Relative Error Reduction Research Articles

Identifying Language Origin of Named Entity With Multiple Information Sources

To identify the language origin of a named entity, morphological information associated with its letter spelling, such as letter N-grams, is commonly employed. However, with this information only, named entities with similar spellings but from different language origins are difficult to differentiate. In this paper, a measure of popularity, in terms of frequency or page count of the named entity in language-specific Web search, is proposed for identifying its language origin. Morphological information, including letter or letter-chunk N-grams, is used to enhance the performance of language identification in conjunction with Web-based page counts. Six languages, including English, German, French, Portuguese, Chinese, and Japanese (Chinese and Japanese named entities are shown in their corresponding phonetic alphabets, i.e., Pinyin and Romaji), are tested. Experiments show that when classifying four Latin languages, including English, German, French, and Portuguese, which are written in Latin alphabets, features from different information sources yield substantial performance improvements in the classification accuracy over a letter 4-gram-based baseline system. The accuracy increases from 75.0% to 86.3%, or a 45.2% relative error reduction.

Tone Recognition of Continuous Mandarin Speech Based on Tone Nucleus Model and Neural Network

A method was developed for automatic recognition of syllable tone types in continuous speech of Mandarin by integrating two techniques, tone nucleus modeling and neural network classifier. The tone nucleus modeling considers a syllable F0 contour as consisting of three parts: onset course, tone nucleus, and offset course. Two courses are transitions from/to neighboring syllable F0 contours, while the tone nucleus is intrinsic part of the F0 contour. By viewing only the tone nucleus, acoustic features less affected by neighboring syllables are obtained. When using the tone nucleus modeling, automatic detection of tone nucleus comes crucial. An improvement was added to the original detection method. Distinctive acoustic features for tone types are not limited to F0 contours. Other prosodic features, such as waveform power and syllable duration, are also useful for tone recognition. Their heterogeneous features are rather difficult to be handled simultaneously in hidden Markov models (HMM), but are easy in neural networks. We adopted multi-layer perception (MLP) as a neural network. Tone recognition experiments were conducted for speaker dependent and independent cases. In order to show the effect of integration, experiments were conducted also for two baselines: HMM classifier with tone nucleus modeling, and MLP classifier viewing entire syllable instead of tone nucleus. The integrated method showed 87.1% of tone recognition rate in speaker dependent case, and 80.9% in speaker independent case, which was about 10% relative error reduction as compared to the baselines.

Automatic Language Identification with Discriminative Language Characterization Based on SVM

Robust automatic language identification (LID) is the task of identifying the language from a short utterance spoken by an unknown speaker. The mainstream approaches include parallel phone recognition language modeling (PPRLM), support vector machine (SVM) and the general Gaussian mixture models (GMMs). These systems map the cepstral features of spoken utterances into high level scores by classifiers. In this paper, in order to increase the dimension of the score vector and alleviate the inter-speaker variability within the same language, multiple data groups based on supervised speaker clustering are employed to generate the discriminative language characterization score vectors (DLCSV). The back-end SVM classifiers are used to model the probability distribution of each target language in the DLCSV space. Finally, the output scores of back-end classifiers are calibrated by a pair-wise posterior probability estimation (PPPE) algorithm. The proposed language identification frameworks are evaluated on 2003 NIST Language Recognition Evaluation (LRE) databases and the experiments show that the system described in this paper produces comparable results to the existing systems. Especially, the SVM framework achieves an equal error rate (EER) of 4.0% in the 30-second task and outperforms the state-of-art systems by more than 30% relative error reduction. Besides, the performances of proposed PPRLM and GMMs algorithms achieve an EER of 5.1% and 5.0% respectively.

