Search

Although an extensive medical informatics research program as well as courses and training experiences in biomedical informatics have existed at the University of Washington (UW) for many years, a formal home did not exist until 1997 when the Division of Biomedical Informatics was created in the Department of Medical Education, School of Medicine. Since that time the expansion of the research, service and teaching programs has been rapid with a key milestone being a university commitment to provide funding, space and faculty to support the development of a new graduate program in Biomedical and Health Informatics. Hallmarks of the biomedical and health informatics program at the University of Washington include: - Strong shared belief that informatics research can contribute to the improvement of healthcare and health; - Large, multidisciplinary faculty including faculty from computer science, library and information science as well as the health sciences schools (dentistry, medicine, nursing, pharmacy, and public health and community medicine); - Comprehensive research and development partnership with the University of Washington Medical Centers information systems group and the UW Primary Care Network to move research from the laboratory to operational clinical systems; - Extensive and diverse regional setting in which to study information needs and develop informatics solutions in primary care settings; - Lack of barriers to interdisciplinary research and teaching.

This paper addresses the problem of adjacent channel interference between minimum-shift keying (MSK)-type signals transmitted through Rayleigh fading channels. A receiver, based on the multiple-input, multiple-output antenna diversity concept and the maximum-likelihood criterion, allows the transmission of MSK-type signals with very small channel spacings, greatly improving the bandwidth efficiency of these signals. The concept of antenna diversity is used to greatly reduce the effects of adjacent channel interference for channels with Rayleigh fading. The results show that it is possible to transmit bit-synchronized MSK-type signals with a very small normalized channel spacing (e.g., /spl Delta/fT/sub b/=/spl plusmn/0.4) between channels, under almost any power imbalance between signals, with very small degradation in performance.

Although an extensive medical informatics research program as well as courses and training experiences in biomedical informatics have existed at the University of Washington (UW) for many years, a formal home did not exist until 1997 when the Division of Biomedical Informatics was created in the Department of Medical Education, School of Medicine. Since that time the expansion of the research, service and teaching programs has been rapid with a key milestone being a university commitment to provide funding, space and faculty to support the development of a new graduate program in Biomedical and Health Informatics.Hallmarks of the biomedical and health informatics program at the University of Washington include:- Strong shared belief that informatics research can contribute to the improvement of healthcare and health;- Large, multidisciplinary faculty including faculty from computer science, library and information science as well as the health sciences schools (dentistry, medicine, nursing, pharmacy, and public health and community medicine);- Comprehensive research and development partnership with the University of Washington Medical Centers information systems group and the UW Primary Care Network to move research from the laboratory to operational clinical systems;- Extensive and diverse regional setting in which to study information needs and develop informatics solutions in primary care settings;- Lack of barriers to interdisciplinary research and teaching.

In this letter, the problem of adjacent channel interference (ACI) caused by the close packing of constant envelope MSK-type users in a given frequency band is considered. An optimal receiver filter, based on the theory of matched filtering, is found, and it serves as an upper bound on the signal-to-interference ratio. The intersymbol interference (ISI) caused by the time response of the matched filter is eliminated by a decision feedback equalizer (DFE) which, however, degrades performance. It was found that the matched-filter upper bound allows about 3-15 dB more ACI than the performance of a classical correlation detector (for additive white Gaussian noise only), depending on the frequency separation between channels. The DFE performance is only a little bit worse than that of the matched filter.

In modern voiceband data communication, the received signal is subject to nonlinear quantization noise due to companding. Under certain conditions, this quantization noise may become dominant and cause serious degradation in performance. In order to design better signal constellations for this environment, it is of interest to characterize this noise. We study the conditional probability density function (PDF) for the quantization noise found on a PCM companding channel. We obtain an analytical expression for the conditional PDF of the nonlinear quantization noise and an algorithm to compute the conditional PDF of the filtered noise, which is the actual noise at the slicer input of the linear receiver. We compare the results to a Gaussian PDF (with second-order moments that depend on the signal point) and conclude that a Gaussian approximation, although not entirely accurate, is not unreasonable.

