Small Amount Of Training Data Research Articles

Detecting movement-related EEG change by wavelet decomposition-based neural networks trained with single thumb movement

Objective The main goal of this study was to develop a real-time detection algorithm of movement-related EEG changes for the naïve subjects with a very small amount of training data. Such an algorithm is vital for the realization of brain–computer interface. Methods The target algorithm developed in this study was based on the wavelet decomposition neural network (WDNN). Surface Laplacian EEG was recorded at central cortical areas and processed with wavelet decomposition (WD) for feature extraction and neural network for pattern recognition. The new algorithm was compared with nother three methods, namely, threshold-based WD and short-time Fourier transform (STFT), and Fourier transform neural network (FTNN), for performance. The trainings of all algorithms were based, respectively, on the changes of μ and β rhythms before and after voluntary movements. In order to investigate whether WDNN could adapt to the nonstationarity of EEG or not, we also compared two training modes, namely, fixed and updated weight. The significances of the success rates were tested by ANOVA (analysis of variance) and verified by ROC (receiver operating characteristic) analysis. Results The experimental data showed that (1) success rates of movement detection were acceptable even when the training set was reduced to a single trial data, (2) WDNN performed better than WD or STFT without optimized thresholds and (3) when weights were updated and thresholds were optimized, WDNN still performed better than WD, while FTNN had a marginal advantage over STFT. Conclusions We developed a detection algorithm based on WDNN with the training set being reduced to a single trial data. The overall performance of this algorithm was better than the conventional methods as such. Significance μ wave suppression could be detected more precisely by the wavelet decomposition with neural network than the conventional algorithms such as STFT and WD. The size of training data could be reduced to a single trial and the success rates were up to 75–80%.

Fast estimation of a precise dereverberation filter based on the harmonic structure of speech

A speech signal captured by a distant microphone is generally smeared by reverberation. This severely degrades both the speech intelligibility and the Automatic Speech Recognition (ASR) performance. We have already proposed a novel dereverberation technique named “Harmonicity based dEReverBeration (HERB),” which utilizes an essential feature of speech, namely harmonics, and estimates an inverse filter for an unknown impulse response. If a large amount of acoustically stable training data is available, HERB is able to estimate an accurate inverse filter even in severely reverberant environments. In general, however, a dereverberation algorithm has to work with small amounts of training data, because the acoustic property of a real world environment changes according to various factors such as the speaker’s position and room temperature. In this paper, we propose a new dereverberation scheme based on HERB, aiming primarily at reducing the amount of training data needed to estimate an inverse filter. The proposed method re-estimates a more accurate source signal based on the output signal of conventional HERB, and re-calculates the inverse filter. We show experimentally that our new dereverberation scheme successfully achieves high quality dereverberation with much smaller amounts of training data, and is very effective at improving both audible quality and ASR performance, even in unknown severely reverberant environments.

Evaluation of eigenvoice conversion based on Gaussian mixture model

Eigenvoice conversion (EVC) has been proposed as a new framework of voice conversion (VC) based on the Gaussian mixture model (GMM) [Toda et al., ‘‘Eigenvoice Conversion Based on Gaussian Mixture Model,’’ ICSLP, Pittsburgh, Sept. 2006]. This paper evaluates the performance of EVC in conversion from one source speaker’s voice to an arbitrary target speakers’ voices. This framework trains canonical GMM (EV‐GMM) in advance using multiple parallel data sets consisting of utterance pairs of the source and many prestored target speakers. This model is adapted to a specific target speaker by estimating a small number of free parameters using a few utterances of the target speaker. This paper compares spectral distortion between converted and target voices in EVC with conventional VC based on GMM when varying the amount of training data and the number of mixtures. Results show EVC outperforms conventional VC when using small amounts of training data. EVC can effectively train a complex conversion model using the information of many prestored speakers. By contrast, conventional VC needs a large‐sized parallel data set for training. It also shows the results of subjective evaluations of speech quality and conversion accuracy for speaker individuality.

