Metropolitan area networks (MANs) are well suited to support integrated voice and data personal communication services (PCS) by interconnecting wireless base stations (BSs). Whereas data transport is best supported by queue-arbitrated (QA) access to the IEEE 802.6 MAN, several alternatives exist for voice transport, including pre-arbitrated (PA) access, bi-state PA access, QA access, and reservation-arbitrated (RA) access. The voice traffic capacity of these alternatives are compared under different data traffic load. The results indicate that RA access, QA access, and bi-state PA access are preferred under low, medium, and high data traffic load, respectively, as they facilitate statistical multiplexing to almost double the capacity of PA access.

Anticipated growth in demand for personal communications services suggests that current facilities will not be adequate for handling the associated growth in message and signalling traffic. A new distributed switch architecture, consisting of an IEEE 802.6 metropolitan area network (MAN) has been proposed, to interconnect wireless radio basesites with each other within a local region. Each MAN is then bridged to adjacent MANs, as well as to a backbone MAN for improved reliability and speed. Queue arbitrated (QA) access is provided to each MAN. A six-MAN network is simulated to determine the delay-throughput performance for both voice and data traffic, subject to voice traffic having priority over data. Of the 310 Mb/s capacity available, in the absence of signalling traffic, 290 Mb/s can be used, and provides source-to-destination transmission delays of less than 5 ms for up to 200 Mb/s of voice traffic.

A distributed circuit-switched approach for supporting the deployment of high speed wireless personal communication services in urban areas through the interconnection of base stations via metropolitan area networks (MANs) is proposed. Broadband MANs minimize traffic congestion by dynamic sharing of link capacity and by serving as distributed switches for partitioning call control functions. While the DQDB protocol readily supports distributed packet-switching over the IEEE 802.6 MAN, isochronous traffic such as voice and video is best supported by circuit-switched connections. The authors present an enhanced bi-state pre-arbitrated (PA) transport mechanism, and associated call control and handoff management techniques, which enable distributed circuit-switching over the MAN. These capabilities are not currently addressed in the 802.6 standards. The bi-state PA transport mechanism facilitates statistical multiplexing of variable rate isochronous traffic sources. The network capacity is constrained by the call setup delay performance, and is analyzed by simulations. Alternative signaling architectures, involving different placements of call control network elements, are evaluated. The effects of erasure nodes, and close bus versus open bus architectures, are considered. The overlap inter-MAN call setup procedure is proposed to reduce delays. Different call handoff procedures are formulated according to the type of handoff and the resulting change in call connectivity. Most handoffs are intra-MAN, requiring simple procedures with short delays.>

A distributed microcellular architecture based on the IEEE 802.6 Metropolitan Area Network (MAN) is proposed, and is shown to meet anticipated personal communications service (PCS) needs. A method is presented to calculate MAN coverage in urban areas, and is used to demonstrate coverage of approximately 50 city blocks per MAN. A distributed subscriber database architecture is proposed to facilitate call setup, tracking of roamers and handoffs. To fully utilize MAN bandwidth, a quick method for the heat stations to switch on/off isochronous slots is proposed to facilitate adaptation to PCS traffic level variations. Call setup and handoff procedures are detailed. The PCS signaling overhead is calculated to be 15% of the capacity required to carry voice traffic.>

An analytical procedure is developed to calculate delay and throughput performance of CSMA/PA (carrier sense multiple access with priority acknowledgments). Time-axis slotting and Markov chain analysis are used to determine the channel backlog per cycle. A cycle includes information packet access and transmission time followed by acknowledgment transmission or time-out. Throughput and delay performance are then calculated in terms of the number of network terminals, the network packet error probability, the packet generation rate, and the mean information packet retransmission delay. Fixed-length information and acknowledgment packets are assumed. Calculated results show excellent agreement with measured delay and throughput performance for two distinct five-node intrabuilding power line networks operating at data rates from 1.2 to 9.6 kb/s.>

Intrabuilding power lines provide a readily available and easily accessed network for personal computer (PC) communications within buildings. The use and performance of carrier-sense multiple access with priority acknowledgements (CSMA/PA) on such networks is described. The communication characteristics of intrabuilding power-line channels are summarized and found to suffer from highly variable and unpredictable signal-to-noise ratios and bit-error rates. The advantages of CSMA/PA relative to polling and token-passing access protocols are discussed. The architecture of a CSMA/PA implementation is described. The choice of the best data-link packet length is determined from throughput measurements to be approximately 1000 bits. Delay-throughput performance is measured at 1.2, 2.4, 4.8, and 9.6 kb/s using a phase-shift-keyed (PSK) spread-spectrum modem. Results are included for the potentially difficult problem of transmission across power phases. A discussion of results obtained and their implications for PC intrabuilding communications is also included.>

