A Hidden Markov Modeling technique is used to evaluate the exact performance of the Viterbi algorithm for estimating states of a finite state system. This technique is applied to exact calculation of the bit error probability of the Viterbi decoder of a convolutional code in channels with independent errors and in channels with memory. The results are compared with the widely used upper bounds. The accuracy of the proposed methods is verified via computer simulation.

This paper describes the results of frequency-domain channel sounding in residential environments. It consists of detailed characterization of complex frequency responses of ultra-wideband (UWB) signals having a nominal center frequency of 5 GHz. A path loss model as well as a second-order autoregressive model is proposed for frequency response generation of the UWB indoor channel. Probability distributions of the model parameters for different locations are presented. Also, time-domain results such as root mean square delay spread and percent of captured power are presented.

This paper addresses the problem of training sequence design for multiple-antenna transmissions over quasi-static frequency-selective channels. To achieve the channel estimation minimum mean square error, the training sequences transmitted from the multiple antennas must have impulse-like auto correlation and zero cross correlation. We reduce the problem of designing multiple training sequences to the much easier and well-understood problem of designing a single training sequence with impulse-like auto correlation. To this end, we propose to encode the training symbols with a space-time code, that may be the same or different from the space-time code that encodes the information symbols. Optimal sequences do not exist for all training sequence lengths and constellation alphabets. We also propose a method to easily identify training sequences that belong to a standard 2/sup m/-PSK constellation for an arbitrary training sequence length and an arbitrary number of unknown channel taps. Performance bounds derived indicate that these sequences achieve near-optimum performance.

An algorithm for covert digital audio watermarking is presented. The method embeds a low-rate (in the neighborhood of 20–30 b/sec) watermark via perceptually insignificant long term multiband phase modulation, The watermark is very difficult to either discover or recover without the original (unmodulated) signal. The algorithm is shown to be robust to perceptual coding.

We propose a method to identify training sequences for multiple-antenna transmissions over quasi-static frequency-selective channels. These sequences are constructed to belong to a standard constant-amplitude 2/sup m/-PSK constellation (such as BPSK, QPSK etc) to simplify the transmitter/receiver implementation. Many practical systems use training of predetermined length. Optimal sequences do not exist for all training sequence lengths and constellation alphabets. The proposed method allows us to identify training sequences that belong to a standard constellation for an arbitrary training sequence length and an arbitrary number of unknown channel taps. Performance bounds derived indicate that these sequences achieve near-optimum performance.

The determinant and rank criteria used for space-time code design apply at high SNR. Code design metrics developed for low SNR assume a channel autocorrelation matrix with equal eigenvalues, which does not hold in many practical scenarios. This paper shows that a space-time code designed to ensure full diversity at high SNR can suffer significant degradation when implemented at low-to-medium SNR because of the channel autocorrelation profile. We examine the effect of the channel autocorrelation matrix on a space-time code's performance and discuss how knowledge of this matrix can be used for code design, particularly from the aspect of space-time trellis code minimum memory requirements. Our discussion applies to both flat-fading and frequency-selective channels that are treated in a unified manner.

This paper addresses the problem of training sequence design for multiple-antenna transmissions over quasi-static frequency-selective channels. As performance metric for channel estimation, mean square error is adopted. To achieve the minimum mean square error, the training sequences transmitted from the multiple antennas must have impulse-like autocorrelation and zero crosscorrelation. We reduce the problem of designing multiple training sequences to the much easier and well-understood problem of designing a single training sequence with impulse-like auto-correlation. To this end, we propose to encode the training sequences with a space-time code, that may be the same or different from the space-time code that encodes the information symbols. Designing one instead of multiple training sequences reduces the search space significantly and simplifies the construction of optimal or suboptimal training sequences.

This paper addresses the problem of soft equalization for space-time-coded transmissions over frequency-selective fading channels. The structure of the space-time code is embedded in the channel impulse response for efficient joint equalization and decoding. The proposed equalization/decoding approach uses a prefilter to concentrate the effective channel power in a small number of taps followed by a reduced-complexity maximum a posteriori probability (MAP) equalizer/decoder to produce soft decisions. The prefilter introduces residual intersymbol interference which degrades the performance of MAP when applied to the trellis of the shortened channel. However, the shape of the overall shortened channel impulse response allows the M-algorithm to approximate the prefiltered MAP performance with a small number of states. Based on this general framework, we investigate several enhancements such as using different prefilters for the forward and backward recursions, concatenating two trellis steps during decoding, and temporal oversampling. The performance is evaluated through simulations over the EDGE typical urban channel.

