A setup involving zero-delay sequential transmission of a vector Markov source over a burst erasure channel is studied. A sequence of source vectors is compressed in a causal fashion at the encoder, and the resulting output is transmitted over a burst erasure channel. The destination is required to reconstruct each source vector with zero-delay, but those source sequences that are observed either during the burst erasure, or in the interval of length W following the burst erasure need not be reconstructed. The minimum achievable compression rate is called the rate-recovery function. We assume that each source vector is independent identically distributed (i.i.d.) across the spatial dimension and is sampled from a stationary, first-order Markov process across the temporal dimension. For discrete sources, the case of lossless recovery is considered, and upper and lower bounds on the rate-recovery function are established. Both these bounds can be expressed as the rate for predictive coding, plus a term that decreases at least inversely with the recovery window length W. For Gauss-Markov sources and a quadratic distortion measure, upper and lower bounds on the minimum rate are established when W = 0. These bounds are shown to coincide in the high resolution limit. Finally, another setup involving i.i.d. Gaussian sources is studied and the raterecovery function is completely characterized in this case.

We introduce a framework to study fundamental limits of sequential coding of Markov sources under an error propagation constraint. An encoder sequentially compresses a sequence of vector-sources that are spatially i.i.d. but temporally correlated according to a Markov process. The channel erases up to B packets in a single burst, but reveals all other packets to the destination. The destination is required to reproduce all the source-vectors instantaneously and in a loss less manner, except those sequences that occur in a window of length B+W following the start of the erasure burst. We define a rate-recovery function R(B, W), the minimum compression rate that can be achieved in this framework, and develop upper and lower bounds for first-order Markov sources. For the special class of linear diagonally correlated deterministic sources, we propose a new coding technique -- prospicient coding -- that achieves the rate-recovery function. Finally, a lossy extension to the rate-recovery function is also studied for a class of Gaussian sources where the source is temporally and spatially i.i.d. and the decoder aims to recover a collection of past K sources with a quadratic distortion measure. The optimal rate-recovery function is compared with the sub-optimal techniques including forward error correction coding (FEC) and Wyner-Ziv coding, and performance gains are quantified.

This paper proposes a content-adaptive blending method, smart compositing, for displaying two overlapped video frames on the same screen while preserving the readability of both. A pixel-wise adaptive blending factor map is generated according to the edge and saturation information of the content of only the overlay frame. Using this blending factor map, regions of the overlay frame with edges or saturated color are assigned to be more opaque and the remaining regions are assigned to be more transparent. A halo is also created around the edges of the overlay content which enhances the edges and disambiguates them from the underlying frame. The proposed method is suitable for overlaying many different types of content (e.g. drawings, slides, texts, and pictures) and does not require any information (e.g., an opacity mask) from the application which generates the content. This method has low computational complexity and has been implemented in real-time.

This paper proposes a method for automatic calibration of a projector-camera system with a see-through screen. Conventional projector-camera systems usually project a calibration pattern on an opaque screen for calibrating the system. Images captured by the camera contain only the calibration pattern. Recently, projector-camera systems with see-through screen have been developed for visual collaboration. Due to the semi-transparent property of the see-through screen, not only the calibration pattern, but also the background on the opposite side of the screen is captured by the camera. Since the calibration pattern is contaminated by the image of the background, conventional feature detection methods have difficulty in detecting the features of the calibration pattern and thus often fail to calibrate the system. In the proposed methods, the calibration pattern is temporally varied so that temporal correlation can be applied on a sequence of captured images to separate the calibration pattern from the background. This method works robustly and accurately even for backgrounds with high dynamic range and large motion. The entire calibration process is fully automatic and does not require any parameter tuning.

