Knowledge about lip and glottal reflection coefficients during phonation is needed to eliminate their distortion effects on the estimates of vocal-tract area functions and glottal waves from vowel sounds. Direct measurements of these coefficients at human mouths are difficult. This paper presents a method for estimating them from vowel sounds. The estimation encounters an ill-defined inverse problem: the number of unknowns is greater than the number of constraints, and non-unique solutions exist for a sound. To overcome this problem, this paper uses a vowel sound produced by a subject whose vocal-tract area function (VTAF) for the sound is known. The estimates of the lip and the glottal reflection coefficients are determined as those that lead to a VTAF solution most similar to the known VTAF for the sound. The lip and the glottal reflection coefficients obtained for/a/and/i/ are presented.

Previously, estimating vocal-tract filters and glottal waves from vowel sounds imposed either the invalid assumption that glottal waves over closed glottal intervals are zero, or parametric models for glottal waves, resulting in biased vocal-tract-filter estimates and glottal-wave estimates lacking information over closed glottal intervals. We obtain unbiased vocal-tract-filter estimates from sustained vowel sounds, for which the glottal waveforms are periodically stationary random processes. Two equations are constructed each relating the vocal-tract filter to the sound signal and the glottal wave over one of two closed glottal intervals. By subtracting one equation from the other, the periodic components of the glottal wave are eliminated from the vocal-tract-filter estimation, and an unbiased vocal-tract-filter estimate is obtained. The average of many such estimates from different closed glottal intervals of the sound is the final estimate, which is used to obtain the glottal wave by inverse filtering the sound. The results obtained from vowel sounds /a/ produced by some subjects are presented. Over closed glottal phases, the glottal waves obtained are nonzero. During vocal-fold colliding, they increase; during vocal-fold parting, they decrease or even increase. The vocal-tract filters obtained yield vocal-tract area functions similar to that measured from an unknown subject's magnetic resonance image.

High-resolution vocal-tract area functions (VTAF) can be derived from vocal-tract filters (VTF) estimated from vowel sound signals with a wide bandwidth. However, the effects of open glottises and frequency-dependent lip reflection coefficients contained in the VTF estimates distort the VTAF estimates. Given VTF estimates obtained over closed glottal phases, we provide a method for eliminating the distortion effects of frequency-dependent lip reflection coefficients on the VTAF estimates. When the VTF estimates contain limited effects of incomplete glottal closures, this method can still obtain reasonable VTAF estimates if the vowel sounds are produced with large lip openings. The VTAF estimates obtained using our method from sounds /a/ produced by different subjects are very similar to that measured using the magnetic resonate imaging method. Theoretically, to eliminate both distortions caused by lip reflection coefficients and incomplete glottal closures in the VTAF estimates, lip-opening areas must be known.

Existing methods that estimate the vocal-tract area functions (VTAF) from vocal-tract filters (VTF) using speech signals suffer from inadequate elimination of the glottal wave, and the influence of non-ideal vocal-tract boundary conditions. To minimize these effects on the VTF estimation, we present a method that jointly estimates the glottal wave and the VTF corresponding to closed glottal phases. Experimental results show that our method yields better estimates. The VTAF obtained for /a/ and /i/ each produced by a female and a male subject show that more detailed and accurate VTAF are obtained using the speech signals corresponding to closed glottal phases.

This paper shows that the chord principle, used to couple chords one-to-one with phonemes, can be applied to the design of a new and powerful keyboard for the stenograph. The new keyboard is called the minimum chord stenograph (MCS). The training time is quite short. Within the limits imposed by no more than 50 hours' training, MCS' input data rates are superior to those from the old stenograph. One can definitely state that MCS requires a modest time to learn the code and, in return, gives a means for rapid entry of phoneme-based speech. A review is given of relevant features of the standard typewriter (QWERTY), chord keyboards, and the normal stenograph. Detailed insight into the operator-and-machine as a unit is provided by interval measurements t/sub p/(r) taken directly from tests when MCS is being operated. t/sub p/(r) is the time measured from the end of one chord and the end of the next chord r. A pseudochannel capacity C/sub p/(r) is used to show that MCS' performance gets better as training time is increased. Although QWERTY works with alphabet letters as input and MCS works with phonemes as input an attempt is made to compare QWERTY and MCS in terms of "words per minute". The two methods are not very different in time to acquire a first competence, but MCS's top speed after 50 hours of training is a little better than QWERTY's speed for operators all of whom had at least a year of training and usage.>

Bayesian networks provide a natural, concise knowledge representation method for building knowledge‐based systems under uncertainty. We consider domains representable by general but sparse networks and characterized by incremental evidence where the probabilistic knowledge can be captured once and used for multiple cases. Current Bayesian net representations do not consider structure in the domain and lump all variables into ahomogeneousnetwork. In practice, one often directs attention to only part of the network within a period of time; i.e., there is “localization” of queries and evidence. In such case, propagating evidence through a homogeneous network is inefficient since the entire network has to be updated each time. This paper derives reasonable constraints, which can often be easily satisfied, that enable a natural{localization preserving)partition of a domain and its representation by separate Bayesian subnets. The subnets are transformed into a set of permanent junction trees such that evidential reasoning takes place at only one of them at a time; and marginal probabilities obtained are identical to those that would be obtained from the homogeneous network. We show how to swap in a new junction tree, and absorb previously acquired evidence. Although the overall system can be large, computational requirements are governed by the size of one junction tree.