In a typical wireless system, communications between a transmitter-receiver pair is subject to impairments such as intracell and intercell interference, as well as multipath fading. For practical wireless data transmission systems, e.g., UMTS or GPRS, it has been demonstrated by simulation and supported by real measurements that these impairments can be adequately modeled by hidden Markov models (HMMs). It has also been demonstrated that various types of wireless data arrival process can be modeled by a batch Markov arrival process (BMAP). This paper presents analytical methods for evaluating the packet queue length and packet delay probability distributions assuming that packet arrivals are modeled by a BMAP and packets are transmitted over channels with bursts of errors which are modeled by HMMs. In contrast with simulations, the analytical approach allows a system designer to find and test proper diversity, source and channel coding schemes, and communication protocols more efficiently. Analytical and simulation results are compared to determine the accuracy of the presented methods.

We compare measured fading data with the popular Rayleigh model. Our results show that this model does not agree with the experimental data. As an alternative, we approximate the experimental data using autoregressive moving average (ARMA) models. To validate the models, we compare several characteristics of fading obtained analytically, by simulation, and from measured data.

We analyze the forward-backward algorithm for the maximum a posteriori (MAP) decoding of information transmitted over channels with memory and propose a suboptimal forward-only algorithm. We assume that both the information source and channel are described using hidden Markov models (HMMs). The algorithm lends itself to parallel implementation and pipelining. We apply the algorithm to MAP decoding of symbols which were trellis-code modulated and transmitted over channels with error bursts.

Abstract The complexity of channel equalization using the BCJR algorithm [Bahl LR et al. IEEE Transactions on Info. Theory 1974; 20: 284–287] grows exponentially with the channel memory. The M‐BCJR algorithm [Franz V, Anderson J. IEEE J. Selected Areas Comm. 1998; 16(2): 186–195] provides a method for reduced‐trellis channel equalization, but its performance varies significantly with the distribution of a channel energy to its taps. This paper proposes a variation of the M‐algorithm, based on the implementation of the BCJR algorithm in the logarithmic domain, which can offer robust performance. Designer choices such as the decision delay, maximum and minimum phase transformation, and selection of states are discussed. Simulation results demonstrate the effect of different parameters on the proposed algorithm's performance. Copyright © 2001 John Wiley & Sons, Ltd.

The expectation-maximization (EM) algorithm is popular in estimating the parameters of various statistical models. We consider applications of the EM algorithm to the maximum a posteriori (MAP) sequence decoding assuming that sources and channels are described by hidden Markov models (HMMs). The HMMs can accurately approximate a large variety of communication channels with memory and, in particular, wireless fading channels with noise. The direct maximization of the a posteriori probability (APP) is too complex. The EM algorithm allows us to obtain the MAP sequence estimation iteratively. Since each step of the EM algorithm increases the APP, the algorithm can improve the performance of any decoding procedure.

We derive a formula for calculating the throughput efficiency of the go-back-N (GBN) protocol when error sources in the forward and backward channels are modeled by a hidden Markov model. The result is presented in the matrix form which makes it applicable to models with any number of states. We consider also the relationship between bit-level and block-level models and obtain the probability distribution of the time that a message spends in the system.

Hidden Markov models (HMMs) are a powerful tool for modeling stochastic random processes. They are general enough to model with high accuracy a large variety of processes and are relatively simple allowing us to compute analytically many important parameters of the process which are very difficult to calculate for other models (such as complex Gaussian processes). Another advantage of using HMMs is the existence of powerful algorithms for fitting them to experimental data and approximating other processes. In this paper, we demonstrate that communication channel fading can be accurately modeled by HMMs, and we find closed-form solutions for the probability distribution of fade duration and the number of level crossings.

