In this paper we present our research activities in the area of digital audio processing and transmission. Today's available teleconference audio solutions are lacking in flexibility, robustness and fidelity. There was a need for enhancing the quality of audio for IP-based applications to guarantee optimal services under varying conditions. Multiple tests and user evaluations have shown that a reliable audio communication toolkit is essential for any teleconference application. This paper summarizes our research activities and gives an overview of developed applications. In a first step the parameters, which influence the audio quality, were evaluated. All of these parameters have to be optimized in order to result into the best achievable quality. Therefore it was necessary to enhance existing schemes or develop new methods. Applications were developed for Internet-Telephony, broadcast of live music and spatial audio for Virtual Reality environments. This paper describes these applications and issues of delivering high quality digital audio services over lossy IP networks.

Previous studies on totally mobile wireless networks (TMWN) have been limited to non-hierarchical architectures. In this paper, we study a two-tier cellular architecture for TMWN. Under the constraints of equal power consumption, the two tier system achieves improvement over the one-tier system, especially at light and medium load levels. Performance tests have also shown that handoff prioritization can be achieved by restricting the use of the umbrella channels. Further improvement for the two-tier system was obtained by load balancing strategies with respect to the allocation of channels to the different cells.

It is now empirically documented that data traffic over networks of various types exhibit fractal or self-similar behavior in many instances. Accurate analysis of traffic density and estimation of buffer size must take into account this self-similar nature. Researchers have investigated procedures for generating self-similar signals to model the traffic. Approaches based on the discrete wavelet transform (DWT) are among those that have been proposed. The basis for using the DWT is that it possesses certain scale-invariance properties and scale-invariance provides the foundation for characterizing self-similarity. However, self-similar processes generated with the DWT demonstrates self- invariance to dyadic scaling factors. Zhao and Rao have proposed novel models for purely discrete-time self-similar processes and linear scale-invariant (LSI) systems based on a new interpretation of the discrete-time scaling (equivalently dilation or contraction) operation which is defined through a mapping between discrete and continuous time. They show that it is possible to have continuous scaling factors through this operation although the signal itself is discrete-time. In this paper, we demonstrate application of these LSI systems to the synthesis of data whose self-similarity parameters match those observed in network traffic. Both theoretical development and experimental results are provided.

Third-generation (3G) wireless networks will support integrated multimedia services based on a cellular extension of a packet-switched architecture using variants of the Internet protocol (IP). Services can be categorized as real- time and delay-sensitive, or non-real-time and delay- insensitive. Each call, arriving to or active within the network, carries demand for one or more services in parallel; each service type with a guaranteed quality of service (QoS). Admission of new calls to the wireless IP network (WIN) from the gateway of a wired network or from a mobile subscriber (MS) is allowed by call admission control procedures. Roaming of the MSs among the nodes of the WIN is controlled by handoff procedures between base stations (BSs), or BS controllers, and the MSs. Metrics such as the probabilities of call blocking and dropping, handoff transition time, processing latency of a call, throughput, and capacity are used to evaluate the performance of network control procedures. The metrics are directly related to the network resources required to provide the QoS for the integrated services.

This paper outlines issues involved in improving the security, survivability and availability of wireless networks. DoD is already investing in the protection of its wired information systems. This paper will outline, from a protocol layer perspective, the unique aspects of extending information assurance to wireless systems, address the Information Assurance/Protection issues associated with software programmable radios (i.e. the Joint Tactical Radio System), and describe a recently initiated project at AFRL in this area.

In a recent paper we have shown that a simple turbo coded modulation scheme provides significant performance improvement with respect to the properly interleaved space- time trellis codes for block Rayleigh fading channels. The turbo coded modulation scheme for the multiple antenna system is based on a simple encoding procedure which uses the encoder for a binary turbo code and maps the encoded bits to a certain constellation, and a simple decoding procedure which is based on a log-likelihood computation of the transmitted bits and the use of these log-likelihoods in a sub-optimal iterative decoding algorithm. Although the computation of log-likelihoods is not a problem for small number of transmit antennas, the computational complexity at the receiver is too high for the case of large number of transmit antennas. In this paper we present a practical method for handling this problem. In particular, we present a new decoding algorithm which does not suffer from the same computational complexity. Our method is similar to the array processing/group interference suppression scheme developed for space-time trellis coding. We present examples for the case of four transmit-four receive antennas and for the case of eight transmit-eight receiver antennas over block fading channels. We observe gains of around 3 dB at a bit error rate of 10^-5 over the multi-layered space-time trellis coding with symbol interleaving. We also observe that although the current approach is computationally efficient, it results in a considerable performance degradation compared to the original (more complex) approach.

