Chair: Magdy Bayoumi, University of Southwestern Louisiana (USA)
Katherine Wang, Apple Computer Inc. (USA)
James Normile, Apple Computer Inc. (USA)
Hsi-Jung Wu, Apple Computer Inc. (USA)
Dulce Ponceleo, Apple Computer Inc. (USA)
Ken Chu, Apple Computer Inc. (USA)
Kah-Kay Sung, Apple Computer Inc. (USA)
Video and audio conferencing over networks is becoming increasingly popular due to the availability of video and audio I/O as standard equipment on many computer systems. So far, many algorithms have concentrated on playback only capability. This generally results in unacceptable real-time performance with respect to latency and encoder complexity. We describe a software-only system that allows full duplex video communication. For our analysis and implementation we chose a DCT based method that uses motion estimation and is modelled on the CCITT H.261 standard. We begin with a brief discussion of the algorithm and follow with an analysis of the computational requirements for each major block. The results presented show the effect of computational simplifications on signal to noise ratio and image quality. In our conclusion, we examine the processing needs for full resolution coding and project when this will become available.
M. Tahernezhadi, Northern Illinois University (USA)
L. Liu, Northern Illinois University (USA)
In this paper, a practical IIR (pole-zero) lattice adaptive acoustic echo canceller (AEC) for teleconferencing application is proposed. The proposed algorithm cosists in two parts: forward lattice and inverse lattice. Collectively, they are referred to as LATIN (Lattice and Inverse Lattice) configuration. While the forward lattice is responsible for acoustic echo cancellation, the inverse lattice is employed in the double-talk (DT) mode only as to undo the distortion of the near-end speech, brought about by forward lattice when suppressing the acoustic echo. Assuming M poles and M zeros for the proposed AEC, the complexity of the proposed algorithm is approximately twice the complexity of an M- tap FIR gradient lattice algorithm. Real-time experimentation, conducted on ADSP21020 floating-point DSP chip, in conjunction with simulation attest to practical usefulness of the proposed algorithm.
H. Klingele, Labor Dr. Steinbichler (GERMANY)
H. Steinbichler, Labor Dr. Steinbichler (GERMANY)
Holographic interferometry offers amplitude data with a high spatial resolution which can be used as vibration boundary condition for calculating the corresponding sound pressure field. When investigating objects with arbitrary 3D-shape this requires contour measuring, performing holographic interferometry for three axes of freedom, combining contour and vibration data into a Boundary Element (BE) model, and then solving the discretized Helmholtz-Kirchhoff integral equation for the surface sound pressure. The latter is done by means of a newly developed Picard-iterative Boundary Element Method (PIBEM), which does not need matrix operations at all and such is capable of also treating large BE models arising from small bending wavelengths at high vibration frequencies. An experimental verification of this method by microphone measurements in an anechoic chamber is presented for a cylindrical example object.
Peter O'Shea, RMIT
Edward Palmer, Queensland University of Technology
Gordon Frazer, Defence Science and Technology Organisation (AUSTRALIA)
This paper considers the problem of measuring and analysing the frequency and phase angle variations which occur in a power system after a fault or disturbance. This is an important problem because the results of the analysis are used in implementing control strategies to avert generator cutout. Commonly, the frequency is not measured directly, but is inferred from the instantaneous power at the substation. Since there is a complicated non-linear relationship between the power and the frequency, some bias is incurred in the measurement. This paper proposes instead that the angle and or frequency be obtained directly via an analytic signal readily constructed from the 3-phase output of the power system. The paper also proposes enhancements for some of the currently employed signal analysis techniques, and a new technique based on an extension of Thompson's harmonic line test.
J. Streit, University of Southampton (UK)
L. Hanzo, University of Southampton (UK)
A highly bandwidth efficient, fixed but arbitrarily programmable rate, perceptually weighted Discrete Cosine Transform (DCT) based video communicator for quarter common intermediate format (QCIF) videophone sequences is presented. Perceptually weighted cost/gain controlled motion compensation and quad-class DCT-based compression is applied without variable rate compression techniques and without adaptive buffering in order to maintain a fixed transmission rate, which can be adjusted to any required value. In this treatise we opted for a source coded rate of 11.36 kbps and the sensitivity-matched forward error correction (FEC) coded rate became 20.32 kbps. A partial forced update technique was invoked in order to keep transmitter and receiver aligned amongst hostile channel conditions. When using coherent pilot symbol assisted 16-level quadrature amplitude modulation (16-PSAQAM), an overall signalling rate of 9 kBd was yielded. Over lower quality channels 4QAM had to be invoked, which required twice as many time slots to accommodate the resulting 18 kBd stream. Over the best Gaussian and worst Rayleigh channels signal-to-noise ratio (SNR) values in the range of 7 to 20 dB were needed for these modems in order to maintain near- unimpaired image quality. In a bandwidth of 200 kHz, similarly to the GSM speech channel, 16 and 8 videophone users can be supported, when using the 16QAM and 4QAM systems, respectively.
