracy for consumer application development. Also, the design and implementation of 3D virtual environments for BCI training purposes has proven to be effective due to the high interaction with the end user and the assistance for recreating a specific type of signal or behavior. The aim of this paper is to present a method and the results of applying a binary Density Based Support Vector Machine (DBSVM) Classifier in a 3D virtual environment designed for interaction with EEG predefined signal patterns. The environment trains the classifier by taking 180 second EPOCHs and classifying them into a successful/unsuccessful attempt per test subject. The applications can be extended for implementing mind-wave pattern password or tracing a specific set of mind-based commands for virtual path tracing purposes. The tested SVM had a success rate of 60%. Further work includes the study of different classifier features and implementation of a dynamic classifier.
Keywords: BMI, MindWave, SVM, Pattern Recognition, Machine Learning
Abstract: This paper presents electrical impedance imaging based on ultrasonic B-mode image for
detecting breast tumor. In the proposed system, a thin film with 16 electrodes was pasted on the
surface of the ultrasound probe. The proposed system allows both the electrical conductivity image by sixteen film electrodes and ultrasound image by the ultrasonic probe at same area. The electrical conductivity image is improved by using ultrasound image as the prior information. In the simulation and experiment, the electrical conductivity distribution at the simple breast model was reconstructed by the proposed system. The comparison between the electrical images with and without the ultrasound image as the prior information was presented. As results, it was found that the electrical conductivity distribution was clearly improved by using ultrasonic image as the prior information although the proposed system must be improved.
Keywords: Ultrasonic imaging, electrical impedance imaging, linear array sensor
Abstract: This paper represents creating and analyzing of an adequate hydraulically balanced Water Distribution Network (WDN) based on Geographical Information System (GIS) and Epanet softwares. Fist a proper and functional WDN has been built and skeletonized by network analyst methodology in ArcGIS. Then the project scenario has been imported into Epanet with simple text or Epanet's readable format consisted of multiple physical or non-physical WDN objects and its characteristics. These all in order to determine analyze and simulate hydraulic parameters of WDN. Second, we analyzed and determined water flows, water in each pipe, water flows direction as well as pressure heads at each node in the network with other required hydraulics parameters. The advantages of this method are simple math and self-correction.
Concisely, the context of this work is to use GIS and Epanet-based methodologies as well as
solution of closed-loop network problems. Used continuity and energy conservation methods due to determination of pressure heads at each node, water flows and water flows direction at each pipe in the network. As a result got a satisfied hydraulically balanced and a systematic WDN appropriately as the implications of this paper approaches consequently.
Keywords: Geographical Information System (GIS); Water Distribution Network (WDN), Hydraulic,
Abstract: Lacking proper theory and design, big data has continued to grow in a chaotic way and are now beyond human recognition and control. To address this, we have been researching an application of the concept of homotopy preservation in homotopy type theory to remove the bottleneck in big data processing, originating from combinatorial explosion of information, which makes use of the seven layers, from homotopy to presentation, of the incrementally modular abstraction hierarchy (IMAH).
We developed the Cellular Data System (CDS) as an implementation of the IMAH in our previous
work. In this paper, we introduce a position information formula, which identifies the relative position of a specified factor in a formula in CDS. This function has an important role for preserving homotopy in actual changes of objects. We show the effectiveness of the formula by taking up a case study of component configuration information management in manufacturing. In the case study, it is demonstrated that design component configuration information and manufacturing component configuration information are consistently managed by preserving homotopy, even when component configuration frequently changes.
Keywords: Position information formula, Component configuration information management,
Incrementally Modular Abstraction Hierarchy, Cellular Data System
Abstract: The classic web-based applications operate only when connected to the network. Many
realities in a field require a web-based application that is applicable even in case of offline. We have proposed the Automated HTML5 Offline Service (AHOS), presented as the integration of advanced services, developing with HTML5 Application Programming Interfaces (APIs) to provide web-based applications with the ability to work offline. When the AHOS web-based application visits at the web server for the first time, the web server will notify the application the list of files required to be downloaded. Then after being downloaded, the web application can work successfully and continuously even though the network connection from the client to the server is unavailable. Moreover, if the connection to the server is re-connected, any changes that have been made during offline will be automatically uploaded. The present study describes the requirements and implementation stages of AHOS concept for web-based applications. Then, several current status and challenges of AHOS concept are also described. The performance analyses of the AHOS concept are performed in a case of web-based maintenance Decision Support System (DSS) for Small and Medium Industries (SMIs). The results of the study are very useful in providing in-depth understanding of the advantages and limitations, and as the future directions in applying this AHOS concept to other webbased applications.
Abstract: Thus far, attention has been paid to the relation between pain and anxiety, which has been studied. On the other hands, the earlier studies have hinted at the importance of considering mental and physical health from a holistic perspective, while taking into consideration the principles that prescribe the chaotic behavior of living organisms. Therefore, the purpose of this study is to use the respiration and pulse waves to examine the nature of the chaotic connection between biosignals involved in people’s mental and physical conditions, particularly those involved in the psychological trait of anxiety under a painful stimulus in this case. The results showed that with the high anxiety group, the extent of the synchronicity between their respiration and pulse waves under a painful stimulus increased, while this decreased for the low anxiety group. In other words, chaos dynamics for living systems are expressed in synchronous phenomena for the LLE for respiration and pulse waves. It also implied that these dynamics are prescribed by trait anxiety under a painful stimulus. This has opened up the possibility that, in the future, the cross-correlation function for LLE in pulse waves and respiration will make contributions to treating and assessing chronic pain in the field of clinical medicine.
Keywords: Pain; Anxiety; Chaos; Synchronization