• 한국전자통신학회논문지
• /
• 제14권6호
• /
• pp.1249-1256
• /
• 2019

본 논문에서는 보청기의 음향궤환 및 잡음을 제거하기 위한 새로운 알고리즘을 제안한다. 이 알고리즘은 기존의 FIR 구조를 이용하는 대신 신경망 적응예측필터를 이용한 심층학습 알고리즘으로 궤환 및 잡음제거 성능을 향상시킨다. 먼저 궤환제거기가 마이크 신호에서 궤환신호를 제거하고, 이어서 Wiener 필터기법을 이용하여 잡음을 제거한다. 잡음 제거는 음성신호가 가진 주기적 성질에 따라 선형예측모델을 이용하여 잡음이 포함된 음성신호로부터 음성을 추정해내는 것이다. 한 루프 안에 포함된 두 적응 시스템의 안정적 수렴을 보장하기 위해 궤환제거기 및 잡음제거기의 계수 업데이트를 분리하여 실시하며 제거 후 생성된 잔차신호를 이용하여 수렴시키는 과정을 진행한다. 본 연구에서 제안한 궤환 및 잡음제거기의 성능을 검증하기 위하여 시뮬레이션 프로그램을 작성하고 모의실험을 수행하였다. 실험 결과, 제안한 심층학습 알고리즘을 사용하면 기존의 FIR 구조를 사용하는 경우보다 궤환제거기에서 약 10 dB의 SFR(: Signal to Feedback Ratio), 잡음제거기에서 약 3 dB의 SNRE(: Signal to Noise Ratio Enhancement) 개선효과를 얻을 수 있는 것으로 확인되었다.

• 한국전문물리치료학회지
• /
• 제11권3호
• /
• pp.11-18
• /
• 2004

This study is aimed at investigating the influence of different quantitative knowledge of results on the measurement error during lumbar proprioceptive sensation training. Twenty-eight healthy adult men participated and subjects were randomly assigned into four different feedback groups(100% relative frequency with an angle feedback, 50% relative frequency with an angle feedback, 100% relative frequency with a length feedback, 50% relative frequency with a length feedback). An electrogoniometer was used to determine performance error in an angle, and the Schober test with measurement tape was used to determine performance error in a length. Each subject was asked to maintain an upright position with both eyes closed and both upper limbs stabilized on their pelvis. Lumbar vertebrae flexion was maintained at $30^{\circ}$ for three seconds. Different verbal knowledge of results was provided in four groups. After lumbar flexion was performed, knowledge of results was offered immediately. The resting period between the sessions per block was five seconds. Training consisted of 6 blocks, 10 sessions per one block, with a resting period of one minute. A resting period of five minutes was provided between 3 blocks and 4 blocks. A retention test was performed between 10 minutes and 24 hours later following the training block without providing knowledge of results. To determine the training effects, a two-way analysis of variance and a one-way analysis of variance were used with SPSS Ver. 10.0. A level of significance was set at .05. A significant block effect was shown for the acquisition phase (p<.05), and a significant feedback effect was shown in the immediate retention phase (p>.05). There was a significant feedback effect in the delayed retention phase (p<.05), and a significant block effect in the first acquisition phase and the last retention phase (p<.05). In conclusion, it is determined that a 50% relative frequency with a length feedback is the most efficient feedback among different feedback types.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1996년도 하계학술대회 논문집 B
• /
• pp.1278-1280
• /
• 1996

In this paper, an Iterative learning control algorithm is presented for a class of non linear system which have external inputs or disturbances. The acceleration of error signal is used to update the next control signal. It is shown that the feedback gain can be deter.ined so that the overall errors are convergent.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제51권1호
• /
• pp.39-42
• /
• 2002

For the purpose of modeling EEG signal which has nonstationary and nonlinear dynamic characteristics, this paper propose a state feedback real time recurrent neural network model. The state feedback real time recurrent neural network is structured to have memory structure in the state of hidden layers so that it has arbitrary dynamics and ability to deal with time-varying input through its own temporal operation. For the model test, Mackey-Glass time series is used as a nonlinear dynamic system and the model is applied to the prediction of three types of EEG, alpha wave, beta wave and epileptic EEG. Experimental results show that the performance of the proposed model is better than that of other neural network models which are compared in this paper in some view points of the converging speed in learning stage and normalized mean square error for the test data set.