Hidden Markov models (HMMs) are popular in many applications, such as automatic speech recognition, control theory, biology, communication theory over channels with bursts of errors, queueing theory, and many others. Therefore, it is important to have robust and fast methods for fitting HMMs to experimental data (training). Standard statistical methods of maximum likelihood parameter estimation (such as Newton-Raphson, conjugate gradients, etc.) are not robust and are difficult to use for fitting HMMs with many parameters. On the other hand, the Baum-Welch algorithm is robust, but slow. In this paper, we present a parallel version of the Baum-Welch algorithm. We consider also unidirectional procedures which, in contrast with the well-known forward-backward algorithm, use an amount of memory that is independent of the observation sequence length.

A method for the automatic verification of on-line handwritten signatures using both global and local features as described. The global and local features capture various aspects of signature shape and dynamics of signature production. The authors demonstrate that with the addition to the global features of a local feature based on the signature likelihood obtained from hidden Markov models (HMM) the performance of signature verification improves significantly. The current version of the program, has 2.5% equal error rate. At the 1% false rejection (FR) point, the addition of the local information to the algorithm with only global features reduced the false acceptance (FA) rate from 13% to 5%.

It is well known that language models are effective for increasing accuracy of speech and handwriting recognizers, but large language models are often required to achieve low model perplexity (or entropy) and still have adequate language coverage.We study three efficient methods for stochastic language modeling in the context of the stochastic pattern recognition problem and give results of a comparative performance analysis.In addition we show that a method which combines two of these language modeling techniques yields even better performance than the best of the single techniques tested.

Hidden Markov model (HMM) based recognition of handwriting is now quite common, but the incorporation of HMM's into a complex stochastic language model for handwriting recognition is still in its infancy. We have taken advantage of developments in the speech processing field to build a more sophisticated handwriting recognition system. The pattern elements of the handwriting model are subcharacter stroke types modeled by HMMs. These HMMs are concatenated to form letter models, which are further embedded in a stochastic language model. In addition to better language modeling, we introduce new handwriting recognition features of various kinds. Some of these features have invariance properties, and some are segmental, covering a larger region of the input pattern. We have achieved a writer independent recognition rate of 94.5% on 3,823 unconstrained handwritten word samples from 18 writers covering a 32 word vocabulary.

Three methods for on-line signature verification are discussed in this paper. They are based on statistical models of features that summarize different aspects of signature shape and the dynamics of signature production. Two of the methods are based on the feature statistics of genuine signatures only. Of these two methods, the simpler one using a Euclidean distance error metric was found to have superior performance when tested on a database of 919 genuine signatures and 330 forgeries. Using a procedure for selecting the individual best 10 out of 22 features, the Euclidean distance method correctly classified 99.5% of the genuine signatures, while rejecting 86% of the forgeries. The third method uses statistical properties of the forgeries as well as the genuine signatures to develop a quadratic discriminant rule for classifying signatures. On the basis of the database used in this study, this method was generally better than the simpler Euclidean distance method. At the same 99.5% acceptance of the genuine signatures, the method rejected 90% of the forgeries.

A modified Baum-Welch algorithm is developed and applied to estimating parameters of error source models that belong to the class of hidden Markov models (HMM). Such models arise in the description of bursty error statistics in communication channels. A key element used repeatedly for estimating parameters of such models is the computation of the likelihood of given sequences of observations. Several recursive methods are available for efficiently computing this likelihood. However, even recursive methods can require prohibitive amounts of computation if the observation sequences are very long. Modifications of the Baum-Welch reestimation algorithm that significantly reduces the computational requirements when the observation sequences contain long stretches of identical observations are discussed. The algorithms are used here to estimate parameters of a binary error source model using the results of computer simulation.>

A general approach to error-source simulation in digital channels is discussed. It is shown that the stochastic sequential machine (SSM) model is general enough to describe error-source statistics in digital communication channels. The SSM model is capable of describing both the correlation between errors in a channel (bursts of errors) and the correlation between errors in different channels (crosstalk). Physical causes of error burstiness in satellite channels are considered, and it is shown that the SSM model can account for differential encoding, scrambling, forward error correction, intersymbol interference, nonlinear signal distortion, adjacent channel interference, modem defects, etc. SSM models of the satellite channel with differential and convolutional codes are simulated. The results of simulation agree with the experimental data for the actual channels. Analytical methods for determining the performance characteristics of SSM-modeled systems are developed.>