• 음성과학
• /
• 제11권1호
• /
• pp.7-20
• /
• 2004

In this paper, we propose a two-step noise compensation algorithm in feature extraction for achieving robust speech recognition. The proposed method frees us from requiring a priori information on noisy environments and is simple to implement. First, in frequency domain, the Harmonics-based Spectral Subtraction (HSS) is applied so that it reduces the additive background noise and makes the shape of harmonics in speech spectrum more pronounced. We then apply a judiciously weighted variance Feature Vector Normalization (FVN) to compensate for both the channel distortion and additive noise. The weighted variance FVN compensates for the variance mismatch in both the speech and the non-speech regions respectively. Representative performance evaluation using Aurora 2 database shows that the proposed method yields 27.18% relative improvement in accuracy under a multi-noise training task and 57.94% relative improvement under a clean training task.

• 방송공학회논문지
• /
• 제20권5호
• /
• pp.728-737
• /
• 2015

스티칭 기법은 여러 영상에서 추출한 특징점의 디스크립터를 생성하고, 특징점들간의 정합 과정을 통해 하나의 영상으로 만드는 것이다. 각각의 특징점은 128 차원의 정보를 가지고 있고, 특징점의 개수가 증가 할수록 데이터 처리 시간이 증가하게 된다. 본 논문에서는 비디오 영상을 입력 했을 때 고속 파노라마 생성을 위한 특징점 추출 및 정합 기법을 제안한다. 빠른 속도로 특징점 추출을 위해서 FAST(Features from Accelerated Segment Test) 기법을 사용한다. 특징점 정합과정은 기존의 방법과는 다른 새로운 방법을 제안한다. Mean shift를 통해 특징점이 포함된 영역을 추적하여 벡터(vector)를 구하고 이 벡터를 사용하여 추출한 특징점들을 정합하는데 사용한다. 마지막으로 이상점(outlier)을 제거하기 위해 RANSAC(RANdom Sample Consensus) 기법을 사용한다. 입력된 두 영상의 호모그래피(homography) 변환 행렬을 구하여 하나의 파노라마 영상을 생성한다. 실험을 통해 제안하는 기법이 기존의 기법보다 속도가 향상되는 것을 확인하였다.

• 전자공학회논문지SC
• /
• 제45권6호
• /
• pp.55-59
• /
• 2008

야외 환경에서 무인차량의 자율주행에 있어서 효과적인 기동제어를 위해서는 장애물 탐지나 지형의 기하학적인 형상 정보외에 탐지된 장애물 및 지형 표면에 대한 재질 유형의 인식 및 분류 또한 중요한 요소이다. 영상 기반의 지표면 분류 알고리듬은 입력 영상에 대한 전처리, 특징추출, 분류 및 후처리의 절차로 수행된다. 본 논문에서는 컬러 CCD 카메라로부터 획득된 야외 지형영상에 대해 색상 및 질감 정보를 이용한 지형분류 기법을 제시한다. 전처리 단계에서 색공간 변환을 수행하고, 색상과 질감 정보를 이용하기 위해 웨이블릿 변환 특징을 사용하였으며, 분류기로서는 SVM(support vector machine)을 적용하였다. 야외 환경에서 획득된 실영상에 대한 실험을 통하여 제시된 알고리듬의 분류 성능을 평가하였으며, 제시된 알고리듬에 의한 효과적인 야지 지형분류의 가능성을 확인하였다.

• ETRI Journal
• /
• 제35권1호
• /
• pp.100-108
• /
• 2013

In this paper, a feature extraction (FE) method is proposed that is comparable to the traditional FE methods used in automatic speech recognition systems. Unlike the conventional spectral-based FE methods, the proposed method evaluates the similarities between an embedded speech signal and a set of predefined speech attractor models in the reconstructed phase space (RPS) domain. In the first step, a set of Gaussian mixture models is trained to represent the speech attractors in the RPS. Next, for a new input speech frame, a posterior-probability-based feature vector is evaluated, which represents the similarity between the embedded frame and the learned speech attractors. We conduct experiments for a speech recognition task utilizing a toolkit based on hidden Markov models, over FARSDAT, a well-known Persian speech corpus. Through the proposed FE method, we gain 3.11% absolute phoneme error rate improvement in comparison to the baseline system, which exploits the mel-frequency cepstral coefficient FE method.

