• 대한음성학회지:말소리
• /
• 제53호
• /
• pp.103-118
• /
• 2005

The performance of speech recognition system is degraded by mismatch between training and test environments. Recently, Stereo-based Piecewise LInear Compensation for Environments (SPLICE) was introduced to overcome environmental mismatch using stereo data. In this paper, we propose several methods to improve the conventional SPLICE and evaluate them in the Aurora2 task. We generalize SPLICE to compensate for covariance matrix as well as mean vector in the feature space, and thereby yielding the error rate reduction of 48.93%. We also employ the weighted sum of correction vectors using posterior probabilities of all Gaussians, and the error rate reduction of 48.62% is achieved. With the combination of the above two methods, the error rate is reduced by 49.61% from the Aurora2 baseline system.

• 한국멀티미디어학회논문지
• /
• 제20권7호
• /
• pp.1046-1053
• /
• 2017

We propose a design method of high accuracy power measuring device and the method of securing the reliability of the measured data. As a design method for the development of the high accuracy power measuring device, the circuit was made by reflecting the high accuracy power measuring ICs and the CT (Current Transformer). To ensure the reliability of the power meter, we requested to the watt-hour meter certified testing organization to measure the error rate, and the error rate of active power based on the "Wattmeter Technical Standard(Revision notify 2014-283)" was measure, and the error rate was confirmed to be less than 0.1%.

• 한국음향학회지
• /
• 제30권2호
• /
• pp.86-91
• /
• 2011

In the underwater communications, the acoustic channel is in poor communication conditions, such as long propagation delay, narrow bandwidth, and high bit-error rate. For these bad acoustic channels, we propose an efficient automatic repeat request (ARQ) for multi-hop underwater network by using the concepts of concurrent bi-directional transmission, multiple sub-packets, and overhearing data packet instead of the acknowledgement signal. Our results show that the proposed ARQ significantly reduces the transmission latency especially in high BER compared with the existing Stop and Wait ARQ.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1999년도 하계학술대회 논문집 G
• /
• pp.2932-2934
• /
• 1999

In this paper, a scheme for recognition of handwritten digits using a multilayer neural network trained with the back-propagation algorithm using generalized delta rule is proposed. The neural network is trained with hand written digit data of different writers and different styles. One of the purpose of the work with neural networks is the minimization of the mean square error(MSE) between actual output and desired one. The back-propagation algorithm is an efficient and very classical method. The back-propagation algorithm for training the weights in a multilayer net uses the steepest descent minimization procedure and the sigmoid threshold function. As an error rate is reduced, recognition rate is improved. Therefore we propose a method that is reduced an error rate.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제4권1호
• /
• pp.62-77
• /
• 2010

Emerging applications with high data rates will need to transport bulk data reliably in wireless sensor networks. ARQ (Automatic Repeat request) or Forward Error Correction (FEC) code schemes can be used to provide reliable transmission in a sensor network. However, the naive ARQ approach drops the whole frame, even though there is a bit error in the frame and the FEC at the bit level scheme may require a highly complex method to adjust the amount of FEC redundancy. We propose a bulk data transmission scheme based on erasure-resilient code in this paper to overcome these inefficiencies. The sender fragments bulk data into many small blocks, encodes the blocks with LT codes and packages several such blocks into a frame. The receiver only drops the corrupted blocks (compared to the entire frame) and the original data can be reconstructed if sufficient error-free blocks are received. An incidental benefit is that the frame error rate (FER) becomes irrelevant to frame size (error recovery). A frame can therefore be sufficiently large to provide high utilization of the wireless channel bandwidth without sacrificing the effectiveness of error recovery. The scheme has been implemented as a new data link layer in TinyOS, and evaluated through experiments in a testbed of Zigbex motes. Results show single hop transmission throughput can be improved by at least 20% under typical wireless channel conditions. It also reduces the transmission time of a reasonable range of size files by more than 30%, compared to a frame ARQ scheme. The total number of bytes sent by all nodes in the multi-hop communication is reduced by more than 60% compared to the frame ARQ scheme.

• 전기학회논문지
• /
• 제60권2호
• /
• pp.455-458
• /
• 2011

4:10 deserializer is proposed to recover 1:10 serial data using 1/4-rate clock. And then, 1/4-rate CDR(Clock and Data Recovery) circuit was designed for SERDES of high-speed serial display interface. The reduction of clock frequency using 1/4-rate clocking helps relax the speed limitation when higher data transfer is demanded. This circuit is composed of 1/4-rate sampler, PEL(Phase Error Logic), Majority Voting, Digital Filter, DPC(Digital to Phase Converter) and 4:10 deserializer. The designed CDR has been designed in a standard $0.18{\mu}m$ 1P6M CMOS technology and the recovered data jitter is 14ps in simulation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권12호
• /
• pp.6017-6037
• /
• 2017