Limiter-discriminator detection of continuous phase modulation (CPM) partial response signals is of special importance in systems in which phase coherence is very difficult to establish and/or maintain (e.g., mobile communications systems). The authors propose a Tomlinson-based approach which copes with the intersymbol interference (ISI) introduced by the baseband partial response frequency modulating pulses of the CPM schemes, (e.g., GTFM, GMSK, raised cosine (RC)). In this approach, the ISI equalization is divided between the transmitter and the receiver. The authors investigate, through simulation, the detectability performance and spectral properties of a number of CPM signals and receiver structures with special emphasis on the Tomlinson approach. It is shown that this novel application of the Tomlinson configuration allows the use of relatively simple receivers which exhibit almost the same performance as more complex Viterbi structured receivers. It is shown that a bandwidth-efficient narrow-band quaternary CPM-2RC modulation with either a Viterbi, a DFE or a Tomlinson receiver structure outperforms classical binary FSK (discriminator-integrate and dump detection) by approximately one db. It is pointed out in the paper that in general it is also possible to achieve detectability results similar to that of classical discriminator (integrate and dump) detection of binary CPFSK, for a number of bandwidth-efficient binary CPM schemes (e.g., GMSK) using the aforementioned receiver structures. >

The performance of a twisted-pair channel is assumed to be dominated by near-end crosstalk (NEXT) from other pairs in the same cable. Both intrabuilding local and central office loop channels can be modeled as NEXT-dominated channels. The capacity of this type of channel is found, using a Gaussian model. It is shown that the capacity is independent of the transmitted power spectral density. The results also indicate that present systems operate far below theoretical capacity. The capacity of a twisted-pair channel with both NEXT and white Gaussian noise present is also addressed. >

The maximum-likelihood sequence estimator (MLSE) for continuous phase modulation (CPM) signals in an additive white Gaussian noise (AWGN) channel is a very efficient method of detection. This paper describes an extension of a relatively simple crosstalk approach for the performance analysis of linear quadrature receivers with cochannel interference (CCI) and adjacent channel interference (ACI) present to the MLSE receiver. Many CPM signals are analyzed, including those using new baseband modulating pulses. One of the new schemes allows an ACI signal to be 62 dB greater than the desired user signal at a frequency separation of one-and-a-half times the bit rate, with just a 2-dB degradation in required E/sub b//N/sub 0/.

This paper presents the simulation results of a trellis code Proposed by Ozarow and Calderbank for the duobinary (1+D) signal. In this code two duobinary pulses are transmitted for each data bit, resulting in bandwidth expansion by a factor of two. The error results are compared to BPSK signaling, with transmitted signal levels of +/- 1. Both bit error and event error results are presented. The results demonstrate significant gains over both binary antipodal signaling, and uncoded duobinary signaling.

Analog modulation as a field of research has been dormant for about the past thirty years. This work presents some new results on the theoretical bounds on the performance of analog communications over Rayleigh fading channels. We consider both single-diversity and dual-diversity receivers. One of the main purposes of this paper is to try to renew interest in the field of analog communications.

This paper addresses the problem of adjacent channel interference between constant-envelope signals, e.g., MSK-type signals. A technique, based on the BLAST concept, is used to greatly reduce the effects of adjacent channel interference (ACI) and co-channel interference (CCI), for channels with Rayleigh fading. The results show that it is possible to transmit these signals with a very small spacing between channels, under almost any power imbalance between signals, with very small degradation in performance! For example, it is shown that for MSK-type signals it is possible to operate at a channel spacing of 0.4/T/sub b/ with very little degradation in detectability performance, due to ACI. All the benefits of antenna diversity are achieved with very little loss due to ACI.

The demand for high data rate communication systems and the lack of available bandwidth has led to much research effort into MIMO systems using multiple transmit and receive antennas. It has been shown that MIMO systems can improve the capacity of rich Rayleigh fading channels by a factor which is linearly proportional to the minimum number of transmit and receive antennas. The V-BLAST architecture has been proposed for realizing high spectral efficiencies over flat Rayleigh fading indoor wireless channels. Multitone transmission, or OFDM, has been used in a number of systems. OFDM eliminates the need for a complex equalizer for overcoming ISI caused by delay spread or frequency-selective fading. However, it has been shown that for a single-input-single-output (SISO) system operating over a frequency-selective channel, if the transmitter "knows" the channel characteristics, the "water-pouring" technique can be used to optimize the transmitted bit rate and approach the channel capacity. In this paper, we use the "water-pouring" approach for combined OFDM-MIMO systems, and show that it is possible to achieve very good bandwidth efficiencies for un-coded systems.