Constructing Modulation Frequency Domain-Based Features for Robust Speech Recognition

Data-driven temporal filtering approaches based on a specific optimization technique have been shown to be capable of enhancing the discrimination and robustness of speech features in speech recognition. The filters in these approaches are often obtained with the statistics of the features in the temporal domain. In this paper, we derive new data-driven temporal filters that employ the statistics of the modulation spectra of the speech features. Three new temporal filtering approaches are proposed and based on constrained versions of linear discriminant analysis (LDA), principal component analysis (PCA), and minimum class distance (MCD), respectively. It is shown that these proposed temporal filters can effectively improve the speech recognition accuracy in various noise-corrupted environments. In experiments conducted on Test Set A of the Aurora-2 noisy digits database, these new temporal filters, together with cepstral mean and variance normalization (CMVN), provide average relative error reduction rates of over 40% and 27% when compared with baseline Mel frequency cepstral coefficient (MFCC) processing and CMVN alone, respectively.

Robust Speech Recognition by Combining Short-Term and Long-Term Spectrum Based Position-Dependent CMN with Conventional CMN

In a distant-talking environment, the length of channel impulse response is longer than the short-term spectral analysis window. Conventional short-term spectrum based Cepstral Mean Normalization (CMN) is therefore, not effective under these conditions. In this paper, we propose a robust speech recognition method by combining a short-term spectrum based CMN with a long-term one. We assume that a static speech segment (such as a vowel, for example) affected by reverberation, can be modeled by a long-term cepstral analysis. Thus, the effect of long reverberation on a static speech segment may be compensated by the long-term spectrum based CMN. The cepstral distance of neighboring frames is used to discriminate the static speech segment (long-term spectrum) and the non-static speech segment (short-term spectrum). The cepstra of the static and non-static speech segments are normalized by the corresponding cepstral means. In a previous study, we proposed an environmentally robust speech recognition method based on Position-Dependent CMN (PDCMN) to compensate for channel distortion depending on speaker position, and which is more efficient than conventional CMN. In this paper, the concept of combining short-term and long-term spectrum based CMN is extended to PDCMN. We call this Variable Term spectrum based PDCMN (VT-PDCMN). Since PDCMN/VT-PDCMN cannot normalize speaker variations because a position-dependent cepstral mean contains the average speaker characteristics over all speakers, we also combine PDCMN/VT-PDCMN with conventional CMN in this study. We conducted the experiments based on our proposed method using limited vocabulary (100 words) distant-talking isolated word recognition in a real environment. The proposed method achieved a relative error reduction rate of 60.9% over the conventional short-term spectrum based CMN and 30.6% over the short-term spectrum based PDCMN.

은닉 마르코프 모델의 다목적함수 최적화를 통한 자동 독순의 성능 향상

This paper proposes a new multiobjective optimization method for discriminative training of hidden Markov models (HMMs) used as the recognizer for automatic lipreading. While the conventional Baum-Welch algorithm for training HMMs aims at maximizing the probability of the data of a class from the corresponding HMM, we define a new training criterion composed of two minimization objectives and develop a global optimization method of the criterion based on simulated annealing. The result of a speaker-dependent recognition experiment shows that the proposed method improves performance by the relative error reduction rate of about 8% in comparison to the Baum-Welch algorithm.

A graphical unitary group approach-based hybrid density functional theory multireference configuration interaction method

A hybrid density functional theory multireference configuration interaction hybrid model (DFT/MRCI) method for computing electronic excitation energies in heavy-element atomic and molecular systems has been developed within a graphical unitary group approach (GUGA) based configuration interaction framework. Implemented in the COLUMBUS suite of programs, the method advances several new capabilities for evaluating systems involving several electronic spin states, relativistic effects and spin–orbit coupling. The method has been evaluated on several types of systems including carbon monoxide, bromine, bromine fluoride, and uranium +4, uranium +5 and uranyl 2+ ions. The method delivers a high level of experimental consistency with absolute accuracies ranging from 11% to 22%. A reduction in relative error ranging from 11% to 42% is observed for the new method when compared with time dependent density functional theory (TDDFT). The new approach provides a significant improvement in computational effort for comparable MRCI calculations with an observed reduction of expansion size by factors up to 64.

