A maximum-a-posteriori approach for enhancing speech signals which have been degraded by statistically independent additive noise is proposed. The approach is based on statistical modeling of the clean speech signal a...
详细信息
A maximum-a-posteriori approach for enhancing speech signals which have been degraded by statistically independent additive noise is proposed. The approach is based on statistical modeling of the clean speech signal and the noise process using long training sequences from the two processes. Hidden Markov models (HMMs) with mixtures of Gaussian autoregressive (AR) output probability distributions (PDs) are used to model the clean speech signal. The model for the noise process depends on its nature. The parameter set of the HMM model is estimated using the Baum or the EM (estimation-maximization) algorithm. The noisy speech is enhanced by reestimating the clean speech waveform using the EM algorithm. Efficient approximations of the training and enhancement procedures are examined. This results in the segmental k-means approach for hidden Markov modeling, in which the state sequence and the parameter set of the model are alternately estimated. Similarly, the enhancement is done by alternate estimation of the state and observation sequences. An approximate improvement of 4.0-6.0 dB in signal-to-noise ratio (SNR) is achieved at 10-dB input SNR.
A digital electronic architecture for parallel processing of the expectation maximization (EM) algorithm for positron-emission-tomography (PET) image reconstruction is proposed. Rapid (0.2-s) EM iterations on high-res...
详细信息
A digital electronic architecture for parallel processing of the expectation maximization (EM) algorithm for positron-emission-tomography (PET) image reconstruction is proposed. Rapid (0.2-s) EM iterations on high-resolution (256*256) images are supported. Arrays of two VLSI chips perform forward and back projection calculations. The architecture is described, including data flow and partitioning relevant to EM and parallel processing. EM images are shown that are produced with software simulating the proposed hardware reconstruction algorithm. Projected cost of the system is estimated to be small in comparison to the cost of current PET scanners.
The paper points out that a certain iterative method of direct deconvolution is a particular manifestation of the so-called expectation-maximisation (EM) algorithm. Recognition of this leads to the establishment of ge...
详细信息
The paper points out that a certain iterative method of direct deconvolution is a particular manifestation of the so-called expectation-maximisation (EM) algorithm. Recognition of this leads to the establishment of general convergence results by reference to existing work. Zusammenfassung Dieser Artikel zeigt dass eine iterative direkte Unfaltungsmethode, die auf dem Bayes Theorem basiert, ein Spezialfall des E.M. algorithmus ist (expectation-Maximization). Diese Bemerkung erlaubt es, die Konvergenz mit Hilfe von Referenzen zu frueren Arbeiten zu zeigen.
暂无评论