This paper takes as point of departure a novel dovetail technique, previously developed, for constructing families of spreading sequences mutually orthogonal in the asynchronous sense. Any two members of the family have periodic cross-correlation values which are identically zero for all values of the shift index, obviating the need for mutual synchronization. Other methods achieve orthogonality between members of a family of sequences at isolated values of shift or delay index, requiring synchronization between the sequences in order to maintain orthogonality, a difficult or impossible task for many users possibly in motion. In light of advances in technology we permit these sequences to be complex-valued with no uniformity constraint on the sequence element absolute values. We have examined communications using these families of sequences, it being well recognized that multiaccess interference (MAI) in SS/CDMA systems is a strongly limiting factor on the performance of these systems. We assume the performance is interference, rather than noise, limited. In order to facilitate comparisons with state of the art CDMA systems, for example IMT-2000, which use spreading factors of 4 512, we give various results for values of sequence length N, equal to 64, 128 and 256. Families of sequences, of size q, are constructed utilizing the dovetail construction guaranteeing that the real and imaginary parts of the periodic cross-correlations between pairs of sequences are identically zero for all shifts. The technique uses random pairwise interchanges, the so-called exchange shuffle, to generate the sequences. The q-1 cochannel interferers, CCIs, determine the probability of error in the communications. Even and odd cross-correlations are used in the simulations and in accordance with the prevailing assumption, the binary information bits form a sequence of independent, identically distributed, i.i.d., random variables each occurring with probability 1/2 . A constant random phase is associated with each cochannel (CCI) signal along with a randomly drawn shift index accounting for the asynchronous nature of the communications. For the majority of the results the voice activity factor (VAF) is 1, assuming all users transmit all the time, although voice communications are active only about 3/8 of the time. A brief glimpse of the effects of VAF < 1 is given to note the effects of lower activity. The simulations indicated that for equal energy sequences with N=256, in a typical family of size 64 the probability of error is of order of magnitude 10/sup -3/ while for a typical family of size 32 the probability of error is of order of magnitude 10/sup -6/. The latter value must have some uncertainty due to the very infrequent occurrence of errors. We noted from the results for N=256 and N=64 that q/N/sup 1/2 / = 2 yields a probability of error of order 10/sup -6/ and q/N/sup 1/2 / = 4 yields a probability of error of order 10/sup -3/. For Gaussian behavior one expects the same probability of error for equal values of (q-1) 1/2 /N 1/2 . An extension of the basic scheme is investigated where information is transmitted over both the real and imaginary channels, an attempt to increase the number of users. Results of this dual channel scheme are given and analyzed in the context of the single channel system and the presence of the signature sequence in the other channel. The PDF of the CCI is modeled and is found through simulations along with the probability of error. An assumption of Gaussian modeling, despite questionable independence, is applied. This is done through the variance as found by simulation with resulting probability of error. In addition a commonly applied Gaussian derived expression for probability of error is compared. For the cases taken it is found that the PDF is different from Gaussian and the probability of error as found by simulations is lower than predicted by both Gaussian based models. The results indicate that one

The ideal near-end crosstalk and thermal noise-limited throughput capabilities of uncoded decision feedback equalized (DFE) and orthogonally frequency division multiplexed (OFDM) PAM digital signals transmitted over nonloaded twisted metallic pairs bundled in a multicore transmission cable are determined. Though OFDM has recently loomed as a possible strong competitor among alternatives to DFE for high-speed data transmission over the digital subscriber loop network, it is revealed to be, at best, of marginal value, even if one allows an infinite number of transmission subchannels. >

It is theoretically shown that the capacity of a voiceband channel operating through a digital network is at least 56 kbps (for a 3.5 kHz baseband bandwidth). The capacity may actually be higher but how close one can come to 64 kbps, the limit imposed by the digital network itself, is unknown. The transmission method is based on generating nonuniformly pulsed PAM (pulse amplitude modulation) signals which at the network sampling instants go through one of the slicing levels of the A/D converter of the network. To accomplish this, both the slicing levels and the sampling clock times must be known. This is a hard practical problem and one can only conjecture if a reasonable solution will be possible. The signal may be recovered using a Viterbi detector to reduce intersymbol interference effects, but at present, the number of states required for this detector is very large. >

In modern voiceband data communication using T1 carrier systems, the received signal is subject to nonlinear quantization noise due to companding. In order to design better signal constellations for this environment, it is of interest to characterize this noise. An analytical expression for the probability density function (PDF) of the nonlinear quantization noise and a numerical result for the PDF of the filtered noise, which is the actual noise at the slicer input of the linear receiver, are obtained. The results are compared to a Gaussian PDF. It is concluded that a Gaussian approximation, although not entirely accurate, is not unreasonable. >

In modern voiceband data communication the received signal is subject to nonlinear quantization noise due to companding. Under certain conditions this quantization noise may become dominant and cause serious degradation in performance. We first calculate the cutoff rate and then use these results to design new "optimal" signal constellations for the companding channel. These new constellations are computed assuming a signal dependent non-circular Gaussian channel transition density.

