LBG Algorithm Research Articles

Pattern Recognition Methods for Emotion Recognition with speech signal

In this paper, we apply several pattern recognition algorithms to emotion recognition system with speech signal and compare the results. Firstly, we need emotional speech databases. Also, speech features for emotion recognition are determined on the database analysis step. Secondly, recognition algorithms are applied to these speech features. The algorithms we try are artificial neural network, Bayesian learning, Principal Component Analysis, LBG algorithm. Thereafter, the performance gap of these methods is presented on the experiment result section.

BER criterion and codebook construction for finite-rate precoded spatial multiplexing with linear receivers

Precoded spatial multiplexing systems with rate-limited feedback have been studied recently based on various precoder selection criteria. Instead of those based on indirect performance indicators, we in this paper propose a new criterion directly based on the exact bit error rate (BER) that is applicable to systems with linear receivers and rectangular/square quadrature-amplitude-modulation constellations. The BER criterion outperforms any other alternative in terms of optimizing the BER performance for an uncoded system with linear receivers. We then develop a precoder codebook construction method based on the generalized Lloyd algorithm from the vector quantization literature. This construction is not directly based on the BER criterion. Hence, it is suboptimal in the BER sense. However, relative to those currently available, our newfound codebooks improve considerably various minimum distances between any pair of codewords of the codebook. Finally, we analyze the BER-optimal precoder in the asymptotic case with infinite-rate feedback that amounts to perfect channel knowledge at the transmitter. The infinite-rate optimal precoder based on the BER criterion is drastically different from the counterparts with other criteria, and it leads to a benchmark performance for finite-rate precoded spatial multiplexing systems. We observe from numerical results that the BER performance of finite-rate feedback with suboptimal codebooks approaches quickly the benchmark performance of infinite-rate feedback. This suggests that i) the number of feedback bits in practical systems need not be large and ii) the room for performance improvement via further codebook optimization shrinks quickly as the codebook size increases.

음성신호를 이용한 감성인식에서의 패턴인식 방법

In this paper, we apply several pattern recognition algorithms to emotion recognition system with speech signal and compare the results. Firstly, we need emotional speech databases. Also, speech features for emotion recognition is determined on the database analysis step. Secondly, recognition algorithms are applied to these speech features. The algorithms we try are artificial neural network, Bayesian learning, Principal Component Analysis, LBG algorithm. Thereafter, the performance gap of these methods is presented on the experiment result section. Truly, emotion recognition technique is not mature. That is, the emotion feature selection, relevant classification method selection, all these problems are disputable. So, we wish this paper to be a reference for the disputes.

Wavelet based compression of medical ultrasound images using vector quantization

In this paper, an efficient technique for compression of medical ultrasound (US) images is proposed. The technique is based on wavelet transform of the original image combined with vector quantization (VQ) of high-energy subbands using the LBG algorithm. First, we analyse the statistical behaviour of wavelet coefficients in US images across various subbands and scales. The analysis show that most of the image energy is concentrated in one of the detail subband, either in the vertical detail subband (most of the time) or in the horizontal subband. The other two subbands at each decomposition level contribute negligibly to the total image energy. Then, by exploiting this statistical analysis, a low-complexity image coder is designed, which applies VQ only to the highest energy subband while discarding the other detail subbands at each level of decomposition. The coder is tested on a series of abdominal and uterus greyscale US images. The experimental results indicate that the proposed method clearly outperforms the JPEG2000 (Joint Photographers Expert Group) encoder both qualitatively and quantitatively. For example, without using any entropy coder, the proposed method yields a peak signal to noise ratio gain of 0.2 dB to 1.2 dB over JPEG2000 on medical US images.

LSP quantization by a union of locally trained codebooks

We present a fixed rate encoding scheme for the line spectrum pair (LSP) representation of an LPC-filter, based on Gaussian mixture (GM) modeling. For each mixture component, we construct a codebook by a union of product quantizers. Each local codebook is trained, independently, using a clustering scheme similar to the generalized Lloyd algorithm (GLA), over synthetic data. The training algorithm iterates fast, due to low complexity encoding, and converges in few iterations. The overall codebook is a combination of local codebooks, and inherits their high performance, while having a moderate complexity. We provide numerical results for average spectral distortion (SD) of the proposed encoder, and benchmark them by a lower bound, according to high-rate theory. We achieve an average SD (full-band measure) of 1 dB at 23 b/frame, for speech signals sampled at 8 kHz and LPC of order 10. By tolerating additional complexity, we reach a SD within 0.01 dB of the lower bound.

