By Topic

Acoustics, Speech and Signal Processing, IEEE Transactions on

Issue 11 • Date Nov. 1989

Filter Results

Displaying Results 1 - 25 of 25
  • Comments on "A systolic array for computing BA/sup -1/

    Publication Year: 1989 , Page(s): 1786 - 1789
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (382 KB)  

    An algorithm and a systolic array for computing BA/sup -1/ was presented by P. Comon and Y. Robert (see ibid., vol.ASSP-35, p.717-23, June 1987). A and B are n by n and p by n matrices, respectively. Such an array computes BA/sup -1/ in (4n+p-2) time units using n(n+1) processing elements (PE's). The commenters apply a graph-based method for the design of systolic arrays to such an algorithm. They systematically derive the original array and another array that performs the computation in the same time but using (n(n+1)/2+pn) units. For p<(n+1)/2, the commenters' array exhibits throughput (n+1), high utilization of PEs, and (n+2p) I/O ports; the original array exhibits poorer performance for these measures.<> View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A frame-synchronous network search algorithm for connected word recognition

    Publication Year: 1989 , Page(s): 1649 - 1658
    Cited by:  Papers (45)  |  Patents (10)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (944 KB)  

    A description is given of an implementation of a novel frame-synchronous network search algorithm for recognizing continuous speech as a connected sequence of words according to a specified grammar. The algorithm, which has all the features of earlier methods, is inherently based on hidden Markov model (HMM) representations and is described in an easily understood, easily programmable manner. The new features of the algorithm include the capability of recording and determining (unique) word sequences corresponding to the several best paths to each grammar node, and the capability of efficiently incorporating a range of word and state duration scoring techniques directly into the forward search of the algorithm, thereby eliminating the need for a postprocessor as in previous implementations. It is also simple and straightforward to incorporate deterministic word transition rules and statistical constraints (probabilities) from a language model into the forward search of the algorithm View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A family of distortion measures based upon projection operation for robust speech recognition

    Publication Year: 1989 , Page(s): 1659 - 1671
    Cited by:  Papers (53)  |  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (972 KB)  

    Consideration is given to the formulation of speech similarity measures, a fundamental component in recognizer designs, that are robust to the change of ambient conditions. The authors focus on the speech cepstrum derived from linear prediction coefficients (the LPC cepstrum). By using some common models for noisy speech, they show analytically that additive white noise reduces the norm (length) of the LPC cepstral vectors. Empirical observations on the parameter histograms not only confirm the analytical results through the use of noise models but further reveal that at a given (global) signal-to-noise ratio (SNR), the norm reduction on cepstral vectors with larger norms is generally less than on vectors with smaller norms, and that lower order coefficients are more affected than higher order terms. In addition, it is found that the orientation (or direction) of the cepstral vector is less susceptible to noise perturbation than the vector norm. As a consequence of the above results, a family of distortion measures based on the projection between two cepstral vectors is proposed. The new measures have the same computational efficiency as the band-pass cepstral distortion measure View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Design of uniform DFT filter banks optimized for subband coding of speech

    Publication Year: 1989 , Page(s): 1672 - 1679
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (596 KB)  

    A new approach for designing uniform DFT analysis/synthesis filter banks, optimized for subband coding (SBC) of speech, is presented. A spectral-domain distortion measure, which consists of a weighted sum of error terms due to filtering, rate conversion, and quantization, is derived. The quantization is embedded by means of a statistical model. For the case of `fine' quantization, a simpler, deterministic distortion function is derived. The optimal filters are designed using an iterative algorithm in which two sets of linear equations are solved in each iteration, aiming at minimizing the distortion function. A 16-kb/s SBC is simulated using a filter bank designed by the new approach, and is found to achieve subjective and objective (SNR) performance similar to that of a conventional QMF-based SBC, with only about half the amount of computation View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Power normalized update algorithm for adaptive filters-without divisions

    Publication Year: 1989 , Page(s): 1782 - 1786
    Cited by:  Papers (8)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (448 KB)  

    Two power-normalized update algorithms for adaptive filters are described that do not require divisions. The algorithms are based on very simple iterative procedures that approximate divisions. Since neither the quotients nor the dividends change quickly, the division procedures need only be updated once each time the coefficients of the adaptive filter are updated. One algorithm requires three extra multiplications per coefficient and results in adaptive filters with convergence times practically identical to those achieved when actual divisions are used. The second algorithm requires only one multiplication per coefficient instead of a division, yet results in adaptive filter convergence times only slightly larger than those obtained with real divisions. Both algorithms result in adaptive filters only slightly more complex than steepest descent adaptive filters. The algorithms are especially applicable to real-time adaptive filters realized using single-chip digital signal processors or custom ASICs where hard-wired dividers are usually not available and fast convergence is still desired View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Silent and voiced/unvoiced/mixed excitation (four-way) classification of speech