Communication has seen enormous advances over the past 100 years including radio, television, mobile phones, video conferencing, and Internet-based voice and video calling. Still, remote communication remains less natural and more fatiguing than face-to-face. The vision of immersive communication is to enable natural experiences and interactions with remote people and environments in ways that suspend disbelief in being there. This paper briefly describes the current state-of-the-art of immersive communication, provides a vision of the future and the associated benefits, and considers the technical challenges in achieving that vision. The attributes of immersive communication are described, together with the frontiers of video and audio for achieving them. We emphasize that the success of these systems must be judged by their impact on the people who use them. Recent high-quality video conferencing systems are beginning to deliver a natural experience-when all participants are in custom-designed studios. Ongoing research aims to extend the experience to a broader range of environments. Augmented reality has the potential to make remote communication even better than being physically present. Future natural and effective immersive experiences will be created by drawing upon intertwined research areas including multimedia signal processing, computer vision, graphics, networking, sensors, displays and sound reproduction systems, haptics, and perceptual modeling and psychophysics.

A rateless code-i.e., a rate-compatible family of codes-has the property that codewords of the higher rate codes are prefixes of those of the lower rate ones. A perfect family of such codes is one in which each of the codes in the family is capacity-achieving. We show by construction that perfect rateless codes with low-complexity decoding algorithms exist for additive white Gaussian noise channels. Our construction involves the use of layered encoding and successive decoding, together with repetition using time-varying layer weights. As an illustration of our framework, we design a practical three-rate code family. We further construct rich sets of near-perfect rateless codes within our architecture that require either significantly fewer layers or lower complexity than their perfect counterparts. Variations of the basic construction are also developed, including one for time-varying channels in which there is no a priori stochastic model.

We consider the problem of efficiently determining whether two microphones are listening to the same acoustic signal. This “reference free” acoustic match problem is useful, for example, for allowing networked computers or phones to automatically rendezvous and collaborate whenever they are within earshot. We describe how this problem differs from the well-studied music identification problem. We cast the problem as a hypothesis test, and argue that an effective test statistic may be found using frequency-domain inner products. We validate our approach using a set of 75 speech samples together with inexpensive microphones built in to notebook computers.

A rateless code, or a rate-compatible family of codes, has the property that the higher rate codes have codewords that are prefixes of those of the lower rate ones. A perfect family of such codes is one in which each of the codes in the family is capacity-achieving. We show by construction that perfect rateless codes with low-complexity decoding algorithms exist for additive white Gaussian noise channels. As an illustration of our framework, we design a practical three-rate code family. We further demonstrate that a rich set of perfect or near-perfect rateless codes may be found via numerical optimization

Media streaming over the best-effort public Internet has been a focus of research for over a decade. Enterprise or corporate streaming is another area of media streaming that is practically very important and has a different set of challenges and feasible solutions. For example, the quality and reliability requirements for enterprise streaming are much stricter than for typical Internet streaming. Furthermore, in typical enterprise streaming scenarios, a single entity has control over most elements of the system, including the end-points and the infrastructure. This entity has the powerful ability to monitor, adapt, and deploy new infrastructure as necessary. The goal of this paper is to describe enterprise streaming and identify the basic differences between typical Internet streaming and enterprise streaming, and how these differences alter the challenges that must be overcome for enterprise streaming to be successful. Specifically, we examine enterprise streaming media content delivery network design and operation, video conferencing, peer-to-peer networking (P2P), voice over IP (VoIP), and briefly touch upon wireless and security issues

Rateless codes are good codes of infinite length that have the property that prefixes of such codes are themselves good codes. This makes them attractive for applications in which the channel quality is uncertain, where systems transmit as much of a codeword as necessary for decoding to be possible. In particular, rateless codes are potentially attractive for wireless communication. In a recent work, a rateless coding scheme was proposed for the AWGN channel, based on layering, repetition and random dithering. We extend this scheme to multiple-input single-output (MISO) Gaussian channels. We show that the rate loss associated with orthogonal design space-time codes may be alleviated by layering and dithering, very similar to the rateless approach for the AWGN channel. We then combine the two schemes and arrive at a close-to-capacity rateless code for MISO channels. The required complexity depends on the fraction of capacity that is targeted, is linear in the capacity of the channel and does not depend on the number of transmit antennas. Furthermore, the coding scheme uses only one base AWGN code