The idea of using multitone signaling for wireless mobile communications has become increasingly popular in the past few years. This article describes the use of multitone signals and antenna combining in a system which is capable of "learning" the characteristics of the wireless transmission channel. The channel is assumed to be a frequency-selective channel which is slowly changing. By utilizing both the "water-pouring" technique of multitone signaling and antenna diversity combining, it is possible to maximize the bit rate through the channel. The new part of this system is the use of antenna diversity combining in this transmission scheme.

The Viterbi decoder is an efficient implementation of the maximum likelihood sequence estimator for continuous phase modulation (CPM) signals in an additive white Gaussian noise channel, therefore its performance analysis is of interest when co-channel and adjacent channel interference (ACI) are also present. Much research has recently been done in this field, towards the goal of finding more bandwidth-efficient CPM modulations for wireless personal communications (PCS). The work described in this paper gives a new and relatively simple approach for performance analysis. It uses the crosstalk approach which was used for analysis of simpler detectors for specific CPM cases, but broadens this approach substantially, to take into account more general CPM cases. The crosstalk approach when combined with an equivalent white Gaussian noise model gives very close approximations to previous analytic results, while requiring much less computer time. This work considers many CPM schemes in an adjacent channel interference and white Gaussian noise environment. The results are compared (when possible) to previous works, and many new CPM modulation schemes are also examined. Some of these new schemes give surprisingly good results, which should be of interest to those searching for bandwidth-efficient CPM signals, for use in PCS systems. One of the new schemes allows an ACI user to be 62 dB greater than the desired user at a frequency separation equal to 1.5 times the bit rate, with just a 2 dB degradation. >

Iterative decoding of a serially concatenated outer convolutional code and an inner continuous phase modulation (CPM) separated by a random interleaver is examined. This scheme, which is known to produce a high coding gain, is examined for various CPMs such as MSK, LRC, GMSK and TFM. CPM signals are constant envelope signals with very high bandwidth efficiency, providing good immunity to adjacent channel interference (ACI). The susceptibility of the serial concatenated system to ACI is examined. Simulations show that this configuration can tolerate much stronger ACI than previous CPM schemes, including those detected with the Viterbi algorithm.

A suboptimal receiver of constant envelope, continuousfrequency MSK-type modulations called Continuous Shift Keying (CSK) is proposed that halves demodulator complexity with no more than a 1.6 dB loss in detectability in AWGN. The technique is especially applicable to a frequency division multiple access system composed of many unsynchronized user terminals.

In many signal processing satellite applications, it is necessary to pack as many bit-time unsynchronized, power imbalanced digital FDM users as possible into the available bandwidth. In certain important cases the on-board demodulator (which simultaneously detects each user bit-stream individually) is not able to maintain phase reference. Phase comparison sinusoidal frequency shift keying (PCSFSK), a phase incoherent demodulation method with low adjacent-channel crosstalk properties, has been proposed as one way of meeting the requirements mentioned above. This paper describes a mathematical model and computer simulation which have been used to analyze the effects of adjacent-channel crosstalk on bit-error rates for a PCSFSK demodulator. Both the model and the simulation are flexible enough to incorporate many users of various strengths and center frequency separations and may be modified to yield results for any phase comparison MSK-type modulation. The results of this work show very close agreement between the model's prediction and the simulation. User scenarios of interest involve from two to eight users both with and without additive white Gaussian channel noise. Of particular interest is the performance of one weak user among many stronger users. A typical example involves eight users, seven of which are 20-dB stronger in power than the user whose probability of error was monitored. In this case, a bit-error rate of 10-3requires that adjacent users be separated in frequency by two times the data rate. If, however, all users are of equal strength, the minimum frequency separation need only be equal to the data rate. The above results indicate that close packing of signals is possible when incoherent detection (PCSFSK) is used, and that optimum user spacings can be determined if the relative power ratio between users is known.

Frequency division multiplexed, orthogonal M -ary FSK modulated signals transmitted over a frequency hopping channel (with one channel use per hop) interfere with each other unless perfectly synchronized in time and frequency with the receiver. In order to maintain the interfering "crosstalk" energy at harmless levels, even in the case of grossly power unbalanced users, the frequency separation between assigned signaling frequencies must be expanded beyond the orthogonal spacing. It is found that the required bandwidth expansion can be significantly reduced by amplitude windowing at the receiver, at a cost of a slight decrease in detectability. This effect is evident through examination of both worst case, single source crosstalk and average, multiple user crosstalk.

