• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.9
• /
• pp.133-141
• /
• 1994

Conventional competitive learning algorithms compute the Euclidien distance to determine the winner neuron out of all predetermined output neurons. In such cases, there is a drawback that the performence of the learning algorithm depends on the initial reference(=weight) vectors. In this paper, we propose a new competitive learning algorithm that dynamically generates output neurons. The proposed method generates output neurons by dynamically changing the class thresholds for all output neurons. We compute the similarity between the input vector and the reference vector of each output neuron generated. If the two are similar, the reference vector is adjusted to make it still more like the input vector. Otherwise, the input vector is designated as the reference vector of a new outputneuron. Since the reference vectors of output neurons are dynamically assigned according to input pattern distribution, the proposed method gets around the phenomenon that learning is early determined due to redundant output neurons. Experiments using speech data have shown the proposed method to be superior to existint methods.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.44 no.4
• /
• pp.453-459
• /
• 1995

In this paper, a novel dead time compensation method is presented which produces inverter output voltages equal to reference voltages. An experimental result is also presented to demonstrate the validity of the proposed method. It shows that the compensation of the dead time is possible up to a sub-microsecond range. And the reference voltage can be used as a feedback value, which is essential for sensorless vector control and flux estimation. The method is based on space vector PWM strategy and can be carried out automatically by an inverter controller for initial set-up without any extra hardware.

• Journal of Power Electronics
• /
• v.5 no.1
• /
• pp.11-19
• /
• 2005

This paper describes the two control techniques to perform the sensorless vector control of a PMSM by injecting the high frequency voltage to the stator terminal. The first technique is the estimation algorithm of the initial rotor position. A PMSM possesses the saliency which produces the ellipse of the stator current when the high frequency voltage is injected into the motor terminal. The major axis angle of the current ellipse gives the rotor position information at a standstill. The second control technique is a sensorless control algorithm that injects the high frequency voltage to the stator terminal in order to estimate the rotor position and speed. The rotor position and speed for sensorless vector control is calculated by appropriate signal processing to extract the position information from the stator current at low speeds or standstill. The proposed sensorless algorithm using the double-band hysteresis controller exhibits excellent reference tracking and increased robustness. Experimental results are presented to verify the feasibility of the proposed control schemes. Speed, position estimation and vector control were carried out on the floating point processor TMS320VC33.

• Journal of radiological science and technology
• /
• v.30 no.4
• /
• pp.419-426
• /
• 2007

In this paper, the input pattern vectors are extracted and the learning algorithms is designed to recognize taste(bitter, sweet, sour and salty) pattern vectors. The signal intensity of taste are used to compose the input pattern vectors. The SOM(Self Organizing Maps) algorithm for taste pattern recognition is used to learn initial reference vectors and the ot-star learning algorithm is used to determine the class of the output neurons of the sunclass layer. The weights of the proposed algorithm which is between the input layer and the subclass layer can be learned to determine initial reference vectors by using SOM algorithm and to learn reference vectors by using LVQ(Learning Vector Quantization) algorithm. The pattern vectors are classified into subclasses by neurons in the subclass layer, and the weights between subclass layer and output layer are learned to classify the classified subclass, which is enclosed a class. To classify the pattern vectors, the proposed algorithm is simulated with ones of the conventional LVQ, and it is confirmed that the proposed learning method is more successful classification than the conventional LVQ.

• Proceedings of the KIEE Conference
• /
• 1994.07a
• /
• pp.262-264
• /
• 1994

In is paper, a novel dead time compensation method is presented which produces inverter output voltages equal to reference voltages. An experimental result is also presented to demonstrate the validity of the proposed method. The reference voltage can be used as a feedback value, which is essential for sensorless vector control and flux estimation. The method can be carried out automatically by an inverter controller for initial set-up without any extra hardware.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.2A
• /
• pp.113-118
• /
• 2012

In this paper, we propose the removal of search point using motion vector correlation and distance between reference frames in H.264/AVC. We remove the search points in full search method and predictive motion vectors in enhanced predictive zonal search method. Since the probability that the reference frame far from the current frame is selected as the best reference frame is decreased, we apply the weighted average based on distance between the current and reference frame to determine the fianl search range. In general, the size of search range is smaller than initial search range. We reduce motion estimation time using the final search range in full search method. Also, the refinement process is adaptively applied to each reference frame. The proposed methods reduce the computational throughput of full search method by 57.13% and of enhanced predictive zonal search by 14.71% without visible performance degradation.

• Smart Media Journal
• /
• v.6 no.2
• /
• pp.9-14
• /
• 2017

A packet loss in wireless environments causes a severe degradation of video quality in video communications. In this paper, a novel video error concealment algorithm is presented by combining boundary errors and a block reliability measure. The block reliability measure decides the reliability of a block by checking residual errors of a block. In the proposed approach, a motion vector of a missing unreliable block in an inter coded frame is obtained initially based on the motion vector of the same block in the reference frame. Furthermore, if the block in the reference frame is unreliable according to the reliability measure, a new motion vector is decided based on block boundary errors around the initial motion vector. According to our simulations, the proposed approach shows promising results for error concealment in error-prone wireless environments.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.40 no.5
• /
• pp.280-288
• /
• 2003

In this paper, we propose a design method of the LVQ network using the SOM. The proposed method determines subclasses and initial reference vectors of the LVQ network using the SOM. The efficacy of the proposed method is verified by means of simulations on iris data of Fisher and character recognition. The results show that the proposed method improves considerably on the performance of the conventional LVQ network.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.5 s.43
• /
• pp.95-103
• /
• 2006

In this paper, the input pattern vectors are extracted and the learning algorithms is designed to recognize EEG pattern vectors. The frequency and amplitude of alpha rhythms and beta rhythms are used to compose the input pattern vectors. And the algorithm for EEG pattern recognition is used SOM to learn initial reference vectors and out-star learning algorithm to determine the class of the output neurons of the subclass layer. The weights of the proposed algorithm which is between the input layer and the subclass layer can be learned to determine initial reference vectors by using SOM algorithm and to learn reference vectors by using LVQ algorithm, and pattern vectors is classified into subclasses by neurons which is being in the subclass layer, and the weights between subclass layer and output layer is learned to classify the classified subclass, which is enclosed a class. To classify the pattern vectors of EEG, the proposed algorithm is simulated with ones of the conventional LVQ, and it was a confirmation that the proposed learning method is more successful classification than the conventional LVQ.

• Proceedings of the KIEE Conference
• /
• 2001.07b
• /
• pp.1172-1175
• /
• 2001

Recently, vector control without a speed sensor for an inverter induction motor system has been developed. In this paper describes the vector control system estimates rotor speed based on FMRAS(Flux Model Reference Adaptive System). Because of improving the initial condition and drift problem by pure integrator is eliminated, we can be expected to rapid responsibility of the speed estimation. The stability of speed estimator is proved on the basis of hyperstability theory. Proposed control system used TMS320C31 DSP for high speed processing. The effectiveness of the proposed system is verified by simulation and experimental results. This result shows that the highly characteristic speed estimation and robust character of load regulation.