9 Articles Published

1. COMPARATIVE ANALYSIS BETWEEN RZ-DPSK AND NRZ-DPSK AT 10GB/S IN OPTICAL COMMUNICATION SYSTEM BY USING DCF

Authors: R.K. SINGH , Shashi Jawla

Abstract: In this paper, the performance ofRZDPSK and NRZ-DPSK modulation techniques with/without dispersion compensating fiberin optical communication system has been analyzed. Dispersion compensating fibers are developingvery fast to gain the dispersion compensation of real value and perfect coupling to fibers to sustainwavelength stability. A effort is shown to improve the performance of the system in terms of bit error rate (BER) and Q-value by proposing a combination of fibre length by using SMF-DCF for RZ-DPSK and NRZ-DPSK modulation techniques. The combination of SMF-DCF improves the optical distance. The BER and Qvalue are observed for RZ-DPSK and NRZ-DPSK at the transmission rate of 10 Gb/s.

Keywords: Dispersion compensation, BER, Qvalue, DCF, SMF, RZ-DPSK and NRZ-DPSK.

Pages: 1 - 3 | DOI: 10.15224/978-1-63248-184-9-01

2. A HEURISTIC GRAPH PARTITIONING METHOD TO MINIMIZE REMOTE COMMUNICATION COSTS

Authors: Shing Ki Wong , Siu Ming Yiu

Abstract: Existing graph partitioning algorithms rarely focus on the choice of the replicas. Most of them make use of random hashing to allocate vertices to partitions due to its convenience and simplicity. However, such random property may result in undesirable partitioning results with huge communication cost if many high degree vertices are being replicated. We address this problem and propose a greedy algorithm to minimize the number of replications of high degree vertices and at the same time minimize the replication factor by sorting the vertices before allocating them to different partitions. We compare our algorithm with one of the well performed existing graph partitioning algorithms, Powers hybrid-cut, and prove that our algorithm gives better results in most practical situations. Experimental results show that our algorithm gives much lower replication factor compared to Powers hybrid-cut algorithm generally in random graphs, power-law graphs and real-life graphs.

Keywords: graph, partitioning, heuristic, replica, insert

Pages: 4 - 10 | DOI: 10.15224/978-1-63248-184-9-02

3. OPTIMAL ROUTING AND SCHEDULING FOR UNRELIABLE MARKOVIAN SYSTEMS MODELED WITH TIMED PETRI NETS.

Authors: Dimitri Lefebvre , Oussama Hayane

Abstract: This paper addresses the topic of robust routing and scheduling for parallel systems that suffer from operation interruptions and unreliable resources. The proposed approach uses a partially controllable extension of timed Petri nets as a model of the deterministic and stochastic behaviours. The mean job durations are first evaluated. Then a mean timed reachability graph is proposed to encode the timing aspects and Dijkstra algorithm is used to solve the routing and scheduling problems related to operation and resource failure rates.

Keywords: Discrete event systems, Timed Petri nets, routing, scheduling, failures.

Pages: 16 - 22 | DOI: 10.15224/978-1-63248-184-9-04

4. A NOVEL FABRICATION TECHNIQUE FOR MOS STRUCTURE BY OXIDATION METHODS USING CHEMICAL SOLUTION

Authors: Kazuki Harada , TOMO UENO , Yositaka Iwazaki

Abstract: So far, the diameter of a Si wafer has been increased to reduce the manufacturing cost of semiconductor chips. However, it is considered that this makes it difficult to form uniform thickness thin films on the entire surface of the wafer by a conventional film forming method. Therefore, we have proposed an oxidation method using chemical solution that can form a film just by immersing it in a chemical solution. In this study, HfO2, a high-k material with a high dielectric constant, was formed on a Si substrate by chemical solution oxidation, and its electrical properties were compared with those of a thermal oxidation method. As a result, the MOS structure was successfully fabricated by oxidation method using chemical solution. However, this method has a problem of improving interface characteristics.

Keywords: Si, HfO2 , chemical solution oxidation method, MOSFET.

Pages: 23 - 25 | DOI: 10.15224/978-1-63248-184-9-05

5. SINGULAR SPECTRUM TRANSFORMATION ON MOTION/EMOTION MATRIX TO MEASURE COLLABORATION

Authors: Fumiko Harada , Hiromitsu Shimakawa , Yuto Hattori

Abstract: In this study, we propose the method estimating collaborative state by the similarity of motion characteristics and emotion characteristics changes. The group learning based on collaboration among students are introduced in many university classes. However, it may not function as a group and teachers evaluate only result of group learning. Our method focusses on similarity of motion and emotion characteristics change. The change points are extracted from these time series data by applying singular spectrum transformation (SST). In addition, the similarity points of change extracted by comparing the change points among students. The collaborative states are estimated in a certain period from the number of similarity points of change using machine learning. The experimental result indicates this method can identify between non-collaborative state and collaborative state. The method facilitates teachers to evaluate and guide the groups.