An analytical expression for evaluating the mean-square crosstalk between adjacent channels in coherent tamed frequency modulation (TFM) systems is presented. Numerical results are obtained and compared to previous crosstalk results for minimum shift keying (MSK); the substantial superiority of TFM is confirmed. An equivalent white Gaussian noise model for crosstalk is included and used to estimate performance degradation as a function of frequency separation between channels. The crosstalk results are also used to show that TFM is about 0.9 dB more sensitive to tone jamming than MSK.

The maximum bit rate of multitone QAM (quadrature amplitude modulation) over a general linear channel is found. First, the overall bit rate for an AWGN channel with a two-level transfer function is maximized, using a multitone QAM system. The power distribution between the tones and the number of bits/symbol per tone is optimized for a given symbol error rate. Extending these results to the general channel, it is shown that the optimum power division for multitone signals is similar to the water-pouring solution of information theory. Furthermore, multitone QAM performance is about 9 dB worse than the channel capacity, independent of the channel characteristics. The multitone results throughout are compared to those of an equivalent single-tone linearly equalized system. The comparison shows that the multitone system is useful for some channels, e.g. those with deep nulls in the transfer function. The maximum bit error rate over a twisted-pair channel which is performance dominated by near-end crosstalk (NEXT) is also found. >

A method is presented for comparing the performance of different quadrature-carrier modulation techniques. It is based on the calculation of the mean-square crosstalk between adjacent digital communications channels. This method was used to determine the crosstalk of a number of constant envelope modulation schemes, including Staggered Quaternary Phase Shift Keying (SQPSK) and Minimum Shift Keying (MSK). The results, as have previous results, show the improved performance of MSK over SQPSK. By utilizing the general equation for mean-square crosstalk as a measure of performance, a search was made for constant envelope modulation methods with reduced crosstalk. A number of methods were found which have better mean-square crosstalk performance than MSK. Finally this paper presents a method for reducing degradation due to crosstalk by using mismatched windows in the standard correlation detector. An example of this technique for detection of MSK signals is analyzed.

This paper deals with the optimization of coherent generalized tamed frequency modulation (GTFM). The optimization is carried out over certain parameters of the modulation scheme to give the best BER (bit error rate) performance, subject to a bandwidth constraint. The optimization is done by an algorithm which uses the upper bound on the squared Euclidean distance as a metric of the BER performance and the power spectra at a certain frequency as a metric of the bandwidth. It is concluded that when the modulation index is not restricted to 0.5, a theoretic improvement of 1.9 dB over the BER performance of GTFM is achievable, with equal bandwidths as a basis of comparison.

In-band generation of synchronous linear data signals eliminates the need for preliminary generation of a baseband signal and sharp cutoff filtering. However, the ratio of carrier frequency to pulsing rate must, in general, be an integer if the signal is to be produced by successive pulsing of a single fixed filter. As a consequence, a singlesideband signal of bandwidth W and transmission rate 2W must occupy the frequency band [ 2NW,(2N +1)W ], N an integer, when it is generated from a single fixed filter. A partial-response signal satisfying this condition is described in this paper. A further analysis demonstrates that a finite number of fixed filters can be used to produce a pulse train in any band where the ratio of carrier frequency to pulsing rate is a rational number. Since this relaxed condition allows any transmission band to be approximated to an arbitrary degree of accuracy, frequency translation can be avoided. A single transversal filter with periodically varied tap weights is shown to be a practical realization.

A QAM signal transmitted over a channel with linear distortion and additive white Gaussian noise can be linearly equalized at the receiver to eliminate intersymbol interference. If the QAM signal is power constrained and a given symbol error rate is required, we show that it is possible to maximize the bit rate of this system by optimizing the symbol rate and the number of bits/symbol. Ideal linear equalization is assumed at the receiver to overcome the distortion introduced by the channel. As an example, a Gaussian channel is chosen, and the bit rate is maximized for this channel. The QAM maximization is especially useful for channels with slowly decaying channel attenuation characteristics, e.g., the twisted-pair channel.

Continuous phase modulation (CPM) techniques for digital communication have been proposed to achieve narrow bandwidth and good bit error rate (BER) performance in coherent systems. These modulation schemes may also be used with noncoherent discriminator detection. However, in this case, the CPM schemes should be designed for noncoherent detection. Using a receiver proposed by Chung for GTFM and "modified GTFM" signals, we show that it is possible to slightly improve BER performance over that of previous GTFM noncoherent schemes. We also show that a form of GMSK can achieve discriminator detectability performance almost equal to that of classical binary FSK (modulation index = 0.7) but with a much narrower bandwidth.

An error to equation (25) in the above-named work (ibid., vol. COM-27, pp. 1519-1527, Oct. 1979) is corrected.