Spectral density estimation

About: Spectral density estimation is a research topic. Over the lifetime, 5391 publications have been published within this topic receiving 123105 citations.

Adaptive two-band spectral subtraction with multi-window spectral estimation

Abstract: An improved spectral subtraction algorithm for enhancing speech corrupted by additive wideband noise is described. The artifactual noise introduced by spectral subtraction that is perceived as musical noise is 7 dB less than that introduced by the classical spectral subtraction algorithm of Berouti et al. (1979). Speech is decomposed into voiced and unvoiced sections. Since voiced speech is primarily stochastic at high frequencies, the voiced speech is high-pass filtered to extract its stochastic component. The cut-off frequency is estimated adaptively. Multi-window spectral estimation is used to estimate the spectrum of stochastically voiced and unvoiced speech, thereby reducing the spectral variance. A low-pass filter is used to extract the deterministic component of voiced speech. Its spectrum is estimated with a single window. Spectral subtraction is performed with the classical algorithm using the estimated spectra. Informal listening tests confirm that the new algorithm creates significantly less musical noise than the classical algorithm.

A Survey of Spectral Analysis

Abstract: A wide variety of applications of spectral analysis have been reported in the literature since spectral estimation methods were introduced by M. S. Bartlett and J. W. Tukey about 15 years ago. In no sense, however, can it be said that spectral analysis is widely used or even understood by statisticians and many of the applications of the technique have in fact been made by physicists and engineers. It is suggested that there are two reasons for this: (1) The genuine difficulties which statisticians (as opposed to physicists and engineers) face in thinking in terms of frequency concepts. (2) The highly mathematical nature of papers written on spectral analysis. This undue emphasis on mathematical work has led many statisticians to believe that spectral analysis is very difficult to apply. This is not the case-in fact the important ideas in spectral analysis are no more difficult than those involved in estimating a probability density function by means of a histogram. In this paper we shall try to present, using the minimum of mathematics, all those ideas in spectral analysis which are necessary in order to be able to apply the technique. In the last resort the only way to understand spectral analysis is to use it and so where possible the main ideas have been illustrated by means of examples. Two forms of spectral analysis are discussed in detail, namely, (1) spectral analysis of a single time-series to be referred to as auto-spectra; (2) spectral analysis of pairs of time-series to be referred to as crossspectra. However other forms of spectral analysis are mentioned briefly in section 7. Cross-spectral analysis is useful in two contexts:

Method and apparatus for enhancing frequency domain analysis

Abstract: A signal processing technique allows accurate interpolation between points of a sampled frequency domain function. A time-based sampling process samples a time waveform of duration T having a sub time interval T'. A signal processor applies a discrete Fourier transform over a time period T to transform the sampled data from the time domain to the frequency domain. The sampled frequency domain data is convolved with one or more convolution kernels to yield a continuous line shape. The result of this convolution permits the spectral composition at arbitrary frequencies to be determined. The disclosed frequency domain interpolation process is characterized by preservation of data in the T' interval of the time domain with an arbitrary but specified degree of accuracy.

A Novel Time–Frequency Transform Based Spectral Energy Function for Fault Detection During Power Swing

Abstract: Fault during a power swing is a challenging task for the distance relay functioning This article presents a spectral energy function for fault detection during a power swing using a novel time frequency transform known as the S-transform, a variable windowed short-time Fourier transform, which combines the elements of short-time Fourier transform and wavelet transform Initially, the current signal is preprocessed using S-transform to generate the S-matrix and corresponding S-contours (time–frequency contours) The spectral energy content of the S-counters is used to register symmetrical and unsymmetrical faults during a power swing and, based on a set threshold on the spectral energy, the relay blocks during a power swing and issue of the tripping signal during fault The proposed technique is tested for different fault conditions during a power swing with possible variations in operating parameters, including the ability to identify the faults with a response time of 125 cycles from the fault