Reduced-Complexity Deterministic Annealing for Vector Quantizer Design

This paper presents a reduced-complexity deterministic annealing (DA) approach for vector quantizer (VQ) design by using soft information processing with simplified assignment measures. Low-complexity distributions are designed to mimic the Gibbs distribution, where the latter is the optimal distribution used in the standard DA method. These low-complexity distributions are simple enough to facilitate fast computation, but at the same time they can closely approximate the Gibbs distribution to result in near-optimal performance. We have also derived the theoretical performance loss at a given system entropy due to using the simple soft measures instead of the optimal Gibbs measure. We use the derived result to obtain optimal annealing schedules for the simple soft measures that approximate the annealing schedule for the optimal Gibbs distribution. The proposed reduced-complexity DA algorithms have significantly improved the quality of the final codebooks compared to the generalized Lloyd algorithm and standard stochastic relaxation techniques, both with and without the pairwise nearest neighbor (PNN) codebook initialization. The proposed algorithms are able to evade the local minima and the results show that they are not sensitive to the choice of the initial codebook. Compared to the standard DA approach, the reduced-complexity DA algorithms can operate over 100 times faster with negligible performance difference. For example, for the design of a 16-dimensional vector quantizer having a rate of 0.4375 bit/sample for Gaussian source, the standard DA algorithm achieved 3.60 dB performance in 16 483 CPU seconds, whereas the reduced-complexity DA algorithm achieved the same performance in 136 CPU seconds. Other than VQ design, the DA techniques are applicable to problems such as classification, clustering, and resource allocation.

Efficient vector quantization using genetic algorithm

This paper proposes a new codebook generation algorithm for image data compression using a combined scheme of principal component analysis (PCA) and genetic algorithm (GA). The combined scheme makes full use of the near global optimal searching ability of GA and the computation complexity reduction of PCA to compute the codebook. The experimental results show that our algorithm outperforms the popular LBG algorithm in terms of computational efficiency and image compression performance.

Achieving the Welch Bound With Difference Sets

Consider a codebook containing N unit-norm complex vectors in a K-dimensional space. In a number of applications, the codebook that minimizes the maximal cross-correlation amplitude (I/sub max/) is often desirable. Relying on tools from combinatorial number theory, we construct analytically optimal codebooks meeting, in certain cases, the Welch lower bound. When analytical constructions are not available, we develop an efficient numerical search method based on a generalized Lloyd algorithm, which leads to considerable improvement on the achieved I/sub max/ over existing alternatives. We also derive a composite lower bound on the minimum achievable I/sub max/ that is effective for any codebook size N.

Mutual information-based analysis of JPEG2000 contexts

Context-based arithmetic coding has been widely adopted in image and video compression and is a key component of the new JPEG2000 image compression standard. In this paper, the contexts used in JPEG2000 are analyzed using the mutual information, which is closely related to the compression performance. We first show that, when combining the contexts, the mutual information between the contexts and the encoded data will decrease unless the conditional probability distributions of the combined contexts are the same. Given I, the initial number of contexts, and F, the final desired number of contexts, there are S(I, F) possible context classification schemes where S(I, F) is called the Stirling number of the second kind. The optimal classification scheme is the one that gives the maximum mutual information. Instead of using an exhaustive search, the optimal classification scheme can be obtained through a modified generalized Lloyd algorithm with the relative entropy as the distortion metric. For binary arithmetic coding, the search complexity can be reduced by using dynamic programming. Our experimental results show that the JPEG2000 contexts capture the correlations among the wavelet coefficients very well. At the same time, the number of contexts used as part of the standard can be reduced without loss in the coding performance.

Novel vector quantiser design using reinforced learning as a pre-process

This paper presents the development and evaluation of a new approach toward the design of optimised codebooks by vector quantisation (VQ). A strategy of reinforced learning (RL) is proposed which exploits the advantages offered by fuzzy clustering algorithms, competitive learning and knowledge of training vector and codevector configurations. RL is used as a pre-process before using a conventional VQ algorithm such as the generalised Lloyd algorithm (GLA) or the fuzzy k-means (FKM) algorithm. At each iteration of RL, codevectors move intelligently and intentionally toward an improved optimum codebook design. This is distinct from simulated annealing (SA) and genetic algorithm (GA) techniques in which a random variation is introduced in the movement of the codevectors. The new strategy reduces the possibility that, in the final design, codevectors will be overcrowded in the high density distribution regions of the training vector space, and also that too few codevectors will be settled in low density regions. Experiments demonstrate that this results in a more effective representation of the training vectors by the codevectors and that the final codebook is nearer to the optimal solution in applications such as image compression. It has been found that GLA and FKM yield improved quality of codebook design in this application when RL is used as a pre-process. The investigations have also indicated that RL is insensitive to the selection of both the initial codebook and a learning rate control parameter, which is the only additional parameter introduced by RL from the standard FKM.