D. Korompis, University of California at Los Angeles (USA)
A. Wang, University of California at Los Angeles (USA)
K. Yao, University of California at Los Angeles (USA)
We perform simulation of various digital signal processing microphone array architectures to compare their suitability for a hearing aid pre-processor. The architectures include fixed narrowband array, General Sidelobe Canceler array and Maximum Energy array. The arrays are all equi-spaced linear array. In particular, arrays with 6 microphones and various number of taps have been simulated under typical hearing aid environment of reverberance, short speaker distance and competing speakers. The sampling frequency is chosen to be 10 KHz, with the physical length of the array being 17 cm. The improvements of various array designs are demonstrated and discussed.
Hsiang-ling Li, Arizona State University (USA)
Chaitali Chakrabarti, Arizona State University (USA)
A novel VLSI architecture is proposed for implementing a long constraint length Viterbi Decoder (VD) for code rate k/n. This architecture is based on the encoding structure where k input bits are shifted into k shift registers in each cycle. The architecture is designed in a hierarchical manner by breaking the system into several levels and designing each level independently. At each level, the number of computation units, the interconnection between the units as well as allocation and scheduling issues have been determined. In-place storage of accumulated path metrics and trace back implementation of the survivor memory have also been addressed. The resulting architecture is regular, flexible and achieves a better than linear tradeoff between hardware complexity and computation time.
Surendra K. Jain, University of Minnesota (USA)
Keshab K. Parhi, University of Minnesota (USA)
A new parallel-in-parallel-out bit-level pipelined multiplier is presented to perform multiplication in GF(2^m). This new multiplier uses m^2 basic cells where each cell has 2 2-input AND, 2 2-input XOR and 3 1-bit latches. The system latency of this multiplier is m+1 compared to 3m in previous architectures. The number of latches has been reduced from 7 to 3. We also present a bit-level pipelined parallel-in-parallel-out squarer. This squarer has a system latency of m/2 compared to 3m in previous designs and is 25% more hardware efficient. The critical path in both these proposed designs are the same as in existing designs.
Jianwei Miao, Georgia Institute of Technology (USA)
Mark A. Clements, Georgia Institute of Technology (USA)
In this paper, we have addressed the problem of analyzing and classifying the digital pulse waveforms from a radiation detector of an alpha-constant air monitor with radon and plutonium sources. With the availability of extremely high-speed A/D conversion with good resolution, it is now possible to look more deeply at the waveform shapes than is currently done. In our studies, a new technique of unsupervised pattern recognition system was investigated for analysis and classification of digital pulse waveforms. This system can extract key multidimensional features from each output pulse waveform and perform an analysis of those features. Finally, a Bayes' optimal classifier was generated based on Gaussian model. The projected accuracy of the Bayes' Gaussian model to classify three classes of digital pulse waveforms is 98.33% in overall. The preliminary results of this new technique demonstrate clearly improved measurement conditions and are therefore very promising.
M. An, Aware Inc.
N. Anupindi, Aware Inc.
Michael Bletsas, Northeastern University
G. Kechriotis, Aware Inc.
C. Lu, Townson State University
E.S. Manolakos, Northeastern University
R. Tolimieri, Aware Inc.(USA)
In this paper we discuss the parallel implementation of multidimensional FFTs on distributed memory multi-processor machines. We introduce a compact notation to describe four equivalent parallel algorithms and discuss their advantages and disadvantages. Two algorithms, suitable for the case when initial and final data are distributed either row- or column- wise, the traditional Row-Column (RC) and a variation of the Vector Radix (VR) that we call partial Vector Radix (PVR) are presented and their efficiency on the Paragon is compared. It is shown that the PVR, although it requires larger amount of interprocessor communication, results in more efficient implementations due to the regularity of local and distributed memory accesses. For the case in which data are partitioned along both dimensions, two suitable parallel algorithms, the Collect-Distribute (CD) and the general full Vector-Radix (FVR), are presented. Again, it is shown that regularity in memory accesses for the case of the FVR, results in more efficient implementations.
Murari Srinivasan, University of Maryland (USA)
Ruth DeFries, University of Maryland (USA)
The mixture model is an attempt to accurately model the ground truth in the case of low-resolution remote-sensed imagery. The model assumes that a pixel in the image does not consist of a single class, but consists instead of the sum of fractions of various classes. We use an iterative least squares approach to estimate these fractions for every pixel. Results are provided on synthetic data as well as real AVHRR data from the African continent.