Scheduled System Maintenance:
On May 6th, system maintenance will take place from 8:00 AM - 12:00 PM ET (12:00 - 16:00 UTC). During this time, there may be intermittent impact on performance. We apologize for the inconvenience.
By Topic

Vision, Image and Signal Processing, IEE Proceedings -

Issue 6 • Date Dec 1994

Filter Results

Displaying Results 1 - 14 of 14
  • Top down structured parallelisation of embedded image processing applications

    Publication Year: 1994 , Page(s): 431 - 437
    Cited by:  Papers (6)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (536 KB)  

    The authors present a general system design method which is intended to support parallelisation of complete image processing applications using MIMD processors. The approach is based upon the utilisation of a generic system level parallel processor architecture, the `pipeline processor farm'(PPF), and is applicable to any embedded application with continuous input/output. The design method is illustrated using applications from the fields of computer vision and image coding. The design model accommodates several commonly exploited parallel processing paradigms, maps conveniently to the software structure of most image processing algorithms, provides incrementally scalable performance, and enables upper-bound speedups to be easily estimated from profiling data generated by the original sequential implementation of the application. It is believed that the approach has significant application in parallel embedded systems design, in the development environment, and in simulation work for computationally intensive image coding algorithms View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Use of multiple vector quantisation for semicontinuous-HMM speech recognition

    Publication Year: 1994 , Page(s): 391 - 396
    Cited by:  Papers (2)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (444 KB)  

    Although the continuous hidden Markov model (CHMM) technique seems to be the most flexible and complete tool for speech modelling. It is not always used for the implementation of speech recognition systems because of several problems related to training and computational complexity. Thus, other simpler types of HMMs, such as discrete (DHMM) or semicontinuous (SCHMM) models, are commonly utilised with very acceptable results. Also, the superiority of continuous models over these types of HMMs is not clear. The authors' group has previously introduced the multiple vector quantisation (MVQ) technique, the main feature of which is the use of one separated VQ codebook for each recognition unit. The MVQ technique applied to DHMM models generates a new HMM modelling (basic MVQ models) that allows incorporation into the recognition dynamics of the input sequence information wasted by the discrete models in the VQ process. The authors propose a new variant of HMM models that arises from the idea of applying MVQ to SCHMM models. These are SCMVQ-HMM (semicontinuous multiple vector quantisation HMM) models that use one VQ codebook per recognition unit and several quantisation candidates for each input vector. It is shown that SCMVQ modelling is formally the closest one to CHMM, although requiring even less computation than SCHMMs. After studying several implementation issues of the MVQ technique. Such as which type of probability density function should be used, the authors show the superiority of SCMVQ models over other types of HMM models such as DHMMs, SCHMMs or the basic MVQs View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural network-based F0 text-to-speech synthesiser for Mandarin

    Publication Year: 1994 , Page(s): 384 - 390
    Cited by:  Papers (2)  |  Patents (1)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (492 KB)  

    A neural-network-based approach to synthesising F0 information for Mandarin text-to-speech is discussed. The basic idea is to use neural networks to model the relationship between linguistic features. Extracted from input text and parameters representing the pitch contour of syllables. Two MLPs are used to separately synthesise the mean and shape of pitch contour, using different linguistic features. A large set of utterances is employed to train these MLPs using the well known back-propagation algorithm. Pronunciation rules for generating F0 information are automatically learned and implicitly memorised by the MLPs. In the synthesis, parameters representing the mean and shape of the pitch contour of each syllable are generated using linguistic features extracted from the given input text. Simulation results confirmed that this is a promising approach for F0 synthesis. The resulting synthesised pitch contours of syllables match well with their original counterparts. Average root mean square errors of 0.94 ms/frame and 1.00 ms/frame were achieved View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Nonlinear cepstral equalisation method for noisy speech recognition

    Publication Year: 1994 , Page(s): 397 - 402
    Cited by:  Papers (1)  |  Patents (1)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (388 KB)  