A New Iterative Decoder of Space-Time Data in Time Selective MIMO Channels

In this paper we present a new iterative decoder of Space-Time Block Coded data in the presence of carrier offsets. This decoder starts the iteration with a small amount of training data to find the initial estimate of carrier offset, channel and noise co-variance. Subsequently it attempts to find the estimate of the STBC data by an alternating maximization (AM) technique. In this way better estimates of Space---Time data are achieved after the final iteration as compared to those found without an iterative method. Thus the new decoder reduces the training overhead required by the conventional data-aided (DA) estimators for efficient estimates of carrier offsets and S---T data, and provides higher data rate.

Unvoiced Speech Recognition Using Tissue-Conductive Acoustic Sensor

We present the use of stethoscope and silicon NAM (nonaudible murmur) microphones in automatic speech recognition. NAM microphones are special acoustic sensors, which are attached behind the talker's ear and can capture not only normal (audible) speech, but also very quietly uttered speech (nonaudible murmur). As a result, NAM microphones can be applied in automatic speech recognition systems when privacy is desired in human-machine communication. Moreover, NAM microphones show robustness against noise and they might be used in special systems (speech recognition, speech transform, etc.) for sound-impaired people. Using adaptation techniques and a small amount of training data, we achieved for a 20 k dictation task a word accuracy for nonaudible murmur recognition in a clean environment. In this paper, we also investigate nonaudible murmur recognition in noisy environments and the effect of the Lombard reflex on nonaudible murmur recognition. We also propose three methods to integrate audible speech and nonaudible murmur recognition using a stethoscope NAM microphone with very promising results.

Bayesian Population Decoding of Motor Cortical Activity Using a Kalman Filter

Effective neural motor prostheses require a method for decoding neural activity representing desired movement. In particular, the accurate reconstruction of a continuous motion signal is necessary for the control of devices such as computer cursors, robots, or a patient's own paralyzed limbs. For such applications, we developed a real-time system that uses Bayesian inference techniques to estimate hand motion from the firing rates of multiple neurons. In this study, we used recordings that were previously made in the arm area of primary motor cortex in awake behaving monkeys using a chronically implanted multielectrode microarray. Bayesian inference involves computing the posterior probability of the hand motion conditioned on a sequence of observed firing rates; this is formulated in terms of the product of a likelihood and a prior. The likelihood term models the probability of firing rates given a particular hand motion. We found that a linear gaussian model could be used to approximate this likelihood and could be readily learned from a small amount of training data. The prior term defines a probabilistic model of hand kinematics and was also taken to be a linear gaussian model. Decoding was performed using a Kalman filter, which gives an efficient recursive method for Bayesian inference when the likelihood and prior are linear and gaussian. In off-line experiments, the Kalman filter reconstructions of hand trajectory were more accurate than previously reported results. The resulting decoding algorithm provides a principled probabilistic model of motor-cortical coding, decodes hand motion in real time, provides an estimate of uncertainty, and is straightforward to implement. Additionally the formulation unifies and extends previous models of neural coding while providing insights into the motor-cortical code.

TEG—a hybrid approach to information extraction

This paper describes a hybrid statistical and knowledge-based information extraction model, able to extract entities and relations at the sentence level. The model attempts to retain and improve the high accuracy levels of knowledge-based systems while drastically reducing the amount of manual labour by relying on statistics drawn from a training corpus. The implementation of the model, called TEG (trainable extraction grammar), can be adapted to any IE domain by writing a suitable set of rules in a SCFG (stochastic context-free grammar)-based extraction language and training them using an annotated corpus. The system does not contain any purely linguistic components, such as PoS tagger or shallow parser, but allows to using external linguistic components if necessary. We demonstrate the performance of the system on several named entity extraction and relation extraction tasks. The experiments show that our hybrid approach outperforms both purely statistical and purely knowledge-based systems, while requiring orders of magnitude less manual rule writing and smaller amounts of training data. We also demonstrate the robustness of our system under conditions of poor training-data quality.

