Narrow Results

The a priori signal-to-noise ratio (SNR) plays an essential role in many speech enhancement systems. Most of the existing approaches to estimate the a priori SNR only exploit the amplitude spectra while making the phase neglected. Considering the fact that incorporating phase information into a speech processing system can significantly improve the speech quality, this paper proposes a phase-sensitive decision-directed (DD) approach for the a priori SNR estimate. By representing the short-time discrete Fourier transform (STFT) signal spectra geometrically in a complex plane, the proposed approach estimates the a priori SNR using both the magnitude and phase information while making no assumptions about the phase difference between clean speech and noise spectra. Objective evaluations in terms of the spectrograms, segmental SNR, log-spectral distance (LSD) and short-time objective intelligibility (STOI) measures are presented to demonstrate the superiority of the proposed approach compared to several competitive methods at different noise conditions and input SNR levels.

A new approach has been presented to characterise phase spectra for simulating realistic nonstationary characteristics in synthetic accelerograms. The phase characteristics of the recorded earthquake accelerograms have been studied for this purpose and it has been found that the phase curve/unwrapped phases exhibit a monotonic downward trend which allows the problem of phase characterisation to be cast as a constrained nonlinear programming problem. The phase spectrum is first characterised by matching mean and variance of the generated distribution of relative phases with those obtained from recorded motions. As a practical application, it is shown how phase spectra can be characterised for an ensemble of synthetic accelerograms so as to maximise the severity of sample realisations.

A family of simulation methods for nonstationary earthquake ground motions is proposed. It employs a univariate model of phase spectrum built up on a time argument associated with the concept of starting-time of phase evolution of frequency components. This phase model allows a feasible phase spectrum just using few variables of the starting-time in numerical implementation. In order to reduce the computational effort of the starting-time, a wave-group propagation formulation is also introduced. Two observed ground motions at the type-II site, i.e. Northridge and Chi-Chi waves, are investigated for illustrative purposes. Inspired from the proposed method, a numerical technique for the spatial variation of ground motions is also developed, and the investigation of coherency function between the example ground motions observed at different stations is carried out. Numerical results prove the validity and applicability of the simulation scheme. This methodology provides a new perspective towards the representation of nonstationary stochastic ground motions.