Abstract
High-quality audio bit-rate reduction systems are widely used in many application areas involving audio broadcast, streaming and download services. With the advent of 3G mobile and wireless communication networks, there is a clear opportunity for new multimedia services, notably those relying on two-way high- quality audio communication. In t his paper we describe a new source/perceptual audio coder that features low-delay, intrinsic error robustness and high subjective audio quality at competitive compression ratios. The structure of the audio coder is described and an emphasis is given on its innovative approaches to semantic signal segmentation and decomposition, independent coding of sinusoidal and noise components, and bandwidth extension using Accurate Spectral Replacement. A few test results are presented that illustrate the operation and performance of the new coder.

Abstract
In this paper we describe the components of a novel audio coding algorithm capable of delivering high-fidelity CDlike stereo audio at the bit rates of 40-48 kbps and natural sounding FM grade mono at the bit rates of 18-22 kbps. Bandwidth Extension has emerged as an important tool for the satisfactory performance of low bit rate audio codecs. Recently we proposed two new bandwidth extension algorithms, Fractal Self-Similarity Model (FSSM) and Accurate Spectral Replacement (ASR), which belong to a new class of Bandwidth Extension techniques which are applied directly to the high resolution frequency representation of the signal (e.g., MDCT or ODFT). The proposed coding scheme uses FSSM and ASR in an adaptive and complementary framework. Another important component of the proposed codec is a wideband psychoacoustic model that makes an explicit use of the Comodulation Release of Masking (CMR) phenomenon. It also includes a novel parametric stereo coding technique. The proposed audio coding scheme is geared towards broadcast applications where codec latency and encoder complexity is generally not an overriding concern. In this paper we present algorithmic details of the new codec, audio demonstrations, and, comparison to other audio coding schemes. Further information and audio demonstrations are available at http://www.atc-labs.com/teslapro.

Abstract
A critical issue in the design of logic programming systems is their memory performance, both in terms of total memory usage and locality in memory accesses. BEAM, as most modern Prolog systems, requires both good emulator design and good memory performance for best performance. We report on a detailed study of the memory management techniques used on our sequential implementation of the EAM. We address questions like how effective are the techniques the BEAM uses to recover and reuse memory space, how garbage collection affects performance and how to classify and unify variables in a EAM environment. We also propose a finer variable allocation scheme to reduce memory overheads that is quite effective at reducing memory pressure, with only a small overhead.

Abstract
In this paper we describe a new family of smooth power complementary windows which exhibit a very high level of localization in both time and frequency domain. This window family is parameterized by a "smoothness quotient". As the smoothness quotient increases the window becomes increasingly localized in time (most of the energy gets concentrated in the center half of the window) and frequency (far field rejection becomes increasing stronger to the order of 150 dB or higher). A closed form solution for such window function exists and the associated design procedure is described. The new class of windows is quite attractive for a number of applications as switching functions, equalization functions, or as windows for overlap-add and modulated filter banks. An extension to the family of smooth windows which exhibits improved near-field response in the frequency domain is also discussed. More information is available at http://www.atc-labs.com/technology/misc/windows.

Abstract
The paper describes how mapping a neural network into a rule-based fuzzy inference system leads to knowledge extraction. This mapping makes explicit the knowledge implicitly captured by the neural network during the learning stage, by transforming it into a set of rules. By applying the method to transformer fault diagnosis using dissolved gas-in-oil analysis, one could not only develop intelligent diagnosis systems, providing better results than the application of the IIEC 60599 Table, but also generate a new rule table whose application also leads to better diagnosis results.

Abstract
Derivative (D) terms in PID controllers are infrequently used in practice due to precise tuning requirements and noisy controller output. The first problem can be alleviated by using several modern tuning methods, while substantial decrease of controller activity can be achieved by filtering controller output. However, introducing additional filter at controller output can destabilise control loop if the PID controller parameters are not properly re-tuned. It has been shown that both problems can be solved simultaneously by using the magnitude optimum (MO) and disturbance rejection magnitude optimum (DRMO) tuning method. Moreover, it is shown that the PID controller performance can be higher than of the PI controller even at lower controller activity. The best controller performance vs. activity ratio is achieved with the second or the third order filter at controller output. © 2005 IEEE.