Offline handwritten Chinese character recognition by radical decomposition

Offline handwritten Chinese character recognition is a very hard pattern-recognition problem of considerable practical importance. Two popular approaches are to extract features holistically from the character image or to decompose characters structurally into component parts---usually strokes. Here we take a novel approach, that of decomposing into radicals on the basis of image information (i.e., without first decomposing into strokes). During training, 60 examples of each radical were represented by "landmark" points, labeled semiautomatically, with radicals in different characteristic positions treated as distinctly different radicals. Kernel principal-component analysis then captured the main (nonlinear) variations around the mean radical. During the recognition, the dynamic tunneling algorithm was used to search for optimal shape parameters in terms of chamfer distance minimization. Considering character composition as a Markov process in which up to four radicals are combined in some assumed sequential order, we can recognize complete, hierarchically-composed characters by using the Viterbi algorithm. This gave a character recognition rate of 93.5% characters correct (writer-independent) on a test set of 430,800 characters from 2,154 character classes composed of 200 radical categories, which is comparable to the best reported results in the literature. Although the initial semiautomatic landmark labeling is time consuming, the decomposition approach is theoretically well-motivated and allows the different sources of variability in Chinese handwriting to be handled separately and by the most appropriate means--either learned from example data or incorporated as prior knowledge. Hence, high generalizability is obtained from small amounts of training data, and only simple prior knowledge needs to be incorporated, thus promising robust recognition performance. As such, there is very considerable potential for further development and improvement in the direction of larger character sets and less constrained writing conditions.

Thresholding method for dimensionality reduction in recognition systems

Often recognition systems must be designed with a relatively small amount of training data. Plug-in test statistics suffer from large estimation errors, often causing the performance to degrade as the measurement vector dimension increases. Choosing a better test statistic or applying a method of dimensionality reduction are two possible solutions to this problem. In this paper, we consider a recognition problem where the data for each population are assumed to have the same parametric distribution but differ in their unknown parameters. The collected vectors of data as well as their components are assumed to be independent. The system is designed to implement a plug-in log-likelihood ratio test with maximum-likelihood (ML) estimates of the unknown parameters instead of the true parameters. Because a small amount of data is available to estimate the parameters, the performance of such a system is strongly degraded relative to the performance with known parameters. To improve the performance of the system we define a thresholding function that, when incorporated into the plug-in log-likelihood ratio, significantly decreases the probability of error for binary and multiple hypothesis testing problems for the exponential class of populations. We analyze the modified test statistic and present the results of Monte Carlo simulation. Special attention is paid to the complex Gaussian model with zero mean and unknown variances.

Introduction to the Special Issue on Learning in Intelligent Algorithms and Systems Design

Introduction to the Special Issue on Learning in Intelligent Algorithms and Systems Design

Determining the Optimal Number of Clusters by an Extended RPCL Algorithm

Determining an appropriate number of clusters is a difficult yet important problem that the rival penalized competitive learning (RPCL) algorithm was designed to solve, but its performance is not satifactory with overlapping clusters or cases where input vectors contain dependent components. We address this problem by incorporating full covariance matrices into the original RPCL algorithm. The resulting extended RPCL algorithm progressively eliminates units whose clusters contain only a small amount of training data. The algorithm is used to determine the number of clusters in a Gaussian distribution. It is also used to optimize the architecture of elliptical basis function networks for speaker verification and vowel classification. We found that covariance matrices obtained by the extended RPCL algorithm have a better representation of clusters than those obtained by the original RPCL algorithm, resulting in a lower verification error rate in speaker verification and a higher recognition accuracy in vowel classification.

An olfactory neuronal network for vapor recognition in an artificial nose.