Gender Classification Based on Integration of Multiple Classifiers Using Various Features of Facial and Neck Images

To reduce the rate of error in gender classification,we propose the use of an integration framework that uses conventional facial images along with neck images.First,images are separated into facial and neck regions,and features are extracted from monochrome,color,and edge images of both regions.Second,we use Support Vector Machines(SVMs) to classify the gender of each individual feature.Finally,we reclassify the gender by considering the six types of distances from the optimal separating hyperplane as a 6-dimensional vector.Experimental results show a 28.4% relative reduction in error over the performance baseline of the monochrome facial image approach,which until now had been considered to have the most accurate performance.

Driver Modeling Based on Driving Behavior and Its Evaluation in Driver Identification

All drivers have habits behind the wheel. Different drivers vary in how they hit the gas and brake pedals, how they turn the steering wheel, and how much following distance they keep to follow a vehicle safely and comfortably. In this paper, we model such driving behaviors as car-following and pedal operation patterns. The relationship between following distance and velocity mapped into a two-dimensional space is modeled for each driver with an optimal velocity model approximated by a nonlinear function or with a statistical method of a Gaussian mixture model (GMM). Pedal operation patterns are also modeled with GMMs that represent the distributions of raw pedal operation signals or spectral features extracted through spectral analysis of the raw pedal operation signals. The driver models are evaluated in driver identification experiments using driving signals collected in a driving simulator and in a real vehicle. Experimental results show that the driver model based on the spectral features of pedal operation signals efficiently models driver individual differences and achieves an identification rate of 76.8% for a field test with 276 drivers, resulting in a relative error reduction of 55% over driver models that use raw pedal operation signals without spectral analysis

Gender Classification Based on Integration of Multiple Classifiers Using Various Features of Facial and Neck Images

To reduce the rate of error in gender classification,we propose the use of an integration framework that uses conventional facial images along with neck images.First,images are separated into facial and neck regions,and features are extracted from monochrome,color,and edge images of both regions.Second,we use Support Vector Machines(SVMs) to classify the gender of each individual feature.Finally,we reclassify the gender by considering the six types of distances from the optimal separating hyperplane as a 6-dimensional vector.Experimental results show a 28.4% relative reduction in error over the performance baseline of the monochrome facial image approach,which until now had been considered to have the most accurate performance.

Robust Distant Speech Recognition by Combining Multiple Microphone-Array Processing with Position-Dependent CMN

We propose robust distant speech recognition by combining multiple microphone-array processing with position-dependent cepstral mean normalization (CMN). In the recognition stage, the system estimates the speaker position and adopts compensation parameters estimated a priori corresponding to the estimated position. Then the system applies CMN to the speech (i.e., position-dependent CMN) and performs speech recognition for each channel. The features obtained from the multiple channels are integrated with the following two types of processings. The first method is to use the maximum vote or the maximum summation likelihood of recognition results from multiple channels to obtain the final result, which is called multiple-decoder processing. The second method is to calculate the output probability of each input at frame level, and a single decoder using these output probabilities is used to perform speech recognition. This is called single-decoder processing, resulting in lower computational cost. We combine the delay-and-sum beamforming with multiple-decoder processing or single-decoder processing, which is termed multiple microphone-array processing. We conducted the experiments of our proposed method using a limited vocabulary (100 words) distant isolated word recognition in a real environment. The proposed multiple microphone-array processing using multiple decoders with position-dependent CMN achieved a 3.2% improvement (50% relative error reduction rate) over the delay-and-sum beamforming with conventional CMN (i.e., the conventional method). The multiple microphone-array processing using a single decoder needs about one-third the computational time of that using multiple decoders without degrading speech recognition performance.

Observation process adaptation for linear dynamic models

This work introduces two methods for adapting the observation process parameters of linear dynamic models (LDM) or other linear-Gaussian models. The first method uses the expectation-maximization (EM) algorithm to estimate transforms for location and covariance parameters, and the second uses a generalized EM (GEM) approach which reduces computation in making updates from O( p 6) to O( p 3), where p is the feature dimension. We present the results of speaker adaptation on TIMIT phone classification and recognition experiments with relative error reductions of up to 6%. Importantly, we find minimal differences in the results from EM and GEM. We therefore propose that the GEM approach be applied to adaptation of hidden Markov models which use non-diagonal covariances. We provide the necessary update equations.