On Finite-State Vector Quantization for Noisy Channels

Finite-state vector quantization (FSVQ) over a noisy channel is studied. A major drawback of a finite-state decoder is its inability to track the encoder in the presence of channel noise. In order to overcome this problem, we propose a nontracking decoder which directly estimates the code vectors used by a finite-state encoder. The design of channel-matched finite-state vector quantizers for noisy channels, using an iterative scheme resembling the generalized Lloyd algorithm, is also investigated. Simulation results based on encoding a Gauss-Markov source over a memoryless Gaussian channel show that the proposed decoder exhibits graceful degradation of performance with increasing channel noise, as compared with a finite-state decoder. Also, the channel-matched finite-state vector quantizers are shown to outperform channel-optimized vector quantizers having the same vector dimension and rate. However, the nontracking decoder used in the channel-matched finite-state quantizer has a higher computational complexity, compared with a channel-optimized vector-quantizer decoder. Thus, if they are allowed to have the same overall complexity (encoding and decoding), the channel-optimized vector quantizer can use a longer encoding delay and achieve similar or better performance. Finally, an example of using the channel-matched finite-state quantizer as a backward-adaptive quantizer for nonstationary signals is also presented.

Medical image compression using topology-preserving neural networks

A novel method based on topology-preserving neural networks is used to implement vector quantization for medical image compression. The described method is an innovative image compression procedure, which differentiates itself from known systems in several ways. It can be applied to larger image blocks and represents better probability distribution estimation methods. A transformation-based operation is applied as part of the encoder on the block-decomposed image. The quantization process is performed by a “neural-gas” network which applied to vector quantization converges quickly to low distortion errors and reaches a distortion error lower than that resulting from Kohonen's feature map or the LBG algorithm. To study the efficiency of our algorithm, we blended mathematical phantom features into clinically proved cancer free mammograms. The influence of the neural compression method on the phantom features and the mammo-graphic image is not visually perceptible up to a high compression rate.

LBGS: a smart approach for very large data sets vector quantization

In this paper, LBGS, a new parallel/distributed technique for Vector Quantization is presented. It derives from the well known LBG algorithm and has been designed for very complex problems where both large data sets and large codebooks are involved. Several heuristics have been introduced to make it suitable for implementation on parallel/distributed hardware. These lead to a slight deterioration of the quantization error with respect to the serial version but a large improvement in computing efficiency.

Hyperspectral data compression using a fast vector quantization algorithm

A fast vector quantization algorithm for data compression of hyperspectral imagery is proposed in this paper. It makes use of the fact that in the full search of the generalized Lloyd algorithm (GLA) a training vector does not require a search to find the minimum distance partition if its distance to the partition is improved in the current iteration compared to that of the previous iteration. The proposed method has the advantage of being simple, producing a large computation time saving and yielding compression fidelity as good as the GLA. Four hyperspectral data cubes covering a wide variety of scene types were tested. The loss of spectral information due to compression was evaluated using the spectral angle mapper and a remote sensing application.

Vector Quantization Approach to Classification of Stressed Speech

In this work, a stressed (emotional) speech recognition system is designed based on Vector-Quantization (VQ) approach using Generalized Lloyd algorithm. Frequency, Amplitude and Phase features are extracted from the Sinusoidal model of speech and these parameters are used for characterization of stressed speech. The stressed conditions considered in this study are anger, happiness, compassion and neutral. Data for analysis and classification of stressed speech are recorded from thirty speakers; each speaker uttered five statements for each emotion in two different languages, English and Telugu (an Indian language). The features of stressed speech signals are compared with the features of neutral speech signals. From the results, it has been observed that the sinusoidal model features can be used successfully for classification of stressed speech. A total of 320 stressed speech data files are used for training and 280 stressed speech data files are used for testing. The performance of the VQ classifier with sinusoidal model features has been tested under normal and noisy conditions. A maximum recognition of 92.8% is achieved with frequency features as the input to the VQ classifier.