    Publication Year: 1989 , Page(s): 1771 - 1774
    Cited by:  Papers (13)  |  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (336 KB)  

    An algorithm is presented for automatically classifying speech into four categories: silent and speech produced by three types of excitation, namely, voiced, unvoiced, and mixed (a combination of voiced and unvoiced). The algorithm uses two-channel (speech and electroglottogram) signal analysis and has been tested on data from six speakers (three male and three female), each speaking five sentences. An overall correct classification accuracy of approximately 98.2% was achieved when compared to skilled manual classification. This is superior to previously reported automatic classification schemes. If word boundary errors, including the beginning and ending of sentences, are excluded, then the algorithm's performance improves to 99.5% View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Error saturation effects in the LMS adaptive line enhancer-transient response

    Publication Year: 1989 , Page(s): 1776 - 1780
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (360 KB)  

    Digital implementation of the least-mean-square (LMS) adaptive line enhancer (ALE) introduces certain nonlinear effects. An investigation is conducted into the effects of feedback error signal saturation on the ALE adaptation. A set of nonlinear coupled difference equations is derived by projecting the mean weight vector onto a set of orthogonal basis functions. These equations are used to study the transient behavior of the ALE for the case of one and two sinusoids in broadband noise. Simulations are presented which support the results of the theoretical model View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Gabriel: a design environment for DSP

    Publication Year: 1989 , Page(s): 1751 - 1762
    Cited by:  Papers (45)  |  Patents (11)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1116 KB)  

    Gabriel is a software system intended to manage the complete development of real-time digital signal processing (DSP) applications, from conception and experimentation to implementation in real-time hardware. It performs non-real-time simulations as well as code synthesis for real-time hardware. It is intended to ease code development for architectures that are not easy targets for conventional compilers, such as multiprocessor systems built with very high-performance microcoded DSPs. The system is designed to be retargetable in two ways. First, it can synthesize code for a variety of multi-DSP architectures where the user specifies the salient features of the architecture. Second, it can target different DSPs. The authors have concentrated on code generation for the Motorola DSP56001, although code generation for the AT&T DSP32 has been demonstrated. At the highest level, an algorithm is described using a hierarchical block diagram. At the lowest level, the user can either simulate the algorithm locally on the workstation, simulate the target architecture running the generated code, or download the code into hardware and run it in real time. Gabriel is capable of handling multiple sample rates, iteration, and recurrences View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Asymptotic properties of the high-order Yule-Walker estimates of sinusoidal frequencies

    Publication Year: 1989 , Page(s): 1721 - 1734
    Cited by:  Papers (33)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (884 KB)  

    The asymptotic properties of the high-order Yule-Walker (HOYW) estimator of sinusoidal frequencies are analyzed. An explicit formula for the covariance matrix of the HOYW frequency estimation error is derived. The effects on estimation accuracy of the number of YW equations, the model order, and a certain weighting matrix are investigated analytically. The optimally weighted HOYW estimator is described. The analysis presented together with the already existing empirical evidence provides enough motivation for the use of an overdetermined and high-order Yule-Walker estimator View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The weighted redundancy transform

    Publication Year: 1989 , Page(s): 1687 - 1692
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (444 KB)  

    An algorithm is introduced for computing the multidimensional finite Fourier transform. The algorithm can be applied to data samples of any size. In most cases, it offers a substantial reduction in the computational complexity View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient computation of the discrete pseudo-Wigner distribution

    Publication Year: 1989 , Page(s): 1735 - 1742
    Cited by:  Papers (19)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (648 KB)  

    A description is given of a novel algorithm, the fast Fourier transform in part (FFTP), for the computation of the discrete pseudo-Wigner distribution (DPWD). The FFTP computes the cosine and sine parts of the discrete Fourier transform (DFT) separately by employing real inverse sinusoidal twiddle factors. Unlike the conventional methods which directly utilize the complex DFT, the FFTP yields real output since the DPWD is always real. In addition, the new method reduces the computational cost by making full use of symmetries and removing redundancies in the FFTP computation. The authors also describe a simple algorithm for computing the discrete Hilbert transform (DHT) to produce the nonaliased DPWD. A pipeline structure for real-time and a bulk processing technique for offline implementations of the method are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Highly parallel recursive/iterative Toeplitz eigenspace decomposition [array processing]

    Publication Year: 1989 , Page(s): 1765 - 1768
    Cited by:  Papers (11)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (368 KB)  