J.G. Daugman's (1988) neural network solution to the Gabor expansion of an image is reformulated as a steepest descent implementation. Nonlinear optimization theory is then applied to select an appropriate convergence factor. Two quasi-Newton-based nonlinear optimization techniques are applied to improve the convergence for certain types of lattice.

The role of the parametric stage is studied under various conditions and the following points have been demonstrated: -With high sampling rate (200 Hz) but otherwise favourable conditions, as the filter order, p, is increased from zero to nineteen the signal-to-noise ratio at the output of the parametric stage remains the same as at the input. Under less favourable conditions, it can fall as p is increased. -We find that comparable performance (in terms of spikes detected) is obtained when the parametric stage is omitted entirely, and detection is based only on the very simple non-parametric stage.

In this paper, some aspects of the implementation of Wigner Distribution (WD) are discussed. In general, it is necessary to smooth the WD in order to study the underlying nature of the signal. It is shown that the WD smoothed for positivity is equivalent to a spectrogram in terms of time and frequency resolutions. An interpretation of the partially smoothed WD is presented. A FORTRAN subroutine to compute the smoothed WD (SWD) is included. The choice of smoothing parameters for partial smoothing and spectral wrap in the WD are discussed.

This paper comments on the optimality of the Laplacian of a Gaussian edge detection filter which localizes edges through zero crossings in the filtered image. The arguments of both Marr and Hildreth, and Dickey and Shanmugam are reviewed to establish that the filter is optimal in the sense of maximizing output image energy near edge features. This filter's principal advantage over other edge detectors is that its response is user-adjustable through selection of a single parameter, the Gaussian standard deviation. However, no clear method for the selection of this parameter has been provided. The problem is addressed here by applying the filter to two ideal periodic edge models blurred by a Gaussian distribution point-spread function. The observed response to the edge spacing and blur standard deviation is then translated into a filter parameter design procedure. The problems of optimum filter performance in the presence of additive Gaussian noise are then addressed. The problem of selecting the sampled filter's coefficient word size is dealt with in a companion paper.

A companion paper describes the design and implementation of the Laplacian of a Gaussian edge detection filter which localizes edges through zero crossings in the filtered image. Accuracy in the presence of noise has been found to be proportional to the square root of the filter's standard deviation. Digital implementation of any continuous filter requires sampling and coefficient quantization. The sampled filter was examined, but a method is proposed here for selection of a minimum coefficient word size for direct-form implementation to satisfy in-band rejection bounds.

The optimal V2G edge detector of D. Marr and E. Hildreth is subjected to the Kitchen-Rosenfeld edge-detector-evaluation method. The ∇∇∇∇2G edge filter was designed to minimize noise and produce continous line-like edges not requiring further thresholding and relaxation. The threshold selection methods of Kitchen and Rosenfeld are applied to judge the value of post processing. Furthermore, the question of whether this optimal filter offers improved performance over prior edge detection methods is addressed. When the results are compared to those compiled by Kitchen and Rosenfeld a definite improvement in performance is indeed found.

Parametric methods which will detect spikes in EEG signal have been suggested by many authors, and encouraging results seem to have been obtained. A main limitation is instrument complexity and most of the results have been obtained relatively slowly off line. The basic theoretical limitations of the approach apart from the factor of instrumentation do not seem to be fully recognized. In this paper we show\bulletthat the parametric method is not optimum in the Weiner sense.\bulletthat anamolous indications of spike activity are given when p, the number of filter parameters, is varied.\bulletthat the mean square error is not an indicator of performance.\bulletthat low p values, perhaps surprisingly, lead to better spike detection than high values. Non-parametric methods have been suggested by several authors and they show that significantly less instrumentation is required and on-line results can be obtained in real time using modest equipment. A 16-channel real-time spike monitor has been realized on an Apple II computer using a method attributed to Ninomija et al. The basic approach is to detect incidence of shapes, all of which could be actual spikes, and then in subsequent filtering stages remove those shapes which do not meet the criteria spelled out by the neurosurgeon.

An important class of nonlinear devices is that based on the\nuth-law characteristic. Mathematical methods already employed in the analysis of this class of nonlinearities can be exploited in a rather straightforward manner to obtain ready-made formulas for autocorrelation functions of arbitrary nonlinear characteristics in response to noise plus signal.

A television system is proposed which sends a succession of high-bandwidth and low-bandwidth pictures over a channel. If flicker is avoided, the eye judges picture quality from successive presentations of the pictures by taking an average. It appears that an exchange can be effected between the power in the high-frequency components and the actual duration of these components. Using only successive high-bandwidth and low-bandwidth pictures, a 1.5:1 compression ratio can be achieved, accompanied by a small drop in subjective quality. Using a "high-frequency boost circuit" to get a compression ratio up to 2.75:1 still gives satisfactory picture quality. Experiments using movie pictures in place of the previously mentioned stills show that high compression ratios are possible.

The perception of flicker by the human eye depends upon the picture content, and this paper describes a relation between the critical flicker frequencyf_cand the spatial frequencyf_x. Provided that the visual picture is presented in such a way that each spatial component, presented at a frequencyf_p, obeys the inequalityf_p \geq f_c = f(f_x) = f_0 \exp (-k f_x), no flicker effect is perceived. The constantsf_0andkare weakly dependent upon the brightness and picture area; they are a little more strongly dependent upon the contrast ratio and the type of picture. The effect can be exploited in motion pictures or television by presenting high-quality pictures interleaved with low-quality ones. In the latter, some high spatial frequencies would be omitted. For television, band reduction of a face by 9:1 is possible. Practical consideration may reduce this compression ratio to a little short of 2:1.