The realization of highly bandwidth-efficient wireless communications over frequency-selective multipath fading channels will be one of the key enabling technologies for future high-speed communication systems that offer ubiquitous access to information. Such bandwidth-efficient communication over broadband wireless channels can be accomplished by a number of different coding, modulation, and framing strategies; in particular, (1) the system can employ a single-carrier or a multicarrier framework, (2) the system can employ adaptive or non-adaptive signaling techniques, and (3) the system can employ iterative or traditional non-iterative decoding. The goal of this ongoing project is to understand the similarities, differences, and applicability of these various techniques in high-speed wireless communications. Such as investigation services not only the specific system design goals here but also aids in the consideration of a number of open issues in the communication theory literature, such as the applicability of adaptive techniques in systems that employ iterative decoding and the robustness to channel uncertainty of iterative single-carrier techniques typically employed in magnetic recording. In this paper, preliminary numerical results from this study are presented for various logical combinations of the aforementioned strategies motivated by previous work of the authors. These numerical results, combined with a complexity and delay analysis of the various receivers, allows the comparison and contrast of the aforementioned techniques for various applications.

The symbol-by-symbol maximum a posteriori (MAP) known also as BCJR algorithm is described. The logarithmic versions of the MAP algorithm, namely, Log-MAP and Max-Log-MAP decoding algorithms along with a new Simplified-Log-MAP algorithm, are presented here. Their bit error rate performance and computational complexity of these algorithms are compared. A new hardware architecture for implementing the MAP-based decoding algorithms suitable for chip design is also presented here.

This work presents modified variants, in a recursive format, of the Kahaner's additive fast Fourier transform algorithm. The variants are presented in Kronecker products algebra language. The language serves as a tool for the analysis, design, modification and implementation of the FFT variants on re-configurable field programmable gate array computational structures. The target for these computational structures are discrete Fourier transform beamforming algorithms for space-time-frequency applications in wireless.

The Neural Multiple Source Tracking (N-MUST) algorithm, which is based on an architecture of a family of radial basis function neural networks (RBFNN), is investigated for multiple source tracking with neural network-based adaptive array antennas. The N-MUST algorithm consists of an initial stage with a number of RBFNN's trained to detect the presence of the sources, while a second stage of networks is trained to estimate the exact locations of the sources. The field of view of the antenna array is divided into separate angular sectors, which are in turn assigned to a different pair of RBFNN's. When a network detects one or more sources in the first stage, the corresponding second stage networks are activated to perform the direction of arrival estimation step. No prior knowledge of the number of present sources is required. Simulation results are performed and experimental data is applied to the networks to investigate the required training criteria to achieve good generalization in the detection mode, with respect to the angular separations and relative SNR of the sources. The results show substantial reduction in the computational complexity of the network training compared to the single network approach.

This paper proposes a novel blind beamforming method for interference mitigation in direct-sequence spread-spectrum (DS/SS) communications based on subspace projection. Instantaneously narrowband interference signals are considered. The interference signals are highly localized in the time-frequency domain, and their characteristics are different from those of the DS/SS signal. The instantaneous frequencies of the nonstationary interference signals are estimated from the time-frequency domain and used in the temporal domain for interference nulling. We derive the receiver SNRs, which shows improved performance in strong interference environments.

The paper presents an adaptive technique to create multiple nulls to suppress multiple jammers. The adaptive technique is based on forward and backward smoothed ESPRIT (estimation of signal parameters via rotational invariance technique), which is capable of handling both coherent and non-coherent narrow-band multiple signals. It is demonstrated to be effective and efficient. Computer simulations for a linear array of 32 elements have shown that the optimization technique can create wide as well as deep nulls, which correspond well to jammer widths and strengths. In all cases the SIR improves substantially and converges quickly.

We investigate the data-record-size requirements to meet a given performance objective in interference suppression and direction-of-arrival (DoA) estimation problems. For interference suppression problems we consider the MVDR (minimum-variance-distortionless-response) beamformer evaluated under desired-signal-present and desired-signal- absent conditions. For the former case we adopt as the figure of merit the ratio between the output variance of the ideal and the estimated SMI (sample-matrix-inversion) MVDR filter, while for the latter we examine the inverse of the corresponding ratio of the output SINRs (signal-to- interference-plus-noise ratio). For DoA estimation problems we consider the conventional and the MVDR DoA estimation algorithm and we adopt a spectrum-based performance measure that is given as a function of the ratio between the estimated and the ideal spectrum. For all cases, we derive closed form expressions that provide the data record size that is necessary to achieve a given performance confidence level in a neighborhood of the optimal performance point as well as expressions that identify the performance level that can be reached for a given data record size. This is done by utilizing close approximations of the involved probability density functions (pdfs) and Markoff-type inequalities. The practical significance of the derived expressions lies in the fact that they are functions of the number of antenna elements only while they are independent of the ideal covariance matrix which is not known in most realistic applications. As a byproduct of the above developments we derive close approximations to the pdf of the output SINR of the MVDR beamformer when the latter is estimated in the presence or in the absence of the desired signal.