In a multi-resolution image encoding system, the image is encoded into a single file as a layer of bit streams, and then it is transmitted layer by layer progressively to reduce the transmission time across a low bandwidth connection. This encoding scheme is also suitable for multiple decoders, each with different capabilities ranging from a handheld device to a PC. In our previous work, we proposed an edge adaptive hierarchical interpolation algorithm for multi-resolution image coding system. In this paper, we enhanced its compression efficiency by adding three major components. First, its prediction accuracy is improved using context adaptive error modeling as a feedback. Second, the conditional probability of prediction errors is sharpened by removing the sign redundancy among local prediction errors by applying sign flipping. Third, the conditional probability is sharpened further by reducing the number of distinct error symbols using error remapping function. Experimental results on benchmark data sets reveal that the enhanced algorithm achieves a better compression bit rate than our previous algorithm and other algorithms. It is shown that compression bit rate is much better for images that are rich in directional edges and textures. The enhanced algorithm also shows better rate-distortion performance and visual quality at the intermediate stages of progressive image transmission.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 1998년도 춘계 학술대회논문집
• /
• pp.1-14
• /
• 1998

Underwater explosion properties for TNT, an ideal explosive, and DXD-04, a nonideal explosive, were numerically modeled with a one-dimensional Lagrangian hydrodynamic code. The equation of state parameters for detonation products for TNT and DXD-04 were obtained from the BKW code, assuming complete reaction. Burn of TNT was modeled by using the Chapman-Jouguet(CJ) volume burn technique, a programmed-burn technique, assuming instantaneous detonation reaction. Burn of DXD-04 was modeled by using the same technique and by using the reaction rate calibrated from two-dimensional steady-state detonation experiments. The calculations for TNT reproduced the experimental peak pressure of the shock wave propagating through water with an error of $3.0\%$ and the experimental oscillation period of the bubble formed of detonation products with an error of $2.3\%$. For DXD-04, the CJ volume burn technique could not reproduce the experimental observations. When the reaction rate calibrated from two-dimensional steady-state detonation experimental data, the calculated peak pressure was slightly higher by $7.3\%$ than the experimental data, but the calculated shock profile was in good agreement. The bubble period was reproduced with an error of $1.8\%$. These results demonstrated that underwater explosion properties for an ideal explosive can be predicted by using a programmed burn technique, and that, however, those for a nonideal explosive can be predicted only when a well-calibrated reaction rate is used.

• 방송공학회논문지
• /
• 제10권4호통권29호
• /
• pp.540-547
• /
• 2005

ATSC DTV 수신기의 결정 궤환 등화기는 양자화기나 지연시간이 없는 트렐리스 복호기에서 에러가 발생하면, 발생된 에러는 후방 필터의 입력으로 사용되어 에러 전파 현상을 야기한다. 에러 전파 현상은 잔류 에러의 증가와 수렴 속도 저하라는 등화기의 성능 열화를 가져온다. 따라서 본 논문에서는 에러 전파 현상을 해결하기 위하여 이중 후방 필터 구조의 결정 궤환 등화기를 제안한다. 제안된 등화기의 후방 필터는 지연 시간이 없는 트렐리스 복호기의 출력을 입력으로 사용하는 결정 후방 필터와 등화기의 출력을 입력으로 사용하는 비결정 후방 필터로 구성되어 있다. 부가적으로 사용된 비결정 후방 필터는 에러 전파 현상이 발생하지 않고, 또한 결정 후방 필터에서 발생되는 에러 전파 현상을 보상한다. 따라서 제안된 등화기는 기존의 등화기에 비해서 수렴 속도는 빨라지고 잔류에러는 감소한다. 전산 모의 실험을 통하여 제안된 이중 후방필터 구조의 등화기의 성능 향상을 분석하였다.

• Journal of the Korean Data and Information Science Society
• /
• 제22권4호
• /
• pp.619-630
• /
• 2011

본 연구에서는 두 분포함수의 혼합된 자료에서 적절한 분류점을 추정하고 평가하기 위하여 많이 사용하는 아홉 종류의 분류정확도 측도인 MVD, Youden지수, (0,1)까지최단기준, 수정된 (0,1)까지 최단기준, SSS, 대칭점, 정확도면적, TA, TR을 다섯 개의 조건범주로 군집시킨다. 신용평가분석에서 정상과 부도상태의 스코어 확률변수가 정규분포를 따르며 전체부도율로 혼합되었다고 가정한다. 다양한 정규혼합분포의 상황에서 군집된 측도들의 최적분류점을 발견하고, 그 분류점에 대응하는 제I종 오류율과 제II종 오류율 그리고 두 종류의 오류율 합을 구하여 각각의 오류율이 최소인 경우를 탐색적으로 살펴본다. 현실자료에 적합한 정규혼합분포를 추정하여 본 연구 결과를 적용하면 최소 오류율이 보장되는 분류정확도를 선택할 수 있으며, 이를 사용하여 모형의 판별력을 향상시킬 수 있다.