We study potential gains in the spectral efficiency of multi-cell uplink CDMA systems that accrue from assigning users in different parts of a cell to different frequency bands. We develop a criterion that enables a fair comparison with a conventional (non-partitioned) system. Two design scenarios are considered: (i) each user requires a fixed data rate; (ii) each user has a fixed SNR. We show that in both scenarios there is a gain in spectral efficiency - regardless of the particular partitioning strategy employed f an MMSE receiver is used. There is no gain if a decorrelator or a matched filter receiver is used. We validate our results through numerical simulations and provide an intuitive explanation based on the coding-spreading tradeoff.

Fundamental results concerning the dynamics of abelian group codes (behaviors) and their duals are developed. Duals of sequence spaces over locally compact abelian (LCA) groups may be defined via Pontryagin duality; dual group codes are orthogonal subgroups of dual sequence spaces. The dual of a complete code or system is finite, and the dual of a Laurent code or system is (anti-)Laurent. If C and C/sup /spl perp// are dual codes, then the state spaces of C act as the character groups of the state spaces of C/sup /spl perp//. The controllability properties of C are the observability properties of C/sup /spl perp//. In particular, C is (strongly) controllable if and only if C/sup /spl perp// is (strongly) observable, and the controller memory of C is the observer memory of C/sup /spl perp//. The controller granules of C act as the character groups of the observer granules of C/sup /spl perp//. Examples of minimal observer-form encoder and syndrome-former constructions are given. Finally, every observer granule of C is an "end-around" controller granule of C.

Wireless local area networks, such as 802.11b, are becoming wide-spread as they provide simple wireless connectivity and data delivery. This paper examines low-latency (conversational) video communication over 802.11b networks. The challenges to enable low-latency video include overcoming the highly variable delays, losses, and bandwidth of 802.11b wireless networks. To overcome these challenges we (1) employ the H.264/MPEG-4 advanced video coding (AVC) standard for high video compression efficiency and good resilience to losses, (2) use low-latency best-effort transport mechanisms, and (3) exploit the potential path diversity between each mobile client and multiple access points in the infrastructure, where we use multiple paths simultaneously or switch between multiple paths (site selection) as a function of channel characteristics. Our results indicate that the proposed system can provide significant benefits over conventional single access point (single path) systems.

We consider transmission over a discrete-time Rayleigh fading channel, in which successive symbols face independent fading, and where neither the transmitter nor the receiver has channel state information. Subject to an average power constraint, we study the capacity-achieving distribution of this channel and prove it to be discrete with a finite number of mass points, one of them located at the origin. We numerically compute the capacity and the corresponding optimal distribution as a function of the signal-to-noise ratio (SNR). The behavior of the channel at low SNR is studied and finally a comparison is drawn with the ideal additive white Gaussian noise channel.

A simple sphere bound gives the best possible tradeoff between the volume per point of an infinite array L and its error probability on an additive white Gaussian noise (AWGN) channel. It is shown that the sphere bound can be approached by a large class of coset codes or multilevel coset codes with multistage decoding, including certain binary lattices. These codes have structure of the kind that has been found to be useful in practice. Capacity curves and design guidance for practical codes are given. Exponential error bounds for coset codes are developed, generalizing Poltyrev's (1994) bounds for lattices. These results are based on the channel coding theorems of information theory, rather than the Minkowski-Hlawka theorem of lattice theory.

Several aspects of the design and optimization of coded multiple-antenna transmission diversity methods for slowly time-varying channels are explored from an information-theoretic perspective. Both optimized vector-coded systems, which can achieve the maximum possible performance, and suboptimal scalar-coded systems, which reduce complexity by exploiting suitably designed linear precoding, are investigated. The achievable rates and associated outage characteristics of these spatial diversity schemes are evaluated and compared, both for the case when temporal diversity is being jointly exploited and for the case when it is not. Complexity and implementation issues more generally are also discussed.