Notwithstanding much effort, explicit analytical results for average message delays for asymmetric token rings are currently unavailable. We develop an excellent and relatively simple analytic approximation which readily demonstrates the effects of the many system parameters on delay-throughput performance. The approximation applies to any cyclically served network (including rings, polling networks, and unidirectional fiber optic broadcast networks) with independent time-slot arrivals and gated service. The approximation becomes exact for symmetric loading. An equation for the moment-generating function of the joint terminal service times is obtained via an imbedded Markov chain. Two matrix equations are then derived, one for the mean terminal service times of theNqueues and another for the joint second central moments. A two-dimensional difference equation for the second central moments is then obtained. This difference equation together with Kleinrock's conservation law is used to approximate the mean delay at each queue. The peak approximation error is typically a few percent for various loading distributions including heavy or highly asymmetric loading, even if the statistics for message arrivals, message lengths, or walk times vary among the queues. Actual delay-throughput results are presented for various terminal loadings. Also included is a procedure to upperbound the error of our approximation for any message and walk time statistics, given the terminal utilizations.

The effects of channel transmission errors as well as packet collisions on delay-throughput performance are determined for ALOHA systems. The results are applied to the optimization of multichannel system throughput when retransmissions from bit errors prevent efficient utilization of a single wide-band power-limited channel.

Localized communication networks for office automation, security monitoring, environmental management of buildings, computer communications, and other applications enjoy every increasing demand. This paper considers communication of data and analog message signals over electric power distribution circuits for such applications, using pseudonoise (PN) modulation. Advantages of power distribution circuits include reasonably universal coverage and easy access via a standard wall plug. Disadvantages include limited communication bandwidth, relatively high noise levels, and varying levels of impedance, noise, and attenuation. PN (spread-spectrum) signaling provides immunity to narrow-band signal impairments and casual eavesdropping, multiplexing capability, addressing capability, and easily implemented error rate versus data rate tradeoffs. The primary disadvantage is synchronization delay and hardware cost normally associated with PN code synchronization at the receiver. Our prototype modem avoids these difficulties by using the 60 Hz power line frequency for carrier, code, and data-bit synchronization. Design issues discussed include processing gain optimization, modem impedance specification, carrier frequency selection, and PN code selection. Prototype modem circuitry costing less than $30.00 is described. Measurements on actual power distribution circuits indicate a bit-error probabilityp = 10^{-4}or less at data rateR_{B} = 6kbits/s for single users or atR_{B} = 60bits/s forM \simeq 22simultaneous users. Performance calculations for analog message signals are included.

Pseudonoise-modulated spread spectrum communication transceivers enable utilzation of broadcast communications facilities including satellite links, radio networks, and coaxial cable networks for point-to-point communications. In this paper, a bound on the acquisition time probability distribution is obtained by calculating the variance of the acquisition time and using Chebyshev's inequality. Recursive computer calculations are used to show that the hold-in time distribution is approximately exponential. The means and variances of the acquisition and hold-in times, together with the distribution estimate and bound, are determined in terms of the number of users and power-to(Gaussian) noise density ratio of the channel. These calculations pertain to a receiver used by Hopkins, which combines acquisition and tracking in a way which facilitates implementation. The utility of the distribution estimates in comparing and relating synchronization times, failure to achieve synchronization, message transmission times, and message error rates is illustrated for two data transmission protocols. The dependence of these quantities on the number of users, information bit rate, data block format, and power-to-noise ratio of the (Gaussian) transmission channel is also illustrated.

An adaptive differential speech encoder was assessed using subjective evaluation procedures. The coder's adaptive quantizer was similar to the one used by Cohn and Melsa [1] and the predictor involved nonadaptive previous-sample feedback. The digital channel used to transmit the quantizer output levels was assumed error-free. Paired comparison tests were used to obtain scaled isopreference contours on theN-fplane, whereNandfdenote, respectively, the number of quantizer output levels and the sampling rate relative to the Nyquist rate. These contours were used to determine the subjective signal-to-noise ratio vs.fandN, maximum subjective signal-signal-to-noise ratios vs. bit rate, optimum values offandN, and bit-rate savings which occur when entropy coding is used instead of natural coding of the quantizer output levels. Entropy coding yielded a bit rate approximately equal to threequarters that for natural coding and Nyquist-rate sampling minimized the bit rate in each case. Savings of from one to two bits occurred when ADPCM was compared with nonadaptive DPCM. The fact that our system was better than others forf \simeq 1.0but worse forf \gsim 2.0indicates the need to modify our quantizer adaptation algorithm as the sampling rate increases relative to the Nyqusit rate.