Robust Speech Recognition by Using Compensated Acoustic Scores

This paper proposes a new compensation method of acoustic scores in the Viterbi search for robust speech recognition. This method introduces noise models to represent a wide variety of noises and realizes robust decoding together with conventional techniques of subtraction and adaptation. This method uses likelihoods of noise models in two ways. One is to calculate a confidence factor for each input frame by comparing likelihoods of speech models and noise models. Then the weight of the acoustic score for a noisy frame is reduced according to the value of the confidence factor for compensation. The other is to use the likelihood of noise model as an alternative that of a silence model when given noisy input. Since a lower confidence factor compresses acoustic scores, the decoder rather relies on language scores and keeps more hypotheses within a fixed search depth for a noisy frame. An experiment using commentary transcriptions of a broadcast sports program (MLB: Major League Baseball) showed that the proposed method obtained a 6.7% relative word error reduction. The method also reduced the relative error rate of key words by 17.9%, and this is expected lead to an improvement metadata extraction accuracy.

Nonparametric Speaker Recognition Method Using Earth Mover's Distance

In this paper, we propose a distributed speaker recognition method using a nonparametric speaker model and Earth Mover's Distance (EMD). In distributed speaker recognition, the quantized feature vectors are sent to a server. The Gaussian mixture model (GMM), the traditional method used for speaker recognition, is trained using the maximum likelihood approach. However, it is difficult to fit continuous density functions to quantized data. To overcome this problem, the proposed method represents each speaker model with a speaker-dependent VQ code histogram designed by registered feature vectors and directly calculates the distance between the histograms of speaker models and testing quantized feature vectors. To measure the distance between each speaker model and testing data, we use EMD which can calculate the distance between histograms with different bins. We conducted text-independent speaker identification experiments using the proposed method. Compared to results using the traditional GMM, the proposed method yielded relative error reductions of 32% for quantized data.

Shape-dependent regularization for the retrieval of atmospheric state parameter profiles

We present an adaptive regularization approach to retrieve vertical state parameter profiles from limb-sounding measurements with high accuracy. This is accomplished by introducing a dedicated regularization functional based on a reasonable assumption of the profile characteristics. The approach results in shape-dependent weighting during least-squares computations and relies on a Cholesky decomposition of a preselected L(T)L matrix. Our method is compared with established regularization functionals such as optimal estimation and Tikhonov with respect to errors and achievable height resolution. The results show an improved height resolution of the retrieved profiles together with a reduction of absolute and relative errors obtained by test-bed simulations.

Classification of acoustic events using SVM-based clustering schemes

Acoustic events produced in controlled environments may carry information useful for perceptually aware interfaces. In this paper we focus on the problem of classifying 16 types of meeting-room acoustic events. First of all, we have defined the events and gathered a sound database. Then, several classifiers based on support vector machines (SVM) are developed using confusion matrix based clustering schemes to deal with the multi-class problem. Also, several sets of acoustic features are defined and used in the classification tests. In the experiments, the developed SVM-based classifiers are compared with an already reported binary tree scheme and with their correlative Gaussian mixture model (GMM) classifiers. The best results are obtained with a tree SVM-based classifier that may use a different feature set at each node. With it, a 31.5% relative average error reduction is obtained with respect to the best result from a conventional binary tree scheme.

Ambiguity Resolution Analysis in Incremental Parsing of Natural Language

Incremental parsing gains its importance in natural language processing and psycholinguistics because of its cognitive plausibility. Modeling the associated cognitive data structures, and their dynamics, can lead to a better understanding of the human parser. In earlier work, we have introduced a recursive neural network (RNN) capable of performing syntactic ambiguity resolution in incremental parsing. In this paper, we report a systematic analysis of the behavior of the network that allows us to gain important insights about the kind of information that is exploited to resolve different forms of ambiguity. In attachment ambiguities, in which a new phrase can be attached at more than one point in the syntactic left context, we found that learning from examples allows us to predict the location of the attachment point with high accuracy, while the discrimination amongst alternative syntactic structures with the same attachment point is slightly better than making a decision purely based on frequencies. We also introduce several new ideas to enhance the architectural design, obtaining significant improvements of prediction accuracy, up to 25% error reduction on the same dataset used in previous work. Finally, we report large scale experiments on the entire Wall Street Journal section of the Penn Treebank. The best prediction accuracy of the model on this large dataset is 87.6%, a relative error reduction larger than 50% compared to previous results.

