This paper presents a new algorithm for wavelet based image compression by exploiting zero tree concept in the wavelet decomposed image. The algorithm has a big edge over previously developed wavelet based image compression algorithms in that it utilizes inter and intra band correlation simultaneously, something that previous algorithms failed to exploit. Besides the improvement in coding efficiency, the algorithm also uses significantly lower memory for computation and coding thereby reducing the complexity of the algorithm. The striking feature of the algorithm is pass independent coding that makes it suitable for application to error protection schemes and makes it less vulnerable to data loss due to noisy communication channel. The algorithm codes all the color bands independently thus enabling differential coding for the color information. The paper starts with the discussion of the concept underlying the algorithm and then sees the algorithm in a broader light. Comparisons have been made to SPIHT, the well known zero tree coder for wavelet based image compression in terms of coding efficiency, memory requirements and error resiliency.

A new approach in the enhancement/suppression of narrowband signals, based on the combination of the maximum output SNR and the least-squares optimizing criteria, has been presented previously. The resulting statistically-optimal null filters (SONFs) have also been presented as recursive filters. Because of their time-varying nature, optimality under the MSE criterion and the similarity to the RLS method, it is anticipated that the SONFs may be related to the well known Kalman filter. In this paper, we show the SONFs to be an alternate implementation of the Kalman filter. For the first order case we show this analytically. Due to mathematical intractability, for the higher order case we set up the mathematical formulations and use simulations for verification.

With the advances in speech recognition and wireless communications, the possibilities for information access in the automobile have expanded significantly. We describe four system prototypes for (i) voice-dialing, (ii) Internet information retrieval-called InfoPhone, (iii) voice e-mail, and (iv) car navigation. These systems are designed primarily for hands-busy, eyes-busy conditions, use speaker-independent speech recognizers, and can be used with a restricted display or no display at all. The voice-dialing prototype incorporates our hands-free speech recognition engine that is very robust in noisy car environments (1% WER and 3% string error rate on the continuous digit recognition task at 0 db SNR). The InfoPhone, voice e-mail, and car navigation prototypes use a client-server architecture with the client designed to be resident on a phone or other hand-held device.

Semantic clusters of a domain form an important feature that can be useful for performing syntactic and semantic disambiguation. Several attempts have been made to extract the semantic clusters of a domain by probabilistic or taxonomic techniques. However, not much progress has been made in evaluating the obtained semantic clusters. This paper focuses on an evaluation mechanism that can be used to evaluate semantic clusters produced by a system against those provided by human experts.

In this paper, we propose an alternate non-parametric statistically optimal method of null filtering. One of the important features of this method lies in its ability to process signals of short record lengths. The optimality criteria for maximum output SNR and the minimum mean-square error are combined to generate the new approach. The method is first designed for the coherent case (where the desired signal shape is a priori known) and later extended to include the non-coherent case based on orthogonal signal expansion. To deal with a non-orthogonal signal expansion, we propose a sliding Gram-Schmidt orthogonalization. An application to separate two closely spaced damped sinusoids is considered. Simulation results are presented comparing the proposed methods with the conventional one based on Constrained Notch Filtering.

This paper presents an approach to identifying conjuncts of coordinate conjunctions appearing in text which has been labelled with syntactic and semantic tags. The overall project of which this research is a part is also briefly discussed. The program was tested on a 10,000 word chapter of the Merck Veterinary Manual. The algorithm is deterministic and domain independent and it performs relatively well on a large real-life domain. Constructs not handled by the simple algorithm are also described in some detail.