Odorant sensitivity and discrimination in the olfactory system appear to involve extensive neural processing of the primary sensory inputs from the olfactory epithelium. To test formally the functional consequences of such processing, we implemented in an artificial chemosensing system a new analytical approach that is based directly on neural circuits of the vertebrate olfactory system. An array of fiber-optic chemosensors, constructed with response properties similar to those of olfactory sensory neurons, provide time-varying inputs to a computer simulation of the olfactory bulb (OB). The OB simulation produces spatiotemporal patterns of neuronal firing that vary with vapor type. These patterns are then recognized by a delay line neural network (DLNN). In the final output of these two processing steps, vapor identity is encoded by the spatial patterning of activity across units in the DLNN, and vapor intensity is encoded by response latency. The OB-DLNN combination thus separates identity and intensity information into two distinct codes carried by the same output units, enabling discrimination among organic vapors over a range of input signal intensities. In addition to providing a well-defined system for investigating olfactory information processing, this biologically based neuronal network performs better than standard feed-forward neural networks in discriminating vapors when small amounts of training data are used.

Affine transformation-based feature normalization for speaker-independent speech recognition

In statistical speech recognition, speaker-independent models are usually trained by using speech samples from a large number of speakers. Those models have a problem in that they have wider feature distributions and hence greater overlaps between different phones than adequately trained speaker-dependent models. In order to cope with the interspeaker variability, a method of speech feature normalization based on affine transformation has been presented [P. Luo and K. Ozeki, Tech. Rep. of IEICE, SP96-10 (1996)]. Prior to HMM training, feature vectors of each speaker are mapped to those of a reference speaker by an affine transformation estimated with a small amount of training data. The transformation, which is phone independent and speaker dependent, is also applied to feature vectors of unknown speakers in the recognition stage. It has been shown experimentally that this method is effective in reducing interspeaker variations in the cepstral domain. In this paper, further discussions about the performance and limitations of the method are given within the framework of continuous HMM. Practical issues related to the selection of an appropriate reference speaker will also be discussed.

Fuzzy network model for part-of-speech tagging under small training data

Recently, most part-of-speech tagging approaches, such as rule-based, probabilistic and neural network approaches, have shown very promising results. In this paper, we are particularly interested in probabilistic approaches, which usually require lots of training data to get reliable probabilities. We alleviate such a restriction of probabilistic approaches by introducing a fuzzy network model to provide a method for estimating more reliable parameters of a model under a small amount of training data. Experiments with the Brown corpus show that the performance of the fuzzy network model is much better than that of the hidden Markov model under a limited amount of training data.

Speaker-independent speech recognition based on tree-structured speaker clustering

We have already proposed the application of tree-structured speaker clustering to supervised speaker adaptation. This paper proposes its application to unsupervised speaker adaptation and speaker-independent (SI) speech recognition. This clustering involves the selection of a speaker cluster from among multiple reference speaker clusters arranged in a tree structure. Cluster selection, unlike parameter training, enables quick adaptation using only a small amount of training data. This method was applied to a hidden Markov network (HMnet) and evaluated in Japanese phoneme and phrase recognition experiments. Results show effective unsupervised speaker adaptation using only 5 s calibration speech. In the SI speech recognition experiments, the method reduced the error rate by 8·5% compared with the conventional speaker-independent speech recognition method.

Speech spectrum conversion based on speaker interpolation and multi-functional representation with weighting by radial basis function networks

This paper describes a speech spectrum transformation method by interpolating multi-speakers' spectral patterns and multi-functional representation with Radial Basis Function networks. The interpolation is carried out using spectral parameters between pre-stored multiple speakers' utterance data to generate new spectrum patterns. Adaptation to a target speaker can be performed by this interpolation, which uses only a small amount of training data to generate new speech spectrum sequences close to those of the target speaker. Moreover, to obtain more precise adaptation by using a larger amount of training data, the transformation is represented by multiple interpolating functions. The multiple functions' outputs are weighted-summed, using weighting values given by RBF networks. The parameters of this multi-functional transformation are adapted by the gradient descent method. Adaptation experiments were carried out using four pre-stored speakers' data. Using only one word spoken by the target speaker for training, the distance between the target speaker's spectrum and the spectrum generated by the single interpolating function was reduced by about 35% compared with the distance between the target speaker's spectrum and the spectrum of the pre-stored speaker closest to the target. Using ten training words, the reduction rate increased to 48% by the multi-functional transformation.