Summary form only given. By jointly optimizing source and channel codes, we can generally get less overall average distortion and more robustness to channel impairments than with separately designed source and channel codes. In many cases, however, it is difficult or sometimes impossible for a single coding scheme to achieve the least possible distortion for a large range of channel variations except the trivial Gaussian case, where the rate distortion bound on the mean square error is achieved for all values of the channel signal to noise ratio (SNR) when the i.i.d. Gaussian source and the additive white Gaussian noise (AWGN) channel have the same bandwidth. In this paper, we show that a simple uncoded system that transmits an unmodified uniform i.i.d. source defined on the unit interval I=[0,1] through a modulo AWGN channel (an AWGN channel followed by a mod-1 mapping such that the channel output is the fractional part of the AWGN channel output) can achieve the rate distortion bound almost optimally for all values of the channel SNR.

We derive performance limits for two closely related communication scenarios involving a wireless system with multiple-element transmitter antenna arrays: a point-to-point system with partial side information at the transmitter, and a broadcast system with multiple receivers. In both cases, ideal beamforming is impossible, leading to an inherently lower achievable performance as the quality of the side information degrades or as the number of receivers increases. Expected signal-tonoise ratio (SNR) and mutual information are both considered as performance measures. In the point-to-point case, we determine when the transmission strategy should use some form of beamforming and when it should not. We also show that, when properly chosen, even a small amount of side information can be quite valuable. For the broadcast scenario with an SNR criterion, we find the efficient frontier of operating points and show that even when the number of receivers is larger than the number of antenna array ...

A class of efficient strategies for exploiting transmit antenna diversity on fading channels is developed. These techniques, which we refer to as linear antenna precoding, asymptotically transform a nonselective Rayleigh fading channel into a nonfading, simple white marginally Gaussian noise channel with no intersymbol interference. Linear antenna precoding requires no additional power or bandwidth, and is also attractive in terms of computation, robustness and delay considerations.

For pt.I see ibid., vol.42, no.3, p.751-65 (1996). In Part I, general results on rotationally invariant codes and encoders were derived assuming no algebraic structure. In Part II, trellis codes based on group systems are considered as a special case for which code and encoder constructions are particularly simple. Rotational invariance is expressed as an algebraic constraint on a group code, and algebraic constructions are found for both "absorbed precoder" encoders and for encoders with separate differential precoders. Finally, the various encoder forms used to achieve rotational invariance are compared based on their performance on an AWGN channel.

We present a theoretical framework for rotational invariance of trellis codes. The distinction between codes and encoders plays a pivotal role. Necessary and sufficient conditions for rotational invariance are derived under general assumptions, and a construction is presented that obtains a rotationally invariant encoder for almost any rotationally invariant code, independent of the code's algebraic structure. Encoders that use a differential precoder are considered as a separate case, where a system-theoretic characterization of precoding is used to find two alternative and slightly less general encoder constructions.

A group code C over a group G is a set of sequences of group elements that itself forms a group under a component-wise group operation. A group code has a well-defined state space Sigma /sub k/ at each time k. Each code sequence passes through a well-defined state sequence. The set of all state sequences is also a group code, the state code of C. The state code defines an essentially unique minimal realization of C. The trellis diagram of C is defined by the state code of C and by labels associated with each state transition. The set of all label sequences forms a group code, the label code of C, which is isomorphic to the state code of C. If C is complete and strongly controllable, then a minimal encoder in controller canonical (feedbackfree) form may be constructed from certain sets of shortest possible code sequences, called granules. The size of the state space Sigma /sub k/ is equal to the size of the state space of this canonical encoder, which is given by a decomposition of the input groups of C at each time k. If C is time-invariant and nu -controllable, then mod Sigma /sub k/ mod = Pi /sub 1>