    A highly parallel algorithm is presented for the Hermitian Toeplitz eigenproblem. It computes recursively, in increasing order, the complete eigendecompositions of the successive submatrices contained in the original matrix. At each order, a number of independent, structurally identical, nonlinear problems is solved in parallel, facilitating fast implementation. The eigenvalues are found with a constrained iterative Newton scheme, and the eigenvectors are obtained by solving Toeplitz systems. In the multiple minimum eigenvalue case, eigenvector information found at the rank before is used to identify all except one of the eigenvectors associated with the multiple eigenvalue instantaneously. The final eigenvector is found by deflation. The performance of the algorithm is evaluated in terms of eigenpair accuracy View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A tight upper bound of the average absolute error in a constant step-size sign algorithm

    Publication Year: 1989 , Page(s): 1774 - 1776
    Cited by:  Papers (12)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (248 KB)  

    A direct performance index of the adaptive filtering sign algorithm (SA) is the average absolute error (AAE) at the output of the filter. Adopting this performance index, an easy analysis of SA is achieved under a weak assumption. It is proved, for both deterministic and random inputs to the filter, that the AAE has a tight upper bound that exceeds the minimum AAE by half the product of the step size and power of the filter input. The assumption used is existence of average squared and average absolute values of filter input signals. A practical interest of the result is that it provides a formula for the biggest step size as a function of tolerable adaptation-noise-to-desired-signal ratio View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Performance of high resolution frequencies estimation methods compared to the Cramer-Rao bounds

    Publication Year: 1989 , Page(s): 1703 - 1720
    Cited by:  Papers (48)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1168 KB)  

    An explicit expression for the Cramer-Rao bounds (CRBs) and a calculation of the statistical perturbation of the covariance matrix due to additive noise are presented. The results are applied to a statistical efficiency analysis of the main frequency estimation methods based on eigenvalue decomposition. For the covariance matrix, in order to characterize the perturbation of the signal subspace, only the component of the perturbation of the eigenvectors orthogonal to the subspace is considered. This gives a simpler and more significant form of the error covariance. The treatment includes the cases of forward-backward and moving averages. The CRB and estimation variances are calculated in the presence of additive random noise, but for a given set of amplitudes characterized by their sample covariance matrix. This approach is more realistic for the evaluation of efficiency in the small-sample case View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Wide sense stability of discrete- and continuous-time linear systems in the complex case

    Publication Year: 1989 , Page(s): 1768 - 1771
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (384 KB)  

    Canonical reflection coefficients are used to determine the number of roots appearing on the unit circle and their multiplicities for a stable polynomial (with no zero outside the unit circle). A criterion so that a stable polynomial has no multiple zeros on the unit circle is then deduced and transposed to the continuous-time linear systems in the complex case View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Discrete convolution by means of forward and backward modeling

    Publication Year: 1989 , Page(s): 1680 - 1686
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (472 KB)  

    An approach to discrete convolution is presented which obviates the zero assumption. The method is structurally similar to the method of J.P. Burg (Stanford Univ., May 1975), which estimates the autocorrelation coefficients of a series in a manner which does not require a predefinition of the behavior of the signal outside of the known interval. The basic principle of the present approach is that each term of the convolution is recursively determined from previous terms in a manner consistent with the optimal modeling of one signal in terms of the other. The recursion uses forward and backward modeling together with the algorithm of M. Morf et al. (ibid., vol.ASSP-25, p.429-43, Oct. 1977) for computation of the prediction error filter. The method is illustrated by application to the computation of the analytic signal and its derivative View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Concurrent algorithms for a class of 1-D and 2-D Wiener FIR filters with symmetrical impulse response

    Publication Year: 1989 , Page(s): 1780 - 1782
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (236 KB)  

    Consideration is given to the development of parallel algorithms for the design of mean-squared-error multichannel zero-phase smoothers and 2-D image noncausal models. The resulting algorithms require approximately O(p) additions and block multiplications, with p being the order of the corresponding filter or model. Efficient order-recursive Levinson-type and Schur-type algorithms are derived. The Schur-type algorithms exhibit a high degree of parallelism and can be performed on a linear array of O(p) processors in O( p) time units View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On implementation of the discrete Fourier transform-the STUSE algorithm for spectral estimation

    Publication Year: 1989 , Page(s): 1763 - 1765
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (284 KB)  

