Speech Enhancement, Gain, and Noise Spectrum Adaptation Using Approximate Bayesian Estimation

Abstract

Abstract-This paper presents a new approximate Bayesian estimator for enhancing a noisy speech signal. The speech model is assumed to be a Gaussian mixture model (GMM) in the log-spectral domain. This is in contrast to most current models in frequency domain. Exact signal estimation is a computationally intractable problem. We derive three approximations to enhance the efficiency of signal estimation. The Gaussian approximation transforms the log-spectral domain GMM into the frequency domain using minimal Kullback-Leiber (KL)-divergency criterion. The frequency domain Laplace method computes the maximum a posteriori (MAP) estimator for the spectral amplitude. Correspondingly, the log-spectral domain Laplace method computes the MAP estimator for the log-spectral amplitude. Further, the gain and noise spectrum adaptation are implemented using the expectation-maximization (EM) algorithm within the GMM under Gaussian approximation. The proposed algorithms are evaluated by applying them to enhance the speeches corrupted by the speech-shaped noise (SSN). The experimental results demonstrate that the proposed algorithms offer improved signal-to-noise ratio, lower word recognition error rate, and less spectral distortion.Index Terms-Approximate Bayesian estimation, Gaussian mixture model (GMM), speech enhancement.

Speech enhancement via low-rank matrix decomposition and image based masking

“…We compare our method with four traditional denoising approaches. "Spectral Subtraction" stands for speech enhancement by spectral subtraction [1], "LSA" means using a minimum mean-square error log-spectral amplitude estimator [9], "ML" is a method based on maximum likelihood estimator [10] and "MAP" denotes for an approach based on maximum a posteriori estimation [11]. To be worthy of attention, we degrade these methods' performances by letting them to estimate the noise by using only the first fifteen frames.…”

Enhancement of performance parameters of speech signal using model order reduction approach

“…This performance parameter is defined as below in terms of spectral coefficients of the input signal and estimated signal (Hao et al 2009). …”

Speech Denoising via Low-Rank and Sparse Matrix Decomposition

“…Examples of such methods include independent component analysis [2], non-negative matrix factorization (NMF) [3]- [5], and K-SVD [6]. However, these always require particular features or prior training for supervised separation.…”