    The authors deal with the problem of automatic speech recognition in the presence of additive white noise. The effect of noise is modelled as an additive term to the power spectrum of the original clean speech. The cepstral coefficients of the noisy speech are then derived from this model. The reference cepstral vectors trained from clean speech are adapted to their appropriate noisy version to best fit the testing speech cepstral vector. The LPC coefficients, LPC derived cepstral coefficients, and the distance between test and reference, are all regarded as functions of the noise ratio (the spectral power ratio of noise to noisy speech). A gradient based algorithm is proposed to find the optimal noise ratio as well as the minimum distance between the test cepstral vector and the noise adapted reference. A recursive algorithm based on Levinson-Durbin recursion is proposed to simultaneously calculate the LPC coefficients and the derivatives of the LPC coefficients with respect to the noise ratio. The stability of the proposed adaptation algorithm is also addressed. Experiments on multispeaker (50 males and 50 females) isolated Mandarin digits recognition demonstrate remarkable performance improvements over noncompensated method under noisy environment. The results are also compared to the projection based approach, and experiments show that the proposed method is superior to the projection approach under a severe noisy environment View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 2D shape signature based on fractal measurements

    Publication Year: 1994 , Page(s): 422 - 430
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (540 KB)  

    The authors describe a novel scheme for 2D shape representation based on fractal geometry. The representation (or fractal signature) of the shape is an external scalar transform which contains information from a wide range of scales and is invariant under translation, rotation and scaling. The performance of the signature is evaluated in the presence of noise and occlusion. Use of the fractal signature for 2D shape matching is also demonstrated and compared with a more established technique based on Fourier shape descriptors View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Frequency sampling method for designing 2-D FIR real-coefficient digital filters using Hermite interpolation

    Publication Year: 1994 , Page(s): 365 - 372
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (532 KB)  

    A novel frequency-sampling method for designing 2-D real-coefficient FIR filters, given the values and slope estimates of the desired frequency response at each of the node points of a rectangular grid, is presented. Based on a new class of bivariate Hermite-type polynomials suitable for interpolating at complex conjugate points, and using Kronecker products, the original 2-D filter design problem is reduced to the solution of two 1-D systems of linear equations. Additional advantages of the method are the securing of the existence and uniqueness of the solution of the design problem, computational efficiency, the use of simple and recursive 1-D algorithms; the guarantee of real accurate results; and the inherent parallelism. The method is also applied to design 2-D symmetric FIR filters and can be extended to m-D design problems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Colour texture analysis using colour histogram

    Publication Year: 1994 , Page(s): 403 - 412
    Cited by:  Papers (8)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (720 KB)  

    The authors describe a new and systematic way of analysing colour texture. They propose to extract both spatial (i.e. texture) and spectral (i.e. colour) attributes separately from a colour image. This involves a parallel system of representation comprising two channels, one for texture representation and the other for colour description. The spatial attributes are gathered by the `local linear transform' for texture representation as it is very suitable for statistical texture representation. The use of the three-dimensional colour histogram as the model for colour representation is proposed. As the storage and feature computation from the colour histogram are both very intensive, the authors advocate and justify the approximation of this histogram by the product of the three first-order histograms computed for the principal axis components of the colour image pixel data. Illumination invariant colour features are then defined as statistical descriptors of these lower dimensional histograms View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generalised approach to parallelising image sequence coding algorithms

    Publication Year: 1994 , Page(s): 438 - 445
    Cited by:  Papers (2)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (560 KB)  

    The author describes the parallelisation of three different versions of the CCITT H.261 encoder algorithm using a generalised parallel design methodology based upon pipelines of processor farms (PPFs). For each algorithm, a theoretical upper-bound scaling model was derived by analysing the execution time profile of the algorithm and its feedback structure. The performance predicted by the model was, in each case, in good agreement with that achieved by the corresponding practical implementation. Practical throughput scaling up to a factor of 11 was achieved, using PPFs containing up to 16 processors. The design examples illustrate the impact which feedback has on potential speedup for image coding algorithms, and the diagnostic role of the model in identifying those algorithm components which restrict scaling performance. It is believed that the techniques presented may be useful both in developing embedded image coders based upon multiple DSP devices, and for simulation work with large image sequences in application areas such as image coding for HDTV and SHDTV View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Textured image segmentation by context enhanced clustering

    Publication Year: 1994 , Page(s): 413 - 421
    Cited by:  Papers (4)  |  Patents (1)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (740 KB)  