    The discrete short-time Fourier transform implemented as the STUSE (short-time unbiased spectrum estimation) algorithm, which includes both Fourier-domain coherent smoothing and power-domain incoherent smoothing, is examined in terms of its resolution capability in power spectrum estimation. An analysis of the STUSE algorithm in terms of the effective spectral window allows insight into how the influence of the finite window length on a spectral estimate can be lessened by linearly combining biased estimates. Comparison is made to the weighted overlapped segment averaging algorithm, which has been previously characterized as a special case of the STUSE algorithm View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive cosine transform coding of images in perceptual domain

    Publication Year: 1989 , Page(s): 1743 - 1750
    Cited by:  Papers (42)  |  Patents (17)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (620 KB)  

    An adaptive cosine transform coding scheme for color images which incorporates human visual properties into the coding scheme is described. It employs adaptive quantization to exploit the statistical nature of the coefficients and adaptive block distortion equalization to reduce the block edge structures inherent in block transform coding schemes. Results show that the subjective quality of the reconstructed images at a bit rate of 0.4 bit/pixel or a compression ratio of 60:1 is very good View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Speaker-independent phone recognition using hidden Markov models

    Publication Year: 1989 , Page(s): 1641 - 1648
    Cited by:  Papers (221)  |  Patents (18)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (752 KB)  

    Hidden Markov modeling is extended to speaker-independent phone recognition. Using multiple codebooks of various linear-predictive-coding (LPC) parameters and discrete hidden Markov models (HMMs) the authors obtain a speaker-independent phone recognition accuracy of 58.8-73.8% on the TIMIT database, depending on the type of acoustic and language models used. In comparison, the performance of expert spectrogram readers is only 69% without use of higher level knowledge. The authors introduce the co-occurrence smoothing algorithm, which enables accurate recognition even with very limited training data. Since the results were evaluated on a standard database, they can be used as benchmarks to evaluate future systems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Phase unwrapping of digital signals

    Publication Year: 1989 , Page(s): 1693 - 1702
    Cited by:  Papers (11)  |  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (784 KB)  

    In order to unwrap the phase of a digital signal, the signal's z-plane zeros, which lie close to the unit circle, are located by repeatedly expanding the z-plane in small steps and detecting a zero's crossing of the unit circle. This detection is accomplished by determining the step which gives rise to a maximum change in the phase at any frequency sampling point and across any segment of the unit circle. This method and its associated computer program have been rigorously tested, and the phase was successfully unwrapped in most cases, with the few exceptions involving sequences generated from high-density clusters of zeros View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multichannel 2-D AR spectrum estimation

    Publication Year: 1989 , Page(s): 1798 - 1800
    Cited by:  Papers (14)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (232 KB)  

    Spectral estimation for multiple 2-D signals by autoregressive modeling is discussed. The procedure computes the entire spectral matrix of autospectra and cross spectra for the set of 2-D signals. Specific differences between AR models for this problem and those for lower dimensional problems are discussed. Experimental results are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A direct algorithm for computing 2-D AR power spectrum estimates

    Publication Year: 1989 , Page(s): 1795 - 1798
    Cited by:  Papers (14)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (312 KB)  

    An algorithm for computing the parameters in a 2-D autoregressive spectral estimate without prior estimation of the correlation is described. The algorithm utilizes the multichannel form of the Burg algorithm and the relation between multichannel and 2-D AR modeling. The procedure permits computation of the spectral matrix for several channels of 2-D data; models with support in different quadrants are combined to form the spectral estimate View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Closed-form recursive estimation of MA coefficients using autocorrelations and third-order cumulants

    Publication Year: 1989 , Page(s): 1794 - 1795
    Cited by:  Papers (25)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (168 KB)  

    The authors derive a simple, recursive, closed-form algorithm for estimating the parameters of a moving-average (MA) model of known order, using only the autocorrelation and the 1-D diagonal slice of the third-order cumulant of its response to excitation by an unobservable, non-Gaussian, IID process. The output may be corrupted by zero-mean, nonskewed white noise of unknown variance. The autoregressive moving-average (ARMA) case is briefly discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Sensitivity considerations in state-space model-based harmonic retrieval methods

    Publication Year: 1989 , Page(s): 1789 - 1794
    Cited by:  Papers (7)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (476 KB)  

    State-space models are considered for estimating the frequencies of multiple sinusoids, and robust coordinate systems for frequency estimation are identified. It is shown that the ideal parameter set from a parameter sensitivity point of view involves estimating a unitary state transition matrix and then computing its eigenvalues to obtain an estimate of the frequencies. Procedures to estimate such matrices from covariance and time-series data are examined. It is shown that two state-space methods, the Toeplitz approximation method and direct data approximation, estimate robust state transition matrices View full abstract»

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

Aims & Scope

This Transactions ceased production in 1990. The current retitled publication is IEEE Transactions on Signal Processing.

Full Aims & Scope