Maximum likelihood adaptive neural controller

A concept of model-based neural controller is described, which incorporates a model of a controlled system into a neural network architecture. This concept results in efficient learning requiring small amounts of training data. This is due to the fact that all synapse weights are determined by a relatively small number of model parameters, which can be learned quickly. The development in this paper is based on the MLANS neural network developed previously for classification. The current paper addresses the development of a model that is generally applicable to the problem of multidimensional nonlinear control and that is useful for model-based neural network development. We describe modifications to the MLANS architecture, which are needed to incorporate the control model into the neural network architecture and for the neural dynamics to converge to the maximum likelihood values of model parameters. Then, we present examples comparing the neural network performance to that of the classical least squares controller.

Formant normalisation for speech recognition and vowel studies

Vowel formant target frequencies from different talkers depend on the details of the vocal tract, sex, regional accent, speaking habits and other factors. Good vowel recognition and studies of vowels from different talkers require an accurate method for compensating for speaker differences in these frequencies. The major variance seen in the data is between males and females. However, even within the same sex class, there are large variations in the formant target frequencies for the same vowel in the same phonetic context. Various methods of compensating for speaker variation in formants were studied. Bark scaled formants and subtraction of Bark fundamental frequency from the first formant was tried first. In spite of recent published papers on the efficacy of this technique, it was found inadequate. The transformations were incapable of improving the clusters of the cardinal vowels, for example. A modification of the Gerstman technique, determining the speaker's formant range and then transforming into an “ideal” talker's range, was found to account for most of the variance due to different talkers given a small amount of training data. This technique was applied to vowel in context studies on American English. Formant ranges were studied for 125 talkers of General American English. Plots of formant ranges for males and females showed interesting patterns. The lower limit of the second formant was not very different, while the lower limit of the first formant was lower for males. Both the first and second formant maxima were larger for females. The modified Gerstman transformation was able to superimpose the formant targets for the same vowel in the same context from different talkers into the same region of F1, F2 space. There remained some residual variance between male and female, even after the transformation. These trends are shown in a series of plots of vowel target frequency data.

Symbolic and neural learning algorithms: an experimental comparison

Despite the fact that many symbolic and neural network (connectionist) learning algorithms address the same problem of learning from classified examples, very little is known regarding their comparative strengths and weaknesses. Experiments comparing the ID3 symbolic learning algorithm with the perception and backpropagation neural learning algorithms have been performed using five large, real-world data sets. Overall, backpropagation performs slightly better than the other two algorithms in terms of classification accuracy on new examples, but takes much longer to train. Experimental results suggest that backpropagation can work significantly better on data sets containing numerical data. Also analyzed empirically are the effects of (1) the amount of training data, (2) imperfect training examples, and (3) the encoding of the desired outputs. Backpropagation occasionally outperforms the other two systems when given relatively small amounts of training data. It is slightly more accurate than ID3 when examples are noisy or incompletely specified. Finally, backpropagation more effectively utilizes a “distributed” output encoding.

Large vocabulary word recognition using context-dependent allophonic hidden Markov models

In this paper, we report our development of context-dependent allophonic hidden Markov models (HMMs) implemented in a 75 000-word speaker-dependent Gaussian-HMM recognizer. The context explored is the immediate left and/or right adjacent phoneme. To achieve reliable estimation of the model parameters, phonemes are grouped into classes based on their expected co-articulatory effects on neighboring phonemes. Only five separate preceding and following contexts are identified explicitly for each phoneme. By grouping the contexts we ensure that they occur frequently enough in the training data to allow reliable estimation of the parameters of the HMM representing the context-dependent units. Further improvement in the estimation reliability is obtained by tying the covariance matrices in the HMM output distributions across all contexts. Speech recognition experiments show that when a large amount of data (e.g. over 2500 words) is used to train context-dependent HMMs, the word recognition error rate is reduced by 33%, compared with the context-independent HMMs. For smaller amounts of training data the error reduction becomes less significant.