• 말소리와 음성과학
• /
• 제7권4호
• /
• pp.9-16
• /
• 2015

In this paper, several experiments are performed according to deep neural network (DNN) based features for the performance comparison of speaker verification (SV) systems. To this end, input features for a DNN, such as mel-frequency cepstral coefficient (MFCC), linear-frequency cepstral coefficient (LFCC), and perceptual linear prediction (PLP), are first compared in a view of the SV performance. After that, the effect of a DNN training method and a structure of hidden layers of DNNs on the SV performance is investigated depending on the type of features. The performance of an SV system is then evaluated on the basis of I-vector or probabilistic linear discriminant analysis (PLDA) scoring method. It is shown from SV experiments that a tandem feature of DNN bottleneck feature and MFCC feature gives the best performance when DNNs are configured using a rectangular type of hidden layers and trained with a supervised training method.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2006년도 Proceedings of ISRS 2006 PORSEC Volume II
• /
• pp.643-646
• /
• 2006

This paper addresses the approach to extract linear features from satellite imagery using an efficient segmentation method. The extraction of linear features from satellite images has been the main concern of many scientists. There is a need to develop a more capable and cost effective method for the Iranian map revision tasks. The conventional approaches for producing, maintaining, and updating GIS map are time consuming and costly process. Hence, this research is intended to investigate how to obtain linear features from SPOT satellite imagery. This was accomplished using a discontinuity-based segmentation technique that encompasses four stages: low level bottom-up, middle level bottom-up, edge thinning and accuracy assessment. The first step is geometric correction and noise removal using suitable operator. The second step includes choosing the appropriate edge detection method, finding its proper threshold and designing the built-up image. The next step is implementing edge thinning method using mathematical morphology technique. Lastly, the geometric accuracy assessment task for feature extraction as well as an assessment for the built-up result has been carried out. Overall, this approach has been applied successfully for linear feature extraction from SPOT image.

• ETRI Journal
• /
• 제24권6호
• /
• pp.469-472
• /
• 2002

Speaker adaptation techniques are generally used to reduce speaker differences in speech recognition. In this work, we focus on the features fitted to a linear regression-based speaker adaptation. These are obtained by feature transformation based on independent component analysis (ICA), and the feature transformation matrices are estimated from the training data and adaptation data. Since the adaptation data is not sufficient to reliably estimate the ICA-based feature transformation matrix, it is necessary to adjust the ICA-based feature transformation matrix estimated from a new speaker utterance. To cope with this problem, we propose a smoothing method through a linear interpolation between the speaker-independent (SI) feature transformation matrix and the speaker-dependent (SD) feature transformation matrix. From our experiments, we observed that the proposed method is more effective in the mismatched case. In the mismatched case, the adaptation performance is improved because the smoothed feature transformation matrix makes speaker adaptation using noisy speech more robust.

• 제어로봇시스템학회논문지
• /
• 제12권9호
• /
• pp.935-944
• /
• 2006

EMG pattern recognition is essential for the control of a multifunction myoelectric hand. The main goal of this study is to develop an efficient feature projection method for EMC pattern recognition. To this end, we propose a linear supervised feature projection that utilizes linear discriminant analysis (LDA). We first perform wavelet packet transform (WPT) to extract the feature vector from four channel EMC signals. For dimensionality reduction and clustering of the WPT features, the LDA incorporates class information into the learning procedure, and finds a linear matrix to maximize the class separability for the projected features. Finally, the multilayer perceptron classifies the LDA-reduced features into nine hand motions. To evaluate the performance of LDA for the WPT features, we compare LDA with three other feature projection methods. From a visualization and quantitative comparison, we show that LDA has better performance for the class separability, and the LDA-projected features improve the classification accuracy with a short processing time. We implemented a real-time pattern recognition system for a multifunction myoelectric hand. In experiment, we show that the proposed method achieves 97.2% recognition accuracy, and that all processes, including the generation of control commands for myoelectric hand, are completed within 97 msec. These results confirm that our method is applicable to real-time EMG pattern recognition far myoelectric hand control.

