The ADM(adaptive delta modulation) speech coder is generally used with a time-invariant low-pass filter at the decoder output. The purpose of this low-pass filter is to reject coder noise at frequencies above the fixed speech passband. The speech spectrum, however, tends to occupy only the lower frequencies within the passband during voiced speech, and is somewhat “high pass” during unvoiced speech. In this paper, we show how the quality of ADM may be significantly improved by adoptively filtering the coder output such as to follow the natural bandwidth of the speech. This was found to reduce drastically the perceived noise in the output of the ADM coding system at low bit rates. The use of an adaptive low-pass filter realizes almost all of this quality gain. (An adaptive high-pass filter seems to reject less audible noise components and seems more prone to introducing objectionable artifacts.) We also discuss a method for reducing the bit rate with little or no sacrifice in quality (relative to normal ADM) by adapting the sampling rate along with the time-varying low-pass filter.

Assume that each item from a set of items is classified into one of several categories. We then use the proportion of items classified into a category to estimate the true proportion of items from the category in the population. This article models the effect of misclassification error on the estimate of the true proportion. We discuss two conditions which can be used to determine the adequacy of a classifier. We present an optimal classification algorithm which can be used when the joint distribution of the variables on which classifications are based is known separately for items from each category.

Simple frequency-division multiplexing (FDM) of two television signals onto a microwave radio channel is frequently unsatisfactory because of nonlinearity-induced crosstalk from one picture into the other. With time-frequency multiplexing (TFM), two successive scan lines (or fields) of one picture are frequency multiplexed so that they can be sent in one line (or field) period. During the next time interval, two successive lines (or fields) from the other picture are transmitted, thus avoiding crosstalk between pictures. To reduce the bandwidth required, one of the two simultaneously transmitted lines (or fields) is sent as an analog differential signal. In this paper, we discuss computer simulations of these techniques with application to a 20-MHz-bandwidth microwave radio channel. Effects of filtering and nonlinearities are included insofar as possible.

This paper reports on results that complement those in an earlier paper by this writer which gives a constructive proof of the existence of an algorithm that, for each right-hand side a, products a sequence which converges globally and superlinearly to a solution x of f(x) = a whenever f is a C <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> -diffeomorphism (i.e., is a continuoualy-differenliable invertible map with continuously-differentiable inverse) of a Banach space B onto itself and either B = R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</sup> or f satisfies certain other conditions that are often met in applications. Here we consider the case in which f is Lipschitz on each bounded subset of B. We give results which, while along the lines of those obtained earlier, concern a fundamentally different Newton-direction algorithm which does not appear to have been introduced previously, and which has the advantage that its implementation does not require the use of certain search procedures.

The International Telegraph and Telephone Consultative Committee (CCITT) has recently recommended two redundancy reduction codes for the digital transmission of two-level facsimile. It is of interest to transmit other types of still pictures using these codes. In this paper we develop a method for bilevel dithered pictures, wherein pictures are preprocessed reversibly to permit efficient coding by the CCITT codes. The preprocessing consists of local rearrangement of picture elements to obtain large contiguous black and white areas. The main advantage of these schemes over the existing ones is that they require the addition of a simple processor to a standardized facsimile coder. One of the schemes has the additional advantage that even in the absence of the postprocessor at the receiver, a distorted but recognizable picture is obtained. A compression ratio of 1.5 to 3.5 is obtained with these methods, which is comparable with other coding schemes.

In this paper we derive explicit expressions for the transient state probabilities of the Kendall birth-death process, with and without immigration, for any initial condition. We then propose this process as a model for special services point-to-point demand, in which the births represent circuit “connects” and the deaths represent “disconnects.” This choice of model is based on intuitive arguments and on the fact that the model can represent the growth and turnover characteristics of special services demand. Thus, the model provides a means by which special services demand, with its inherent uncertainty, may be approximately represented in various facility network studies, to obtain, at the very least, useful qualitative results. In particular, we evaluate the probability of a held order (ie., the probability that a service request is held for lack of spare facilities) with Blocked Customers Held (BCH) as the queue discipline. We also apply the model to capacity expansion problems, introduce the concept of margin, the extra capacity needed to meet the demand within a given held-order probability, and examine its sensitivity with respect to growth, turnover (or churn), and system size. We find that aggregating small demands into a single larger demand produces significant reduction of the margin, because of improved statistical properties.

The space planning process for central office switching, transmission, and power systems begins once the equipment need is forecast and the particular type of system is decided. This process is concerned with determining the kinds and quantities of the system equipment to meet ultimate forecast circuit requirements, devising a layout and growth plan for the determined equipment, and planning and coordinating the cable distribution, lighting, and cooling schemes associated with the equipment layout. The entire procedure of establishing a space utilization plan requires many iterations because of the large number of engineering aspects that must be examined in the decision-making process. To simplify, optimize, and streamline space planning for large equipment systems, various planning aids have been developed. We describe the general approach and rationale in developing two of the more fundamental space planning aids: office plan algorithms and office planning guidelines. Office plan algorithms allow a space planning engineer to translate long-range circuit forecasts into specific equipment frame requirements and their associated building space, cooling, and power needs. Office planning guidelines help the engineer to devise an efficient layout of the equipment in the building space, and to help plan and coordinate the cabling, lighting, and cooling schemes for the equipment layout. The general developmental approach of the space planning aids, and their usage benefits, are clarified relative to T1 offices that conform to the New Equipment-Building System (NEBS) standards.

In earlier papers, we have presented motion-compensated television coding schemes in which the displacement of objects was recursively estimated using a steepest descent algorithm that minimized the square of the intensity prediction error at each picture element. In this paper, we present extensions in which displacement is estimated by considering the prediction error at several picture elements. These extensions are more complex, but they significantly improve the performance of the displacement estimation in those cases where the displacement is spatially uniform. However, in real scenes containing large spatial variations of displacement, only a small improvement is obtained. For one scene containing a head-and-shoulders view of a person engaged in active conversation, an improvement of about 10 percent in average bit rate was obtained over our previous motion-compensation scheme.

The properties, fabrication and performance of planar (i.e., without an etched well), single heterojunction Ga <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1–x</inf> Al <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</inf> As:Te − GaAs:Ge fast LEDs, developed for use in optical data links, are discussed in this paper. The LEDs consist totally of epitaxially grown material. Current confinement achieved by contact area restriction increases current density and localizes electroluminescence for efficient coupling into a fiber. The peak wavelength of emission is at ≈900 nm with a full width at half-maximum of ≈50 nm. Electroluminescent rise and fall times are ≈3 ns each with pre-bias, making these LEDs adequate for data transmission rates as high as 100 Mb/s. At 700 A/cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , the LEDs launch 100 μW of optical power into a 0.36 numerical aperture (NA) and 125-μm core-diameter, step-index fiber or 10 μW into a 0.23 NA, 55-μm core-diameter, graded-index fiber. The LEDs are highly reliable with extrapolated mean time to failure of 2.7 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">7</sup> hours at junction current density of 700 A/cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and junction temperature of 74°C. These LEDs have been used as sources in prototype subsystems under development for use in electronic switching systems.