Development of land surface albedo parameterization based on Moderate Resolution Imaging Spectroradiometer (MODIS) data

A new dynamic‐statistical parameterization of snow‐free land surface albedo is developed using the Moderate Resolution Imaging Spectroradiometer (MODIS) products of broadband black‐sky and white‐sky reflectance and vegetation and the North American and Global Land Data Assimilation System (LDAS) outputs of soil moisture during 2000–2003. The dynamic component represents the predictable albedo dependences on solar zenith angle, surface soil moisture, fractional vegetation cover, leaf plus stem area index, and greenness, while the statistical part represents the correction for static effects that are specific to local surface characteristics. All parameters of the dynamic and statistical components are determined by solving nonlinear constrained optimization problems of a physically based conceptual model for the minimization of the bulk variances between simulations and observations. They all depend on direct beam or diffuse radiation and visible or near‐infrared band. The dynamic parameters are also functions of land cover category, while the statistical factors are specific to geographic location. The new parameterization realistically represents surface albedo variations, including the mean, shape, and distribution, around each dependent parameter. For composites of all temporal and spatial samples of the same land cover category over North America, correlation coefficients between the dynamic component of the new parameterization and the MODIS data range from 0.39 to 0.88, while relative errors vary within 8–42%. The gross (i.e., integrated over all categories) correlations and errors are 0.57–0.71 and 17–26%, changing with direct beam or diffuse radiation and visible or near‐infrared band. The static local correction results in a further reduction in relative errors, producing gross values of 11–21%. The new parameterization is a marked improvement over the existing albedo scheme of the state‐of‐the‐art Common Land Model (CLM), which has correlation coefficients from −0.57 to 0.71 and relative errors of 18–140% for individual land cover categories, and gross values of 0.03–0.32 and 37–71%, respectively.

Improvement of Thomson Scattering Diagnostics Using Stimulated-Brillouin-Scattering-Based Phase Conjugate Mirrors

In order to improve the measurement performance of incoherent Thomson scattering diagnostics, a high performance phase conjugate mirror based on stimulated Brillouin scattering (SBS-PCM) is applied to a Thomson scattering system for the first time in the JT-60U tokamak. We have demonstrated that a SBS-PCM which uses heavyfluorocarbon liquid showed a high reflectivity of 95% at a high input-power of 145 W. Using the SBS-PCM, two newly developed methods were employed to increase the amount of scattered light. In the first method, we first developed a new optical design to provide a double-pass scattering scheme with the SBS-PCM. In this new optical design, a laser beam passing through the plasma is reflected by the SBS-PCM, and the reflected beam is returned via the same path by means of the phase conjugate effect, and is then passed through the plasma again, in order to increase the scattered light. A double-pass Thomson scattering scheme using the SBS-PCM was demonstrated in JT-60U ohmic plasma, resulting in an increase of the scattered light by a factor of 1.6, and the reduction of relative error by 2/3 for electron temperature measurement in contrast to single-pass scattering. A multi-pass Thomson scattering scheme is also proposed based on the results of double-pass scattering. It is estimated that multi-pass scattering allows the generation of several times the amount of scattered light, and the reduction of the relative error for electron temperature measurement by 37% in contrast to single-pass scattering. Regarding the second method, a high average-power of YAG laser system was developed by applying the SBS-PCM to a existent diagnostic laser. As a result, the average-power was increased by over 8 times in contrast to the average power of the original system, achieving up to 368 W (7.4 J × 50 Hz).

Taxon sampling, bioinformatics, and phylogenomics.

Taxon sampling, bioinformatics, and phylogenomics.