    An unsupervised textured image segmentation technique based on multidimensional feature vector clustering is described, where the features are the parameters of an autoregressive model, The benefits of incorporating spatial contextual information are demonstrated on both true cluster number estimation and actual image segmentation. A simple within-cluster distance is used for cluster validity analysis, where feature vectors are modified through local spatial dependency. This greatly reduces the dispersion in the raw feature data fed to the clustering process, and improves the true cluster number estimation. At the segmentation stage, three schemes incorporating contextual information at feature vector and label levels are proposed to enhance the segmentation accuracy. One is a development of a technique due to Mardia and Hainsworth (1988). The proposed approaches are tested on a four-class textured image View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Image sequence coding using quadtree-based block-matching motion compensation and classified vector quantisation

    Publication Year: 1994 , Page(s): 453 - 460
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (616 KB)  

    The authors propose a new image sequence coding algorithm based on two crucial methods: quadtree segmentation and classified vector quantisation (CVQ). Overall coding rates are efficiently lowered by quadtree segmentation while visual quality is well preserved by a CVQ method. A moving-block extraction technique is employed to greatly improve the coding efficiency in the interframe coding mode. A quadtree efficiently segments the stationary background regions of interframe differential signals with various large-sized blocks, and the moving regions are extracted from the smallest blocks of 4×4 size during the growth of the quadtree. These moving regions are motion-compensated using a block-matching method based on 4×4 blocks and the residual signals of the motion-compensated moving regions are coded by CVQ. The stationary regions are simply replenished from the previous frame. The proposed coding scheme is effective for coding the sequential signals of video telephony or video conferencing at low bit rates View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Nonlinear echo cancelling using look-up tables and Volterra systems

    Publication Year: 1994 , Page(s): 357 - 364
    Cited by:  Patents (2)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (516 KB)  

    Two classical approaches to nonlinear echo cancellation are to use look-up tables combined with FIR filters and Volterra systems; their practical interest is higher when adapted with LMS-type algorithms, because results are easily implementable and offer good tracking capabilities. A complete analysis of look-up-table-based schemes is presented, and a new one is introduced which offers near optimal characteristics but without requiring variable steps: in practice a piecewise-constant adaptation algorithm is enough. The theoretical equivalence of these schemes with those based on Volterra kernels is established; a comparison of operational characteristic is then made. The important option of applying growing Volterra structures is proposed for the first time, and its possibilities are verified (as well as the previous analysis) by means of simulation examples. The paper ends with suggestions to be explored to obtain more efficient nonlinear echo cancellers View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generalised block-matching motion estimation using quad-tree structured spatial decomposition

    Publication Year: 1994 , Page(s): 446 - 452
    Cited by:  Papers (12)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (524 KB)  

    A generalised block-matching motion estimation with variable-size blocks is proposed. The location and size of each block are determined from a quad-tree spatial decomposition algorithm. Experimental results with head-and-shoulders test image sequences show that motion-compensated errors with this method have lower entropy than the conventional fixed or variable-size block-matching techniques. The quality of the coded pictures under the proposed method with an H.261 codec, both subjectively and in terms of PSNR, outperforms the conventional use of fixed and variable block-size motion estimators View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive allpass filtering for nonminimum-phase system identification

    Publication Year: 1994 , Page(s): 373 - 379
    Cited by:  Papers (5)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (484 KB)  

    Efficient algorithms for the recursive identification of nonminimum-phase systems using an adaptive allpass filter together with an adaptive transversal filter are introduced. Both the output and input sequences are assumed to be available, and the proposed technique basically uses a linear predictor at the output of the system to equalise its magnitude response and an allpass filter to match its phase response. The mirror image property of the numerator and denominator polynomials in an allpass transfer function reduces the number of adaptive parameters needed compared to an unconstrained recursive identification scheme. Computer simulations are used to support the claims made in the paper View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Complex gradient and Hessian

    Publication Year: 1994 , Page(s): 380 - 383
    Cited by:  Papers (55)  |  Patents (2)
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (184 KB)  

    The gradient and Hessian are often used in analytical and numerical function optimisation complex valued parameter estimation problems. In a number of signal processing applications the function is a real function of complex variables. Then the optimisation is usually carried out with respect to the real and imaginary part of these variables; therefore, the gradient and Hessian concerned are real. The reason for this approach is to avoid difficulties with the definition and interpretation of the gradient and Hessian with respect to the complex variables. Definitions of a complex gradient and Hessian are proposed to solve these difficulties. The proposed and the real gradient and Hessian are fully compatible and are related by simple linear transformations. The results presented are an extension of a result by Brandwood (1983) concerning a complex gradient View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.