• IEMEK Journal of Embedded Systems and Applications
• /
• v.5 no.2
• /
• pp.57-66
• /
• 2010

In this paper, we propose an enhanced voice recognition algorithm using adaptive MFCC(Mel Frequency Cepstral Coefficients) and neural network. Though it is very important to extract voice data from the raw data to enhance the voice recognition ratio, conventional algorithms are subject to deteriorating voice data when they eliminate noise within special frequency band. Differently from the conventional MFCC, the proposed algorithm imposed bigger weights to some specified frequency regions and unoverlapped filterbank to enhance the recognition ratio without deteriorating voice data. In simulation results, the proposed algorithm shows better performance comparing with MFCC since it is robust to variation of the environment.

• Proceedings of the KIEE Conference
• /
• 2006.04a
• /
• pp.132-134
• /
• 2006

To improve equipment throughput and device yield, a malfunction in plasma equipment should be accurately diagnosed. A recognition model for plasma diagnosis was constructed by applying neural network to scanning electron microscope (SEM) image of plasma-etched patterns. The experimental data were collected from a plasma etching of tungsten thin films. Faults in plasma were generated by simulating a variation in process parameters. Feature vectors were obtained by applying direct and wavelet techniques to SEM Images. The wavelet techniques generated three feature vectors composed of detailed components. The diagnosis models constructed were evaluated in terms of the recognition accuracy. The direct technique yielded much smaller recognition accuracy with respect to the wavelet technique. The improvement was about 82%. This demonstrates that the direct method is more effective in constructing a neural network model of SEM profile information.

• Proceedings of the KIEE Conference
• /
• 2006.04a
• /
• pp.138-140
• /
• 2006

A new diagnosis model was constructed by combining atomic force microscopy (AFM), wavelet, and neural network. Plasma faults were characterized by filtering AFM-measured etch surface roughness with wavelet. The presented technique was evaluated with the data collected during the etching of silicon oxynitride thin film. A total of 17 etch experiments were conducted. Applying wavelet to AFM, surface roughness was detailed into vertical, horizon%at, and diagonal components. For each component, neural network recognition models were constructed and evaluated. Comparisons revealed that the vertical component-based model yielded about 30% improvement in the recognition accuracy over others. The presented technique was evaluated with the data collected during the etching of silicon oxynitride thin film. A total of 17 etch experiments were conducted

• Proceedings of the KIEE Conference
• /
• 1995.11a
• /
• pp.293-295
• /
• 1995

In this article a fuzzy controller and neural network adaptive observer is proposed and applied to the case of induction motor control. The proposed observer which comprises neural network flux observer and neural network torque observer is trained to learn the flux dynamics and torque dynamics and subjected to further on-line training by means of a backpropagation algorithm. Therefore it has been shown that the robust control of induction motor neglects the rotor time constant variations.

• Proceedings of the KIEE Conference
• /
• 2008.07a
• /
• pp.958-959
• /
• 2008

This paper is proposed a high performance speed control of the Interior Permanent Magnet Synchronous Motor through the Neural Network SV-PWM. SV-PWM is controlled using Neural Network control. SV-PWM can be maximum used maximum dc link voltage and is excellent control method due to characteristic to reducing harmonic more than others. Neural Network control has a advantage which can be robustly controlled. Simulation results are presented to show the validity of the proposed algorithm.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.15 no.3
• /
• pp.39-44
• /
• 2006

In this study, we nodestructive test based on ultrasonic test as inspection method and compared backpropagation neural network(BPNN) with probabilistic neural network(PNN) as pattern recognition algorithm of welding flasw. For this purpose, variables are applied the same to two algorithms. Where, feature variables are zooming flaw signals of reflected whole signals from welding flaws in time domain. Through this process, we confirmed advantages/disadvantages of two algorithms and identified application methods of two algorithms.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.04a
• /
• pp.484-490
• /
• 2000

The abjective of the study is the development of the system for effective prediction of residual stresses using the backpropagation algorithm from the neural network. To achieve this goal, the series experiment were carried out and measured the residual stresses using the sectional method. Using the experimental results, the optional control algorithms using a neural network should be developed in order to reduce the effect of the external disturbances on during GMA welding processes. Then the results obtained from this study were compared between the measured and calculated results, the neural network based on backpropagation algorithm might be controlled weld quality. This system can not only help to understand the interaction between the process parameters and residual stress, but also improve the quantity control for welded structures.

• Proceedings of the KIPE Conference
• /
• 2003.11a
• /
• pp.178-181
• /
• 2003

This paper presents a position sensorless control system of Switched Reluctance Motor(SRM) using Neural Network. The basic algorithm of this scheme is based on the flux linkage characteristic according to the phase currents and rotor position. A sufficient experimental data was used for neural network training. The proposed position sensorless control system was realized using TMS320F240 DSP. The experimental result shows some good results, and verifies the possibility of this algorithm.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.10b
• /
• pp.218-222
• /
• 2006

우수한 능력의 인공지능 개체로 구성된 게임은 그렇지 못한 게임에 비해 더 나은 흥미를 사용자에게 제공할 수 있다. 미국 Valve사의 Half-Life, Counter-Strike 및 한국 Dragonfly사의 Special-Force와 같은 실시간 FPS 전투게임에서 상대편에 대한 검색 및 목표 화하는(Targeting) 기법은 인공개체의 전투력에 중요한 하나의 요소이다. 하지만 이 같은 경우의Targeting은 정적인 대상에 대한 것이 아니라 동적인 대상에 대한 것이므로 단순한 산술 계산으로는 실용적인 효과를 내기 힘들다. 본 논문에서는 Neural Network를 이용한 학습기법을 사용하여 동적인 개체에 대한 효과적인 Targeting기법을 제안한다. 제안한 기법은 매 순간 변화하는 상황정보와 Virtual bullet이라는 가상 미사일 개념을 활용하여 학습 Data를 모델링한 후 Neural Network로 학습시켜 효과적인 Targeting이 가능하도록 구현하였다. 제안한 기법은 Java기반의 탱크전투 시뮬레이션 Framework인 Robocode에 적용하여 그 성능을 평가하였다. 제안된 기법으로 제작된 Robot(Crystal 1.0)은 ‘2006 Robocode Korea Cup에서 우승을 차지하였다.

• Proceedings of the KIEE Conference
• /
• 1990.07a
• /
• pp.485-491
• /
• 1990

We propose a minimum-time path planning soheme for the robot manipulator using Hopfield neural network. The minimum-time path planning, which can allow the robot system to perform the demanded tasks with a minimum execution time, may be of consequence to improve the productivity. But most of the methods proposed till now suffers from a significant computational burden and thus limits the on-line application. One way to avoid such a difficulty is to apply the neural network technique, which can allow the parallel computation, to the minimum-time problem. This paper propose an approach for solving the minimum-time path planning by using Hopfield neural network. The effectiveness of the proposed method is demonstrarted using the PUMA 560 manipulator.