In this paper we find the theoretical slowest convergence mode for the Multidimensional Adaptive Linear Receiver (MALR) with a received signal consisting of additive white Gaussian noise, multi-access interference, and spatially diverse multitone interference. This convergence mode is used in a decaying exponential that approximates the MSE transient which is directly related to the net effective interference suppression capability of the MALR. We show here that the slowest mode of convergence is related to the AWGN variance.

Conventional phased-array antenna schemes operate by combining the signals from individual antennas in an array. Normally, the signals are passed through receivers with common local oscillators to ensure phase coherence prior to signal weighting and combination. With an appropriate delay added to each signal path, the weights can be adjusted to emphasize the signal that one of attempting to receive while minimizing the contributions from intentional or unintentional jammers. By applying DSP techniques to the signals after downconversion in receivers that do not share local oscillators, it is possible to detect differences in local oscillator frequency and then `virtually' phase-lock the signal streams together. In order to detect the local- oscillator offsets between receivers, we have proposed and experimentally demonstrated a correlation method using received `signals of opportunity' available from the antenna array. Examples of `signals of opportunity' are time beacons, radio broadcasts and even external noise. By comparing the signals from both receivers, it is possible to determine the relative frequency offset between them. We present both the theory of operation of this technique as well as experimental results.

Second-order multipath channel estimation procedures for direct-sequence code-division-multiple-access communications induce phase ambiguity that necessitates differential phase- shift-keying (DPSK) modulation and detection. The maximum likelihood (ML) single-symbol multiuser DPSK/CDMA detector is derived with direct generalization to multiple-symbol (block) multiuser DPSK/CDMA detection. Exponential complexity requirements limit the use of the ML rule to theoretical lower-bound bit-error-rate benchmarking. Linear filter DPSK demodulators are viewed as a practical alternative. Phase-ambiguous RAKE filtering followed by RAKE-output differential detection is considered. The familiar minimum-variance-distortionless-response (MVDR) PSK/CDMA filter (designed for minimum filter output energy under the constraint of distortionless response in a given RAKE vector direction) adds the valuable feature of active interference suppression; however, minimum disturbance variance at the differential logic output can be claimed formally only in the absence of multipath (no inter-symbol- interference). Short-data-record adaptive alternatives to costly and slow adaptive MVDR implementations are sought in the context of auxiliary-vector filtering. Numerical and simulation studies illustrate the developments.

This paper describes an approach and preliminary results associated with the use of pattern recognition techniques to identify transmitted information (symbols) in a temporally variant acoustic channel. The method allows the observation of the transmitted signal simultaneously in both time and frequency space and does not necessary rely on the application of adaptive algorithms for reception. The observation of the symbol energy from the Wigner-Ville Distribution as 2D pattern can allow the determination of channel characteristics over short symbol sequences and can provide a means for symbol detection. For the QPSK modulation used, wavelet filtering provides a basis for noise reduction and WVD cross tem separation. The process used for development of the pattern classifier is described and results are presented for shallow water acoustic data on a limited data set.

We propose a method to classify symbol constellations when the unknown signal of interest has propagated through an unknown frequency selective multipath fading channel. The proposed approach is based on a receiver structure composed of a set of parallel adaptive blind equalizers each of which minimizes a cost function matched to a specific constellation. The method works in batch mode, and an average gradient rather than a stochastic gradient approach is used to obtain smoother convergence. The channel is assumed to be FIR; to avoid instability problems, the equalizer is also constrained to be FIR, even though this creates an unavoidable residual inter-symbol interference (ISI). The proposed algorithm may also be used in conjunction with cumulant-based classifiers (to be triggered after the multipath channel has been partially equalized), and it can be extended to antenna arrays receivers to avoid residual ISI even using FIR filters. The performance of the proposed approach is demonstrated via simulations.

This work deals with synchronous CDMA transmission using orthogonal channel codes in frequency selective multipath. The motivating application is the Forward Link in 3G CDMA Cellular Systems. In saturated systems, the intra-cell Multi-user Access Interference created by the multipath causes the RAKE receiver to yield extremely poor performance. The chip-level MMSE estimate of the (multi- user) synchronous sum signal transmitted from the base followed by a correlate and sum with the desired user's spreading code, Walsh-Hadamard channel code multiplied by appropriate portion of long code, has been shown to yield superior performance to the RAKE receiver. This work considers reduced-rank, chip-level MMSE estimation based on the Multi-Stage Nested Weiner Filter (MSNWF) of Goldstein and Reed. It is shown that only a small number of stages of the MSNWF are needed in order to achieve near full-rank MMSE performance over a practical SNR range. This implies rapid adaptation in the case where the chip-level MMSE equalizer is adapted based on the pilot channel.