다중 시계열 패턴인식을 이용한 반도체 생산장치의 지능형 감시시스템

본 논문에서는 다중 시계열 패턴인식 사용하여 생산장치의 상태자료부터 공정결과를 예측하여 정상 또는 비정상을 판정하는 지능형 감시시스템에 관하여 기술한다. 제안하는 감시스템은 초기화, 학습 그리고 인식의 세 단계로 구성된다. 초기화 단계에서는 감시대상의 생산장치가 가지는 인사들 각각의 가중치와 각 인자들이 가지는 시계열 자료 중에서 학습과 인식에 유효단계를 설정한다. 학습단계에서는 LBG알고리즘을 사용하여 이 생산장치에 의하여 생성되고 수집된 패턴들을 군집화 한다. 각 패턴은 시계열 형태의 자료와 처리 완료 후 계측기에 의하여 측정된 ACI로 구성된다. 인식단계에서는 DTW를 사용하여 실시간으로 입력된 패턴과 군집화된 패턴들 사이의 대응을 수행하여 가장 잘 정합되는 패턴을 찾는다. 다음은 이 패턴이 가지는 ACI, 차 그리고 가중치들의 조합으로 예측된 ACI 값을 산출한다. 최종적으로 예측된 ACI가 정상으로 수용할 수 있는 값 범위에 없는지 여부를 결정한다. 제안하는 시스템의 성능평가를 위하여 식각장치로부터 획득된 자료를 대상으로 실험하였다. 실험결과에서는 학습횟수가 증가함에 따라 예측 ACI값과 실측ACI값 사이의 오차가 현저히 감소함을 볼 수 있다 This paper describes an intelligent real-time monitoring system of a semiconductor processing equipment, which determines normal or not for a wafer in processing, using multiple time-series pattern recognition. The proposed system consists of three phases, initialization, learning and real-time prediction. The initialization phase sets the weights and tile effective steps for all parameters of a monitoring equipment. The learning phase clusters time series patterns, which are producted and fathered for processing wafers by the equipment, using LBG algorithm. Each pattern has an ACI which is measured by a tester at the end of a process The real-time prediction phase corresponds a time series entered by real-time with the clustered patterns using Dynamic Time Warping, and finds the best matched pattern. Then it calculates a predicted ACI from a combination of the ACI, the difference and the weights. Finally it determines Spec in or out for the wafer. The proposed system is tested on the data acquired from etching device. The results show that the error between the estimated ACI and the actual measurement ACI is remarkably reduced according to the number of learning increases.

Joint Matrix Quantization of Face Parameters and LPC Coefficients for Low Bit Rate Audiovisual Speech Coding

A key problem for videophony, that is telephony including the processing of images of the speaker's face in addition to acoustic speech, concerns signal compression for transmission. In such systems, audio and video compression are separately achieved by using both audio and video coders. In this paper, an audio-visual approach to this problem is considered, since we claim that the fundamental property of coherence (redundancy) between the two modalities of speech should be exploited by coding systems. We consider the framework of parametric analysis, modeling and synthesis of talking faces, which allows efficient representation of video information. Thus, we propose to jointly encode several face parameters, namely lip shape geometric descriptors, together with sets of audio coefficients, namely quite usual LPC parameters. The definition of an audiovisual distance between vectors of concatenated audio and video parameters allows to generate audiovisual single stage vector and matrix quantizers by using the generalized Lloyd algorithm. Calculation of video and audio mean distortion measures shows a significant gain in quantization accuracy and/or resolution compared to separate video and audio quantization. An alternative sub-optimal tree-like structure for audiovisual joint coding is also tested and yields interesting results while decreasing the computational complexity of the quantization process.

An Enhanced Generalized Lloyd Algorithm

Vector quantization is an effective means of data compression for it maps a set of real numbers into a single integer. In this letter an enhanced generalized Lloyd algorithm (GLA) for vector quantizer design is presented. Experiment results show that the proposed scheme outperforms the conventional GLA and recently reported modification of the K-means algorithm. Moreover, the achievement in coding performance is not accompanied by significant increase in computational complexity.

Adaptive vector quantization in SVD MIMO system backward link with limited number of active sub channels

This paper presents combination of Channel Optimized Vector Quantization based on LBG algorithm and sub channel power allocation for MIMO systems with Singular Value Decomposition and limited number of active sub channels. Proposed algorithm is designed to enable maximal throughput with bit error rate bellow some tar- get level in case of backward channel capacity limitation. Presence of errors effect in backward channel is also considered.

High-frequency error concealment in wavelet-based image coding

A new algorithm for high-frequency subband error concealment in wavelet-based picture coding is presented. It is based on a wavelet patch repetition approach: the LBG algorithm, given by Linde et al., is used to generate a codebook of patches and, according to a boundary distance measure, one of these patches is selected to mask the damaged area. Experiments show noteworthy results.