• 대한원격탐사학회지
• /
• 제5권2호
• /
• pp.133-145
• /
• 1989

This paper describes an implementation of linear feature extraction algorithms for satellite images and a method of automatic GCP(Ground Control Point) filing using the extracted linear feature. We propose a new linear feature extraction algorithm which uses magnitude and direction information of edges. The result of applying the proposed algorithm to satellite images are presented and compared with those of the other algorithms. By using the proposed algorithm, automatic GCP filing was successfully performed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권3호
• /
• pp.1243-1263
• /
• 2018

The two-stage linear discrimination analysis (TSLDA) is a feature extraction technique to solve the small size sample problem in the field of image recognition. The TSLDA has retained all subspace information of the between-class scatter and within-class scatter. However, the feature information in the four subspaces may not be entirely beneficial for classification, and the regularization procedure for eliminating singular metrics in TSLDA has higher time complexity. In order to address these drawbacks, this paper proposes an improved two-stage linear discriminant analysis (Improved TSLDA). The Improved TSLDA proposes a selection and compression method to extract superior feature information from the four subspaces to constitute optimal projection space, where it defines a single Fisher criterion to measure the importance of single feature vector. Meanwhile, Improved TSLDA also applies an approximation matrix method to eliminate the singular matrices and reduce its time complexity. This paper presents comparative experiments on five face databases and one handwritten digit database to validate the effectiveness of the Improved TSLDA.

• 말소리와 음성과학
• /
• 제1권3호
• /
• pp.19-28
• /
• 2009

This paper reports experimental findings about the duration feature of the acoustic components of English stops in Korean speakers' voicing perception. In our experiment, 35 participants discriminated between recorded stimuli and digitally transformed stimuli with different duration features from the original stimuli. 72 sets of paired stimuli are generated to test the effects of the duration feature in various phonetic contexts. The result of our experiment is a complicated cross-tabulation with 540 cells defined by five categorical independent variables plus one response variable. To find a meaningful generalization out of this complex frequency table, we ran logit log-linear regression analyses. Surprisingly, we have found that there is no single effect of the duration feature in all phonetic contexts on Korean speakers' perception of the voicing contrasts of English stops. Instead, the logit log-linear analyses reveal that there are interaction effects among phonetic contexts (=C), the places of articulation of stops (=P), and the voicing contrast (=V), and among duration (=T), phonetic contexts, and the places of articulation. To put it in mathematical terms, the distribution of the data can be explained by a simple log-linear equation, logF=${\mu}+{\lambda}CPV+{\lambda}TCP$.

• 말소리와 음성과학
• /
• 제1권4호
• /
• pp.127-132
• /
• 2009

The feature vectors which are used in conventional speaker recognition (SR) systems may have many correlations between their neighbors. To improve the performance of the SR, many researchers adopted linear transformation method like principal component analysis (PCA). In general, the linear transformation of the feature vectors is based on concatenated form of the static features and their dynamic features. However, the linear transformation which based on both the static features and their dynamic features is more complex than that based on the static features alone due to the high order of the features. To overcome these problems, we propose an efficient method that applies linear transformation and temporal information of the features to reduce complexity and improve the performance in speaker verification (SV). The proposed method first performs a linear transformation by PCA coefficients. The delta parameters for temporal information are then obtained from the transformed features. The proposed method only requires 1/4 in the size of the covariance matrix compared with adding the static and their dynamic features for PCA coefficients. Also, the delta parameters are extracted from the linearly transformed features after the reduction of dimension in the static features. Compared with the PCA and conventional methods in terms of equal error rate (EER) in SV, the proposed method shows better performance while requiring less storage space and complexity.

• 제어로봇시스템학회논문지
• /
• 제22권10호
• /
• pp.833-838
• /
• 2016

This paper presents a performance evaluation of various feature-initialization algorithms for forward-viewing mono-camera based simultaneous localization and mapping (SLAM), specifically in indoor environments. For mono-camera based SLAM, the position of feature points cannot be known from a single view; therefore, it should be estimated from a feature initialization method using multiple viewpoint measurements. The accuracy of the feature initialization method directly affects the accuracy of the SLAM system. In this study, four different feature initialization algorithms are evaluated in simulations, including linear triangulation; depth parameterized, linear triangulation; weighted nearest point triangulation; and particle filter based depth estimation algorithms. In the simulation, the virtual feature positions are estimated when the virtual robot, containing a virtual forward-viewing mono-camera, moves forward. The results show that the linear triangulation method provides the best results in terms of feature-position estimation accuracy and computational speed.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1996년도 하계학술대회 논문집 B
• /
• pp.1353-1355
• /
• 1996

This paper presents classification of high impedance fault pattern using linear prediction coefficients. A feature of neutral phase current is extracted by the linear predictive coding. This feature is classified into faults by a multilayer perceptron neural network. Neural network successfully classifies test data into three faults and one normal state.