Jens Hannemann

Performance of phase transform for detecting sound sources with microphone arrays in reverberant and noisy environments

The performance of sound source location (SSL) algorithms with microphone arrays can be enhanced by processing signals prior to the delay and sum operation. The phase transform (PHAT) has been shown to improve SSL images, especially in reverberant environments. This paper introduces a modification, referred to as the PHAT-@b transform, that varies the degree of spectral magnitude information used by the transform through a single parameter. Performance results are computed using a Monte Carlo simulation of an eight element perimeter array with a receiver operating characteristic (ROC) analysis for detecting single and multiple sound sources. In addition, a Fishers criterion performance measure is also computed for target and noise peak separability and compared to the ROC results. Results show that the standard PHAT significantly improves detection performance for broadband signals especially in high levels of reverberation noise, and to a lesser degree for noise from other coherent sources. For narrowband targets the PHAT typically results in significant performance degradation; however, the PHAT-@b can achieve performance improvements for both narrowband and broadband signals. Finally, the performance for real speech signal samples is examined and shown to exhibit properties similar to both the simulated broad and narrowband cases, suggesting the use of @b values between 0.5 and 0.7 for array applications with general signals.

Audio signal delay estimation using partial whitening

This work examines time and frequency domain implementations for estimating delays between acoustic signals arriving at spatially distributed microphones. A parametric variant of the phase-only transform (PHAT) is introduced for partially whitening the signal before estimating the delay. The PHAT variant is referred to as the PHAT-beta and is shown to be advantageous when processing signals corrupted by both independent noise and reverberation effects. Simulations show superior performance for the time-domain implementation under conditions of independent noise for time-limited broadband signals, achieving low estimation errors at signal-to-noise ratios 8 to 13 dB lower than that required for a frequency-domain implementation. Extensive Monte Carlo simulations are also performed for the time-domain delay estimator using the PHAT-beta on speech signals corrupted by reverberation and independent noise. Performance metrics include percent anomalous detections as well as the root mean square estimation error. Results show that partial whitening leads to significant improvements over zero or total whitening (as in the case of the standard PHAT). Simulations indicate that robust performance can be achieved for beta values near 0.4 when both reverberations and independent noises are present.

A fast integral equation solution technique for printed circuits in layered media

A fast solution for the full electromagnetic simulation of printed circuits in layered media is presented. The method is based on a Galerkin solution of the mixed potential integral equation (MPIE) with high-order vector basis functions and arbitrary discretization. A fast iterative solution is performed via the quadrature sampled precorrected FFT (QSPCFFT) (Gedney, S.D. et al., 2003). The QSPCFFT is in the same class of techniques as the adaptive integral method (AIM) (Bleszynski, E. et al., 1996), the precorrected FFT (Phillips, J.R. and White, J.K., 1997), or the sparse matrix/canonical grid algorithm (Li, S.Q. et al., 2001). Near interactions are computed using the traditional integral equation formulation, and far interactions are computed via the FFT. The proposed technique has distinct advantages over previous methods in that it does not explicitly require the computation of moments or an expansion of the Greens function. It also has the advantage over fast solution methods such as the FMM (fast multipole method) (Ling, F. et al., 1999) in that the size of the near field block is a function of discretization rather than electrical dimensions. It is shown that the QSPCFFT method is highly accurate and efficient. The solution scales with a computational complexity of O(N log N) and memory as O(N).

Scientific Programming in Field Theory, Part 2

For pt.1 see ibid., vol.3, no.3, p.66-74 (2001). As the first part of this two part article explained, TETlib, the theoretical electrical engineering toolkit, is an object oriented C++ library that we designed for basic purposes in field theory. The last issue described its software engineering aspects. The second part addresses its field theory aspects, introducing its main classes to give an overview of the functionality involved

A cluster-based computing infrastructure for wide-area multi-modal surveillance networks

Wide-area surveillance networks of diverse sensors present unique challenges for processing the massive amount of data generated. This paper presents a general, flexible computing infrastructure based on free and open-source software components to tackle the problem of quasi-realtime processing of time-sensitive surveillance data. The integration of NISTs SmartFlow system enables the transport of data to a Rocks-based cluster of computers for more CPU-intensive computations. Services, metadata, and events are made available to the network via a central server called the Blackboard. This allows application programmers to access the raw sensor data within a user-settable time-to-live. A general processing framework based on Trolltechs Qt4 signal and slot mechanism hides the complexity of multi-threaded programming and enables users to fully exploit the potential of the SMP machines in the cluster. Case studies are presented for an independent agent detector that uses multiple cameras as well as scalability results and for the processing of massively multi-channel audio data from a microphone array to achieve e.g. sound-source location in near-realtime.

