• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.3
• /
• pp.135-138
• /
• 2014

This paper devices a facial expression recognition method robust to overfitting using 2D-DCT and EHMM algorithm. In particular, this paper achieves enhanced recognition performance by setting up a large window size for 2D-DCT feature extraction and extracting the observation vectors of EHMM. The experimental results on the CK facial expression database and the JAFFE facial expression database showed that the facial expression recognition accuracy was improved according as window size is large. Also, the proposed method revealed the recognition accuracy of 87.79% and showed enhanced recognition performance ranging from 46.01% to 50.05% in comparison to previous approaches based on histogram feature, when CK database is employed for training and JAFFE database is used to test the recognition accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.3
• /
• pp.139-146
• /
• 2018

Monostatic sonar systems localize targets using the time information of pulse transmission and receipt. Whereas, in asynchronous bistatic sonar systems, receivers need to detect direct blast to localize targets, since a source doesn't share pulse information with receivers. In this paper, we propose a direct blast detection algorithm, which estimates PRI (Pulse Repetition Interval) of direct blast and adaptive thresholds. Experimental results show the proposed algorithm has robust direct blast detection performance in the environment where strong background noise and target signal exist.

• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.213-216
• /
• 2000

본 연구에서는 마이크로폰 어레이를 이용하여 화자의 음성신호로부터 화자의 위치를 추정하는 기존의 대표적인 알고리듬인 CPSP(Cross Power Spectrum Phase)로부터 보다 반향에 강인한 알고리듬인 저주파 위상 복원 알고리듬을 제안한다. CPSP 함수는 상호 상관관계(Cross Correlation)가 정규화 되어있는 형태를 갖는데, CPSP 함수의 최대 값 인덱스로부터 화자의 공간정보인 TDOA(Time Difference Of Arrival)를 추출한다. 그러나 CPSP 함수를 이용한 공간정보 추정 알고리듬은 실내환경에서 심각하게 일어나는 반향신호에 대해서 취약한 단점을 갖고 있다. 본 논문에서 제안하는 저주파 위상복원 알고리듬은 주파수 측면에서 반향신호가 CPSP 함수에 미치는 영향을 분석하여 반향으로 인하여 왜곡된 위상 성분을 복원함으로써 보다 신뢰도 있는 TDOA 추정을 가능하게 한다. 반향신호로 인한 CPSP의 위상은 저주파보다 고주파에서 심하게 왜곡되는데, 각각의 반향신호의 도달 시간을 기하학적 분포를 갖는 확률변수로 모델링하여 이를 수학적으로 증명하였다. 또한 실제 환경에서 채집한 음성신호를 이용한 모의 실험을 통해 개선된 알고리듬의 성능 개선을 확인하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.2
• /
• pp.163-170
• /
• 2015

In this paper, it is shown that the performance of the frequency-domain multi-reference narrowband noise canceller is determined by the narrowband component to the broadband disturbance power ratio in the reference signals. To overcome this problem, a new narrowband ANC is proposed, where the update of the adaptive filter is determined based on SNR of the reference inputs being measured using the magnitude squared coherence (MSC) between the primary and the reference signals. Simulation results show that the proposed ANC has superior performance over the conventional one.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.9 no.8
• /
• pp.1840-1846
• /
• 2005

In this paper, does by purpose DC servo motor speed controller design about PID coefficient tuning techniques that use genetic algerian. DC servo motor is used in application field of a peat many control machine or robot etc. and in this field, selection of controller parameters requires user's expert knowledge. Therefore, general amount of work engineers must continuously iteration tuning in controller parameters by trial and error. With this, when must tuning parameter coefficient about change of dynamic system or disturbance, can improve the efficiency according to following that is more precised and parameter coefficient value that is optimized by using genetic algorithm. In this paper, from dynamic character modeling get in analyze dynamic character of DC motor desist controller drive control possible that is fast response character md improved speed precision using a Genetic Algorithms.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.12 no.4
• /
• pp.324-331
• /
• 2007

This paper studies new estimation method for state of charge (SOC) of the hybrid electric vehicle lithium battery using sliding mode observer. A simple R-C Lithium battery modeling technique is established and the errors caused by simple modeling was compensated by the sliding mode observer. The structure of the sliding mode observer is simple, but it shows robust control property against modeling errors and uncertainties. The performance of the system has been verified by the UUDS test. The test results of the proposed observer system shows robust tracking performance under real driving environments.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.13-19
• /
• 2022

Video Object Segmentation (VOS) task requires an amount of labeled sequence data, which limits the performance of the current VOS methods trained with public datasets. In this paper, we propose two effective data augmentation schemes for VOS. The first augmentation method is to swap the background segment to the background from another image, and the other method is to play the sequence in reverse. The two augmentation schemes for VOS enable the current VOS methods to robustly predict the segmentation labels and improve the performance of VOS.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.5
• /
• pp.477-482
• /
• 2009

The emotion recognition system in commercial environments such as call-center undergoes severe system performance degradation and instability due to the speech characteristic differences between the system training database and the input speech of unspecified customers. In order to alleviate these problems, this paper extends traditional method of emotion recognition of neutral/anger into two-step hierarchical structure by using emotional characteristic changes and differences of male and female. The experimental results indicate that the proposed method provides very stable and successful emotional classification performance about 25% over the traditional method of emotion recognition.

• The Journal of the Acoustical Society of Korea
• /
• v.27 no.5
• /
• pp.239-243
• /
• 2008

This paper proposes a new audio identification method based on a combination of the instantaneous and dynamic spectral features of the audio spectrum. Especially we propose the spectro-temporal subband centroids that are easy to compute and effective to summarize the instantaneous and dynamic spectral variations. Experimental results demonstrate that the identification performance can be greatly improved by combining both the spectral and the temporal subband centroids.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.1
• /
• pp.4-8
• /
• 1998

본 논문에서는 다양한 전화선 채널을 통하여 수집된 음성 데이터에 포함된 잡음 및 채널 왜곡을 제거하여 음성인식 시스템의 성능을 향상시키는 방법에 관하여 연구하였다. 전 화선을 통과한 음성에 포함된 채널 잡음 및 왜곡을 제거하는 방법으로는 음성신호를 보상하 는 방법으로 CMS(Cepstral Mean Subtraction), SBR(Signal Bias Removal)과 SM(Stochastic Matching)의 성능을 비교 평가하였다. 잡음제거 방식의 성능을 평가를 위하 여 음소 단위의 반연속 HMM을 이용한 화자독립 단독음 인식을 수행하였다. 인식 실험 결 과, 멜 켑스트럼을 사용한 경우에 CMS가 가장 우수한 성능을 내었고 다음으로 SM과 SBR 순으로 나타났다. 또한 특징벡터를 주변 잡음에 강인하게 하는 가중함수(RPS, BPL)를 사용 한 켑스트럼 계수와 잡음제거 방식을 함께 사용한 경우에 인식 성능이 더욱 향상되었다.