A new network decomposition-optimization algorithm is presented and evaluated via computer simulation. The network is first partitioned into subnetworks on the basis of lines fitted via linear regression to the node locations weighted by their traffic loads. Each node is then connected via a data line of appropriate capacity to a concentrator-multiplexer. The concentrators are then regarded as nodes, and the process is repeated as often as required. The resulting singly connected network consists of a hierarchy of concentrators whose number, capacity, location, and interconnection are selected as an inherent part of the design procedure to minimize network cost. The linear regression clustering (LRC) design procedure is evaluated via computer simulation by comparing the costs and performance of the resulting networks with that which results from use of an algorithm based on the generally applicable network design approach of iterative local optimizations or "search" procedures. The data supplied to the LRC and search algorithms include randomly generated node locations and traffic matrices, and specific (realistic) cost vs. capacity schedules for data lines and concentrators. Comparisons of network costs, queuing and transmission delays, network reliability and network design costs show the LRC algorithm to be markedly superior to the search algorithm. The paper includes a brief discussion of our results, and their implications.

The present study is a sequel to the authors' 1972 paper [1] where the output signal-to-noise ratio SNRois obtained for nonadaptive differential pulse-code modulation (DPCM) systems with an arbitrary linear predictor operating on noisy digital channels. Important stability constraints for an arbitrary linear predictor are obtained. A complete analytical study of previous line-line-and-sample feedback systems is presented. The SNRoimprovement over previoussample feedback is found to increase from approximately 2.6 dB for error-free channels to 4.1 dB for noisy channels. Optimization of the predictor for noisy channel usage is shown to greatly reduce the sensitivity of SNRoto variations in message and channel parameters, while use of the resulting predictor on error-free channels yields SNRovalues which are almost as good as those obtained when the predictor is optimized for error-free channels. Reduction of the effects of digital channel errors on SNRousing various methods, including periodic or pseudorandom resetting are considered briefly. Hardcopy results from computer-simulated DPCM monochrome image transmission systems corroborate our analytical results.

Cost-effective design of networks linking many remote terminals to a central computer (CPU) involves use of low-speed data lines to link geographically close terminals to concentrators. The concentrators are connected via high-speed data lines to the CPU. A design algorithm based on clustering of terminals followed by optimization of location, capacity and number of concentrators in each cluster is developed and evaluated. Evaluation is based on network designs for sets of 20 randomly (uniformly) generated locations of up to 500 terminals, with specific (realistic) cost versus capacity schedules being used for data lines and concentrators. In comparison with the popular Add algorithm, our linear regression clustering (LRC) algorithm has the following advantages: 1) the total cost of the concentrators, low-speed terminal lines, and high-speed CPU lines is typically 8 percent less; 2) the average transmission time delay at the terminals is typically 40 percent less; 3) the cost of adding low-speed data lines to connect additional terminals to concentrators in existing networks is typically 50 percent less; 4) the computational cost of design is typically 20 times less for 100-terminal networks and 150 times less for 500-terminals networks. Implications of the results and suggestions for further work are discussed.

Systems which locate and identify moving vehicles without the active cooperation of either the vehicle operator or its passengers are being considered for various applications, including identification and subsequent billing for vehicles passing toll booths, and location and identification of railroad locomotives. This paper considers the design of identification signals transmitted from (to) a moving vehicle to (from) roadside receivers (transmitters) at various known locations. The paper includes: 1) discussion of some of the important design factors; 2) consideration of modulation/demodulation schemes suitable for vehicle identification and location (VLI) systems; 3) analysis of inexpensive and easily implemented schemes for improving vehicle-detection accuracy by averaging repeated sequences of vehicle-identification symbols; 4) evaluation of binary codes for vehicle identification; 5) analysis and evaluation of schemes for synchronizing binary-coded VLI sequences; and 6) a design example illustrating concepts developed in the paper. It is shown that use of simple binary codes, including constant ratio or single parity check codes, in conjunction with signal averaging, often yields better performance at less cost than systems which employ complex coding/decoding schemes.

Sequential decision algorithms with on-line feature ordering and a limited look-ahead approximation are considered. The algorithms can be used with or without contextual constraints for multicategory pattern recognition problems. Computational complexity due to on-line ordering of features is analyzed and related to system performance. Computer simulated results are obtained using a standard data set (Munson's multiauthor handprinted character files) and a careful test procedure.

