A new scheme for coding speech at low bit rates (4.8-16 kb/s) but still maintaining high quality is described. Speech is regarded as a piecewise-stationary random signal and its synthesis is accomplished by means of a Kalman estimator at the decoder. The Kalman estimator requires for its operation a signal model and a sequence of measurements of the states of the model. A two-stage, time-varying, all-pole filter excited by white noise is used as the speech signal model. Linear combinations of speech samples taken at sparse but periodic intervals and provided in the form of innovations serve as measurements. The role of the encoder in the proposed scheme is seen as that of extracting the signal model parameters as well as forming the measurements and transmitting this information to the decoder. An optimum measurement strategy is developed for the estimator. A procedure for shaping the error spectrum of the synthesized speech is also described. Simulation studies show that coders based on the proposed scheme can provide high-quality speech at low bit rates. Important implementation details of such coders as well as their performance results for different choices of coder parameters are given.>

An efficient search procedure for CELP coders is described. Based on the phase spectra of the codevectors and the ideal excitation vector, a subset of the codebook is preselected. This subset is then exhaustively searched to find the best, perhaps optimum, codevector. Computational savings of the order of 30-40 are achieved while still maintaining good output speech quality.>

The near-lossless, i.e., lossy but high-fidelity, compression of medical Images using the entropy-coded DPCM method is investigated. A source model with multiple contexts and arithmetic coding are used to enhance the compression performance of the method. In implementing the method, two different quantizers each with a large number of quantization levels are considered. Experiments involving several MR (magnetic resonance) and US (ultrasound) images show that the entropy-coded DPCM method can provide compression in the range from 4 to 10 with a peak SNR of about 50 dB for 8-bit medical images. The use of multiple contexts is found to improve the compression performance by about 25% to 30% for MR images and 30% to 35% for US images. A comparison with the JPEG standard reveals that the entropy-coded DPCM method can provide about 7 to 8 dB higher SNR for the same compression performance.

The authors investigate the use of conditioning events (or contexts) in improving the performances of known compression methods by building a source model with multiple contexts to code the decorrelated pixels. Three methods for reversible compression, namely DPCM (differential pulse code modulation), WHT (Walsh-Hadamard transform), and HINT (hierarchical interpolation), employing, respectively, predictive decorrelation, transform decorrelation, and multiresolution decorrelation, are considered. It is shown that the performance of these methods can be enhanced significantly, sometimes even up to 40%, by using contexts. The enhanced DPCM method is found to perform the best for MR and UT (ultrasound) medical images; the enhanced WHT method is found to be the best for X-ray images. The source models used in the enhanced models employ several hundred contexts.

A scheme for the construction of m-out-of-n codes based on the arithmetic coding technique is described. For appropriate values of n, k, and m, the scheme can be used to construct an (n,k) block code in which all the codewords are of weight m. Such codes are useful, for example, in providing perfect error detection capability in asymmetric channels such as optical communication links and laser disks. The encoding and decoding algorithms of the scheme perform simple arithmetic operations recursively, thereby facilitating the construction of codes with relatively long block sizes. The scheme also allows the construction of optimal or nearly optimal m-out-of-n codes for a wide range of block sizes limited only by the arithmetic precision used.>

A scheme for compressing computer data by treating them as sequences of bytes is described. For each individual sequence to be compressed, a source model of predetermined complexity is built adaptively starting from a memoryless model. An alphabet reduction technique which permits handling of each bit within a byte separately is introduced. Variable-order Markov contexts are generated for each bit within a byte by a process of selective context splitting. The selection of a context for splitting is based on the context's probability as well as the bit entropy under the context. Estimation of bit statistics under the different contexts is made adaptively and encoding is accomplished by an arithmetic code. The scheme allows the complexity of the source model, and thereby the compression performance, to be altered easily. Experiments on typical computer files show that the present scheme, at a moderate complexity level, often outperforms some of the existing schemes.>