• 제어로봇시스템학회논문지
• /
• 제6권11호
• /
• pp.987-994
• /
• 2000

A Linear Pulse Motor (LPM) is a direct drive motor that has good performance in terms of accuracy, velocity and acceleration compared to the conventional rotating system with toothed belts and ball screws. However, since an LPM needs supporting devices which maintain constant air-gap and has strong nonlinearity caused by leakage magnetic flux, friction and cogging, etc., there are many difficulties in improvement on accuracy with conventional control theory. Moreover, when designing the position controller of LPM, the modeling error and load variations has not been considered. In order to compensate these components, the neural network with conventional feedback controller is introduced. This neural network of feedback error learning type changes the current commands to improve position accuracy. As a result of experiments, we observes that more accurate position control is possible compared to conventional controller.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.85.4-85
• /
• 2001

A learning controller is presented for repetitive walking motion of biped robot. The learning control scheme learns the approximate inverse dynamics input of biped walking robot and uses the learned input pattern to generate an input profile of different walking motion from that learnt. In the learning controller, the PID feedback controller takes part in stabilizing the transient response of robot dynamics while the feedforward learning controller plays a role in computing the desired actuator torques for feedforward nonlinear dynamics compensation in steady state. It is shown that all the error signals in the learning control system are bounded and the robot motion trajectory converges to the desired one asymptotically. The proposed learning control scheme is ...

• 전자공학회논문지S
• /
• 제34S권7호
• /
• pp.56-65
• /
• 1997

In this paper, we propose a reference input odification-type iterative learning control law for a class of discrete-time nonlinear systems and prove the convergence of the output error. We can get the high-precision in case of the trajectroy control when the proposed control law is properly combined with a feedback controller, and we can easily implement the learning control law compared to the control input modification-type learning control law. To show the validity and the convergence perfodrmance of the proposed control law, we perform experimentations on the trajectroy control and rejection of periodic disturbance for a 2-axis SCARA-type direct drive robot.

• 전자공학회논문지B
• /
• 제32B권12호
• /
• pp.1652-1662
• /
• 1995

In this paper, a hybrid neural network is proposed to improve the learning ability of a neural network. The union of the characteristics of a Self-Organizing Neural Network model and of multi-layer perceptron model using the backpropagation learning method gives us the advantage of reduction of the learning error and the learning time. In learning process, the proposed hybrid neural network reduces the number of nodes in hidden layers to reduce the calculation time. And this proposed neural network uses the fuzzy feedback values, when it updates the responding region of each node in the hidden layer. To show the effectiveness of this proposed hybrid neural network, the boolean function(XOR, 3Bit Parity) and the solution of inverse kinematics are used. Finally, this proposed hybrid neural network is applied to the visual tracking control of a PUMA560 robot, and the result data is presented.

• 한국어교육
• /
• 제25권1호
• /
• pp.203-234
• /
• 2014

This paper demonstrates how to incorporate the teaching of communication strategies into a large class of English-speaking learners of the Korean language. The method proposed here was developed to overcome the difficulty of conducting language activities involving communicative interactions amongst students and also between teacher and students in a large classroom. As a way of compensating the minimal opportunities for interactions in the classroom, students are given the task of expressing in Korean the English translations of authentic Korean comics via Twitter, which was later replaced with the feedback feature on Moodle, and then their Korean expressions are collected and projected onto a big screen. These collected expressions by students naturally differ from one another, helping students to realize that it is possible for them to express the same message or meaning in many different ways. The results of two separately conducted questionnaires show that this method is an effective way of providing students with significantly increased chances of producing 'comprehensible output' that requires them to think of how to communicate with their limited knowledge of the Korean language. Many students also commented that the teachers' feedback on errors provides them with the opportunity to learn about common errors as well as their own errors.

• 정보처리학회논문지B
• /
• 제11B권1호
• /
• pp.77-82
• /
• 2004

강화학습에서 temporal-credit 할당 문제 즉, 에이전트가 현재 상태에서 어떤 행동을 선택하여 상태전이를 하였을 때 에이전트가 선택한 행동에 대해 어떻게 보상(reward)할 것인가는 강화학습에서 중요한 과제라 할 수 있다. 본 논문에서는 조합최적화(hard combinational optimization) 문제를 해결하기 위한 새로운 메타 휴리스틱(meta heuristic) 방법으로, greedy search뿐만 아니라 긍정적 반응의 탐색을 사용한 모집단에 근거한 접근법으로 Traveling Salesman Problem(TSP)를 풀기 위해 제안된 Ant Colony System(ACS) Algorithms에 Q-학습을 적용한 기존의 Ant-Q 학습방범을 살펴보고 이 학습 기법에 다양화 전략을 통한 상태전이와 TD-오류를 적용한 학습방법인 Ant-TD 강화학습 방법을 제안한다. 제안한 강화학습은 기존의 ACS, Ant-Q학습보다 최적해에 더 빠르게 수렴할 수 있음을 실험을 통해 알 수 있었다.