Leah H. Jamieson

High-quality audio compression using an adaptive wavelet packet decomposition and psychoacoustic modeling

This paper presents a technique to incorporate psychoacoustic models into an adaptive wavelet packet scheme to achieve perceptually transparent compression of high-quality (34.1 kHz) audio signals at about 45 kb/s. The filter bank structure adapts according to psychoacoustic criteria and according to the computational complexity that is available at the decoder. This permits software implementations that can perform according to the computational power available in order to achieve real time coding/decoding. The bit allocation scheme is an adapted zero-tree algorithm that also takes input from the psychoacoustic model. The measure of performance is a quantity called subband perceptual rate, which the filter bank structure adapts to approach the perceptual entropy (PE) as closely as possible. In addition, this method is also amenable to progressive transmission, that is, it can achieve the best quality of reconstruction possible considering the size of the bit stream available at the encoder. The result is a variable-rate compression scheme for high-quality audio that takes into account the allowed computational complexity, the available bit-budget, and the psychoacoustic criteria for transparent coding. This paper thus provides a novel scheme to marry the results in wavelet packets and perceptual coding to construct an algorithm that is well suited to high-quality audio transfer for Internet and storage applications.

On the complexity of explicit duration HMM's

Introduces a new recursion that reduces the complexity of training a semi-Markov model with continuous output distributions. It is shown that the cost of training is proportional to M/sup 2/+D, compared to M/sup 2/D with the standard recursion, where M is the observation vector length and D is the maximum allowed duration. >

Endpoint detection of isolated utterances based on a modified Teager energy measurement

The authors have developed an energy measure based on Teagers energy algorithm, and have applied it to the problem of endpoint detection. This energy measure is important in that it appears to be more suitable for describing the source energy associated with the production of speech sounds than the acoustic energy typically measured, and explores a new way of viewing and using Teagers energy algorithm. Experiments were conducted on 400 utterances on which endpoint detection was expected to be difficult. Typical examples that show that this new measure is more effective than traditional measures in capturing speech events such as initial and final fricatives and plosives are presented. Whereas traditional endpoint detectors have used both (acoustic) energy and zero crossing rate, the new measure effectively combines this information into a single measure. The experimental results demonstrate that the measure can be used to improve the performance of endpoint detection algorithms and should be effective for the detection of speech in noisy environments.<<ETX>>

FFT algorithms for SIMD parallel processing systems

SIMD (single instruction stream - multiple data stream) algorithms for one- and two-dimensional discrete Fourier transforms are presented. Parallel structurings of algorithms for efficient computation for a variety of machine size/problem size combinations are presented and analyzed. Through these algorithms, techniques for exploiting relationships between problem size and machine size are demonstrated. The algorithms are evaluated in terms of the number of arithmetic operations and interprocessor data transfers required. The ability of various interconnection networks presented in the literature to perform the needed transfers is examined. It is shown that the efficiency of a particular data distribution/algorithm decomposition approach is a function of the machine-size/problem-size relationship.

Acoustic-phonetic analysis of loud and Lombard speech in simulated cockpit conditions

Acoustic-phonetic differences between normal and abnormal speech in reference to recognition systems for the aircraft cockpit are studied. The speech used in this research was collected in an anechoic chamber with an oxygen mask, and the speaker spoke (1) normally, (2) loudly ( approximately 10 dB above normal), and (3) with 90 dB of pink noise injected into the ears (invoking the Lombard reflex). Eighteen features were analyzed for approximately 11000 phonemes (from five speakers) to determine significant differences in the three types of speech. The most reliable trends observed were energy migrations in the frequency domain and the associated changes in spectral tilt for the sonorants.<<ETX>>

Modeling duration in a hidden Markov model with the exponential family

A procedure for modeling duration with some PDF (probability density function) or PMF (probability mass function) in the exponential family is presented. A means of selecting an appropriate member of the exponential family is suggested. The parameter estimation procedure presented here offers several advantages over other methods of duration modeling. First, the duration PMF can be found directly, rather than sampling and truncating the optimum density. Secondly, the optimum duration parameters are found from Fergusons nonparametric PMF. This simplifies reestimation because the operation of casting a nonparametric PMF to the desired parametric family can be completely separated from the forward and backward algorithms. Thirdly, several competing members of the exponential family can be evaluated quickly for each state in the HMM. This makes it possible to model each states duration with the best member from a set of parametric PMFs in the exponential family. Finally, the solution holds for an PDF or PMF in the exponential family, which includes a large number of promising candidates.<<ETX>>

A software environment for parallel computer vision

A software environment tailored to computer vision and image processing (CVIP) that focuses on how information about the CVIP problem domain can make the high-performance algorithms and the sophisticated algorithm techniques being designed by algorithm experts more readily available to CVIP researchers is presented. The environment consists of three principle components: DISC, Cloner, and Graph Matcher. DISC (dynamic intelligent scheduling and control) supports experimentation at the CVIP task level by creating a dynamic schedule from a users specification of the algorithms that constitute a complex task. Cloner is aimed at the algorithm development process and is an interactive system that helps a user design new parallel algorithms by building on and modifying existing library algorithms. Graph Matcher performs the critical step of mapping new algorithms onto the target parallel architecture. Initial implementations of DISC and Graph Matcher have been completed, and work on Cloner is in progress.<<ETX>>

A probabilistic approach to AMDF pitch detection

The authors present a probabilistic error correction technique to be used with an average magnitude difference function (AMDF) based pitch detector. This error correction routine provides a very simple method to correct errors in pitch period estimation. Used in conjunction with the computationally efficient AMDF, the result is a fast and accurate pitch detector. In performance tests on the CSTR (Center for Speech Technology Research) database, probabilistic error correction reduced the gross error rate from 6.07% to 3.29%.

Overview of parallel processing approaches to image and video compression

In this paper we present an overview of techniques used to implement various image and video compression algorithms using parallel processing. Approaches used can largely be divided into four areas. The first is the use of special purpose architectures designed specifically for image and video compression. An example of this is the use of an array of DSP chips to implement a version of MPEG1. The second approach is the use of VLSI techniques. These include various chip sets for JPEG and MPEG1. The third approach is algorithm driven, in which the structure of the compression algorithm describes the architecture, e.g. pyramid algorithms. The fourth approach is the implementation of algorithms on high performance parallel computers. Examples of this approach are the use of a massively parallel computer such as the MasPar MP-1 or the use of a coarse-grained machine such as the Intel Touchstone Delta.

A model for an intelligent operating system for executing image understanding tasks on a reconfigurable parallel architecture

Abstract Parallel processing is one approach to achieving the large computational processing capabilities required by many real-time computing tasks. One of the problems that must be addressed in the use of reconfigurable multiprocessor systems is matching the architecture configuration to the algorithms to be executed. This paper presents a conceptual model that explores the potential of artificial intelligence tools, specifically expert systems, to design an Intelligent Operating System for multiprocessor systems. The target task is the implementation of image understanding systems on multiprocessor architectures. PASM is used as an example multiprocessor. The Intelligent Operating System concepts developed here could also be used to address other problems requiring real-time processing. An example image understanding task is presented to illustrate the concept of intelligent scheduling by the Intelligent Operating System. Also considered is the use of the conceptual model when developing an image understanding system in order to test different strategies for choosing algorithms, imposing execution order constraints, and integrating results from various algorithms.