Pii: S0165-1684(01)00128-1
نویسندگان
چکیده
In this paper, we present an optimally-modi#ed log-spectral amplitude (OM-LSA) speech estimator and a minima controlled recursive averaging (MCRA) noise estimation approach for robust speech enhancement. The spectral gain function, which minimizes the mean-square error of the log-spectra, is obtained as a weighted geometric mean of the hypothetical gains associated with the speech presence uncertainty. The noise estimate is given by averaging past spectral power values, using a smoothing parameter that is adjusted by the speech presence probability in subbands. We introduce two distinct speech presence probability functions, one for estimating the speech and one for controlling the adaptation of the noise spectrum. The former is based on the time–frequency distribution of the a priori signal-to-noise ratio. The latter is determined by the ratio between the local energy of the noisy signal and its minimum within a speci6ed time window. Objective and subjective evaluation under various environmental conditions con6rm the superiority of the OM-LSA and MCRA estimators. Excellent noise suppression is achieved, while retaining weak speech components and avoiding the musical residual noise phenomena. ? 2001 Elsevier Science B.V. All rights reserved.
منابع مشابه
Multichannel filtering by gradient information
In this paper, multichannel image "ltering using local gradient information is studied and evaluated which is simpler and more appropriate than the traditional approaches that have been addressed by means of groupwise vector ordering information. Two adaptive weighted multichannel "lters based on local gradient information to noise removal in color images are introduced. The "rst proposed multi...
متن کاملDetection of point landmarks in multidimensional tensor data
This paper describes a unified approach to the detection of point landmarks-whose neighborhoods convey discriminant information-including multidimensional scalar, vector, and higher-order tensor data. The method is based on the interpretation of generalized correlation matrices derived from the gradient of tensor functions, a probabilistic interpretation of point landmarks, and the application ...
متن کاملImproved FIR filters for discrimination of damped sinusoids
It is shown that the FIR filter based discrimination of damped sinusoids can be improved by removing the linear phase constraint. The resulting filter is not only economical and better suited to processing of shorter data records, but also offers superior numerical conditioning and therefore reduced error and sensitivity to perturbations. Zusammenfassung Gezeigt wird, da13 die auf FIR-Filtern b...
متن کاملPii: S0165-1684(01)00098-6
A popular technique for time delay estimation is to use an FIR #lter to model the time di-erence and the #lter weights are interpolated with a sinc function to obtain the delay estimate. However, the sinc interpolator requires a su/ciently long #lter length for accurate delay estimation. In this paper, we propose to process the #lter weights via a least-squares-based method in order to acquire ...
متن کامل