• 한국컴퓨터산업학회논문지
• /
• 제4권10호
• /
• pp.631-642
• /
• 2003

본 논문에서는 웹상에서 자신을 대신하는 아바타 제작시 본인의 얼굴과 닮은 얼굴을 생성하기 위해 사진으로부터 개인의 특징정보를 추출하는 방법과 추출된 특징정보에 따라 해당하는 이목구비를 준비된 분류기준에 의해 특정 클래스로 분류해 내는 방법을 제안한다. 특징정보 추출은 눈, 코, 입, 턱선으로 나누어 진행되어졌으며, 각 이목구비의 특징점과 분류기준을 각각 제시하였다. 추출 된 특징정보들은 전문 디자이너에 의해 그려진 이목구비 이미지들과 유사도를 계산하는데 사용되었으며, 여기서 가장 유사한 이미지를 턱선 벡터이미지에 합성하여 아바타 얼굴을 얻어낼 수 있었다.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2005년도 BIOINFO 2005
• /
• pp.237-242
• /
• 2005

Predicting the cellular location of an unknown protein gives a valuable information for inferring the possible function of the protein. For more accurate prediction system, we need a good feature extraction method that transforms the raw sequence data into the numerical feature vector, minimizing information loss. In this paper, we propose new methods of extracting underlying features only from the sequence data by computing pairwise sequence alignment scores. In addition, we use composition based features to improve prediction accuracy. To construct an SVM ensemble from separately trained SVM classifiers, we propose specificity based weighted majority voting. The overall prediction accuracy evaluated by the 5-fold cross-validation reached 88.53% for the eukaryotic animal data set. By comparing the prediction accuracy of various feature extraction methods, we could get the biological insight on the location of targeting information. Our numerical experiments confirm that our new feature extraction methods are very useful for predicting subcellular localization of proteins.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1999년도 추계종합학술대회 논문집
• /
• pp.581-584
• /
• 1999

In order to retrieve the rotated image within database by the content based image retrieval system, the algorithms with rotation robustness is usually applied in the procedure of the feature extraction. In that case, it requires much calculation time for feature extraction and much indexed data for feature indexing. Thus. in this paper. we propose the rotation robust algorithm using the block variance of the projected vector. The algorithm does not require additional calculation for feature extraction and is executed within query time by comparing the extracted data. Proposed method can be processed through database including various size of images with shape information and executed with fast response time in implementation.

• International Journal of Computer Science & Network Security
• /
• 제24권4호
• /
• pp.113-118
• /
• 2024

Pursuance Sentiment Analysis on Twitter is difficult then performance it's used for great review. The present be for the reason to the tweet is extremely small with mostly contain slang, emoticon, and hash tag with other tweet words. A feature extraction stands every technique concerning structure and aspect point beginning particular tweets. The subdivision in a aspect vector is an integer that has a commitment on ascribing a supposition class to a tweet. The cycle of feature extraction is to eradicate the exact quality to get better the accurateness of the classifications models. In this manuscript we proposed Term Frequency-Inverse Document Frequency (TF-IDF) method is to secure Principal Component Analysis (PCA) with Naïve Bayes Classifiers. As the classifications process, the work proposed can produce different aspects from wildly valued feature commencing a Twitter dataset.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1997년도 한국자동제어학술회의논문집; 한국전력공사 서울연수원; 17-18 Oct. 1997
• /
• pp.1436-1439
• /
• 1997

The objective of this paper is to present an efficient and practical approach to the automatic classification of power quality(PQ) disturbances, where and orthogonal polynomial approximation method is emloyed for the detection and localization of PQ disturbances, and a feature vector, newly extracted form the bispectra of the detected signal, is utilized for the automatic rectgnition of the various types of PQ disturbances. To demonstrae the performance and applicabiliyt of the proposed approach, some simulation results are provided.

• 전기전자학회논문지
• /
• 제17권3호
• /
• pp.241-247
• /
• 2013

This paper presents the design of a System on Programmable Chip (SoPC) based on Field Programmable Gate Array (FPGA) for speech recognition in which Mel-Frequency Cepstral Coefficients (MFCC) for speech feature extraction and Vector Quantization for recognition are used. The implementing process of the speech recognition system undergoes the following steps: feature extraction, training codebook, recognition. In the first step of feature extraction, the input voice data will be transformed into spectral components and extracted to get the main features by using MFCC algorithm. In the recognition step, the obtained spectral features from the first step will be processed and compared with the trained components. The Vector Quantization (VQ) is applied in this step. In our experiment, Altera's DE2 board with Cyclone II FPGA is used to implement the recognition system which can recognize 64 words. The execution speed of the blocks in the speech recognition system is surveyed by calculating the number of clock cycles while executing each block. The recognition accuracies are also measured in different parameters of the system. These results in execution speed and recognition accuracy could help the designer to choose the best configurations in speech recognition on SoPC.