Keywords: Group learning, Time series data

Pages: 26 - 31 | DOI: 10.15224/978-1-63248-184-9-06

6. ESTIMATION OF THE ELDERLY GETTING OUT OF BED USING A SHEET-TYPE PRESSURE SENSOR

Authors: Fumiko Harada , Hiromitsu Shimakawa , Atsushi Hagihara

Abstract: In this paper, we propose a method of estimating the care recipient's posture on the bed in detail and predicting the care recipient's getting out of the bed. In order to grasp the movement and posture of the care recipient on the bed in real time, a classifier that estimates the posture of the care recipient on the bed is created by machine learning. The explanatory variables were the position of the center of gravity of each part of the care recipient, the average value of the velocity and acceleration of the position of the center of gravity over a certain period of time, and the variance and covariance of the position of the center of gravity, and 16 types of body positions were classified. As a result of Random Forest, we were able to classify with an F value of 0.72 or more. This suggests that it is possible to estimate the getting out of the bed of a care recipient in real time.

Keywords: Machine Learning, Elderly Care, Sensing, Pressure Sensor

Pages: 32 - 37 | DOI: 10.15224/978-1-63248-184-9-07

7. FABRICATION OF HIGH-PERFORMANCE GE-MOS STRUCTURE USING TIO2 INSERTED LAYER IN GATE INSULATOR FOR HIGH SPEED SWITCHING DEVICE

Authors: Daichi Hasegawa , Kenichiro Toyoda , Ryuichiro Watanabe , TOMO UENO , YOSHITAKA IWAZAKI

Abstract: In recent years, combination of high mobility (High- μ) materials and high dielectric constant (High-k) insulating films has attracted attention as a technique for future switching devices such as high speed switching, low voltage operation and so on. Using a Ge substrate having higher carrier mobility than that of Si as a high-μ material, it is expected that the operation speed of the device will be increased. On the other hand, it is necessary to introduce a high-k material, for example HfO2, Al2O3, in order to increase the capacitance value while maintaining the insulating film thickness, and the like have been studied so far. However, it is difficult to obtain good interface characteristics by direct deposition of HfO2 or Al2O3 on Ge substrates. In this study, we tried to introduce another high dielectric constant material TiO2, whose dielectric constant value is approximately 15 times higher(~60) than that of SiO2(~3.9). However, the energy band gap of TiO2 is very small at 3.2

Keywords: Switching device, Ge-MOS, TiO2 , Al2O3

Pages: 38 - 40 | DOI: 10.15224/978-1-63248-172-6-08

8. STABLE PATH SELECTION NODES BASED ON REINFORCEMENT LEARNING ALGORITHM TO ENHANCE THE NODE AND LINK STABILITY IN MANET

Authors: G Sulakshana , M.V.Bramhananda Reddy , MAmutha , MS Deepthi Reddy

Abstract: In MANET the route discovery in AODV (Ad hoc on demand distance vector routing protocol) is performed on demand and it involves flooding the route request packet. The existing path may be unavailable as the node moves out of coverage due to its rapid node mobility or node fail as it energy level is drained. The proposed system implemented by a Reinforcement Learning using machine learning in which establishing the stable path by selecting these nodes that have node and link stability. A node performs the activity as action and a state. Action is apply the external condition and when given condition is satisfying than it provides reward is a state. This process will eliminate re-route discovery process and there by reduces additional routing overhead caused by the re route discovery. It provides a high degree of stability, increases packet delivery ratio and reduces routing overhead and delay

Keywords: STABLE, PATH, SELECTION, NODES, REINFORCEMENT, LEARNING

Pages: 41 - 49 | DOI: 10.15224/978-1-63248-172-6-09

9. AUTOMATIC DETECTION OF STACKING IN TASK USING VR SPACE

Authors: Fumiko Harada , Hiromitsu Shimakawa , Shota Hashimura

Abstract: Lessons tailored to the situation of understanding have a high learning effect, but it is difficult to keep track of students' understanding. In this paper, we propose a method that use VR to estimate whether students in a task have outlooks for the answer from body movements. As a result of the experiment, we succeeded in estimating the outlook of the answer with 75% accuracy but cannot found that the prospect itself was correct or not. This makes it possible to detect a state that the student is stuck in the task and the understanding is stagnant, and it is possible to provide a lesson tailored to the student's understanding state.

Keywords: Education, Virtual Reality

Pages: 11 - 15 | DOI: 10.15224/978-1-63248-184-9-03