Time-frequency masking for speaker of interest extraction in an immersive environment

Distributed microphone systems can be used to enhance intelligibility for a speaker of interest (SOI) in a noisy environment of multiple speech sources (cocktail party scenario). For finite microphone distributions, however, interfering speech sources leak into the beamformed signal and degrade intelligibility. This article introduces an auditory inspired post-processing algorithm for beamformed signals using spectro-temporal cues to enhance SOI intelligibility. Spatial power ratios obtained through beamforming on multiple locations are used to identify and mask out time-frequency regions dominated by the interfering speech. Performance results based on planar microphone array simulations show consistent increases in the Speech Intelligibility Index (SII) over the beamformed signal for various configurations of speakers using 2 to 16 microphones. In cases of critically low SII (<; 0.25), the application of interference masking achieves critical enhancements in SII, increasing it beyond .3 for the case of 2 microphones to above .5 for the 16 microphone case. Experimental recording were also performed and examples presented. The experimental recordings show similar improvements consistent with the simulation.

A monte-carlo study on the convergence of Multipole-Matched audio Rendering

Multipole-Matched Rendering (MMR) is a novel method to render spatial audio in a sweet spot around a listeners head. It is based on matching multipole expansions of a virtual and several actual sound sources using the Singular-Value Decomposition of a matrix. A critical parameter of the algorithm is the number of radial modes used in the multipole expansion. It is computationally expensive to determine convergence based on actual sound source renderings. This paper shows that in order to detect convergence of the algorithm when increasing the number of radial modes, looking at the difference of the matrix condition numbers of two subsequent computations is sufficient. A Monte- Carlo experiment has been designed to verify this. Results show that using the condition number is a valid and efficient way to determine convergence and additionally give a heuristic estimate on the number of radial modes required for convergence in a typical immersive environment setup.

A Quadrature-Sampled Pre-Corrected FFT for the Analysis of Microwave Circuits in Layered Media

Abstract This paper presents a fast iterative solution based on the quadrature sampled pre-corrected fast fourier transform (QSPCFFT) algorithm used to accelerate a method of moment discretization of the mixed potential integral equation (MPIE) for the analysis of circuits printed in a layered medium. The QSPCFFT algorithm is applied to accelerate the computation of all far interactions, and has a complexity that scales as O (N log N) and memory that scales as O(N). The algorithm is efficiently applied to broad-band analysis of printed circuits since the far-field is not constrained by the wavelength, but is determined rather by the local discretization.

Real-time server overloaded monitoring algorithm using back propagation artificial neural network

In recent years, downtime and information loss problems of server computers have become more critical. Even if a server has antivirus and CPU overload checking programs, it may occasionally be broken or slowed down. Hypothetically, there are possible indications of the problems under specific external situations such as high temperature, low fan speed, and extreme main board vibration. In order to recognize the correlation between external conditions and the overloaded problems, a monitoring computer collects data from external sensors. By using an accelerometer, an anemometer, and a temperature sensor, the monitoring algorithm is able to predict the target computers status. A web application has been developed to help server managers to remotely know how server computers are operating. This paper proposes a monitoring algorithm to diagnose server overload for a target computer, based on the Fast Fourier Transform, multivariate linear regression, and learning algorithms. As a result, this paper suggests that a monitoring algorithm can be implemented with an artificial neural network that warns of possible malfunction cases.

A comparative study of perceptional quality between wavefield synthesis and Multipole-Matched Rendering for spatial audio

This paper introduces a new algorithm to render virtual sound sources with spatial properties in immersive environments. The algorithm, referred to as multipole-matched rendering, uses the method-of-moments and singular-value decomposition to optimally match a spherical-multipole expansion of the virtual source to the field resulting from a spatially distributed speaker set. The flexibility of this method over other approaches, such as wavefield synthesis, allows for complex speaker geometries, and requires a smaller number of speakers to achieve a similar spatial rendering performance for listeners in immersive environments. The trade-off for the enhanced performance is a smaller area of faithful reproduction. This limited area, however, can be focused around listener locations for a sweet-spot solution. Experimental results are presented from perceptual tests comparing multipole-matched rendering to both wavefield synthesis and stereo rendering using a linear speaker array. The experiments included 13 subjects and demonstrated that the perceived direction of a virtual sound source for the new method is comparable to that of wavefield synthesis (no significant difference). The results demonstrate the potential of multipole-matched rendering as an efficient technique for rendering virtual sound sources in immersive environments.