The number of black points in each of the 25 nonoverlapping square regions of a size-normalized character matrix were used to recognize the 3822 uppercase handprinted alphabetic characters from Munson's multiauthor data set. The recognition accuracy obtained using a Bayes' classifier, which assumes statistically independent features, compares favorably with earlier results obtained using recognition systems having complexity comparable to ours. Included are results and a recommendation regarding system evaluation procedures.

A closed-form analytical expression is obtained for the output signal-to-noise ratio (SNR0) of differential PCM systems operating on noisy digital channels. A procedure is then proposed for maximizing SNR0by joint optimization of the quantizer, predictor, and sampling rate. Several examples are considered, including one- and two-sample feedback systems excited by speech and video (Gaussian) signals. The predictor used in the DPCM system is linear and time invariant, but otherwise arbitrary. As with PCM, contributions to output signal distortion can be expressed as a sum of three separate terms resulting, respectively, from quantization errors, channel transmission errors, and mutual errors arising from interaction between quantization and channel errors. When a conventional optimum linear predictor is used, transmission errors are shown to be no more serious for DPCM than for PCM. The results show that SNR0for a well-designed DPCM system is considerably higher than for a well-designed PCM system operating on the same digital channel, even if the channel is noisy. In considering examples, particular attention is devoted to determining maximum values for SNR0and to examining the dependence of SNR0on predictor coefficient values, number of quantization levels, sampling rate, message statistics, and channel error probability. Implications of this dependence on system design are noted.

Isopreference tests were used to study the subjective effects of channel transmission errors on computer-simulated PCM and previous-sample feedback DPCM voice systems. Numerically scaled isopreference contours were obtained and plotted on planes having as axes the number of quantization bitsNand the bit error probabilitypof the binary symmetric transmission channel. For anyNandpconsidered, the quality of DPCM speech was found to exceed that of PCM speech. WithNfixed andpdecreasing, the improvement of DPCM relative to PCM increased to the point where quantization noise limited the performance of both systems. Two nonredundant codes, natural binary and folded binary, were used to encode the quantizer output. The two codes yielded comparable speech quality. The maximum subjective quality obtainable for white Gaussian and Rayleigh fading channels was determined as a function of channel capacity. Use of optimum channel encoding was shown to reduce considerably the channel capacity required to obtain a given speech quality. The subjective ratings were compared with measured system signal-to-noise ratios. Finally, the effect of the overload level of the quantizer on the system signal-to-noise ratio was determined.

Series representations, as well as exact expressions, are obtained for correlation functions and power density spectra of quantized Gaussian signals transmitted over discrete memoryless channels. The mean-square reconstruction error that results from linear filtering the received samples is expressed in terms of these power density spectra and is then examined.

A contour-tracing technique described in an earlier paper [9] was used along with monogram, bigram, and trigram contextual constraints to recognize handprinted characters. Error probabilities decreased by factors ranging from two to four over those which resulted when contextual constraints were ignored. A simple technique for searching over only the most probable bigrams and trigrams was used to significantly reduce the computation without reducing the recognition accuracy.

A contour-tracing technique originally devised by Clemens and Mason was modified and used with several different classifiers to recognize upper case handprinted characters. Recognition accuracies obtained compare favorably with other published results, particularly when additional simple tests are performed to differentiate commonly confused characters. One suboptimum classifier, in addition to yielding near optimum performance when tested on training data, uses much less statistical information than the optimum Bayes classifier and is significantly better than the optimum classifier when training and test data are limited.

The way in which messages are coded into sensory stimuli affects the speed and accuracy with which the messages are transmitted to a human. This paper describes experiments which were conducted in order to determine some of the properties of favorable coding schemes. The experiments showed that, in general, the stimuli should vary along many dimensions, and should assume one of two coordinate values along each dimension. When there are only two allowable coordinates per dimension, the probability of error is often independent of the actual coordinate values, provided the distance between the coordinates is well above threshold at low stimulus rates. The results of these experiments were used as a guide in coding letters of the alphabet into easily identifiable auditory and tactile stimuli. The reading rates for various coding schemes were then measured. In one code, each letter of the alphabet was represented by a unique auditory or tactile stimulus. After only12hours of practice, subjects were able to read English text chosen from several sources at a rate of20words per minute. Included in the paper is a description of a tactile-kinesthetic sensory display, and some suggestions for further research.

Approximate formulas for information transmitted by discrete communication channel obtained by averaging tight and weaker bounds respectively from transition probability matrices