Results 1 -
3 of
3
Microsoft Word - Logmmse_L1_EM.docx
"... Abstract-Voiced speeches have a quasi-periodic nature that allows them to be compactly represented in the cepstral domain. It is a distinctive feature compared with noises. Recently, the temporal cepstrum smoothing (TCS) algorithm was proposed and was shown to be effective for speech enhancement in ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract-Voiced speeches have a quasi-periodic nature that allows them to be compactly represented in the cepstral domain. It is a distinctive feature compared with noises. Recently, the temporal cepstrum smoothing (TCS) algorithm was proposed and was shown to be effective for speech enhancement in nonstationary noise environments. However, the missing of an automatic parameter updating mechanism limits its adaptability to noisy speeches with abrupt changes in SNR across time frames or frequency components. In this paper, an improved speech enhancement algorithm based on a novel expectationmaximization (EM) framework is proposed. The new algorithm starts with the traditional TCS method which gives the initial guess of the periodogram of the clean speech. It is then applied to an L 1 norm regularizer in the M-step of the EM framework to estimate the true power spectrum of the original speech. It in turn enables the estimation of the a-priori SNR and is used in the E-step, which is indeed a logmmse gain function, to refine the estimation of the clean speech periodogram. The M-step and Estep iterate alternately until converged. A notable improvement of the proposed algorithm over the traditional TCS method is its adaptability to the changes (even abrupt changes) in SNR of the noisy speech. Performance of the proposed algorithm is evaluated using standard measures based on a large set of speech and noise signals. Evaluation results show that a significant improvement is achieved compared to conventional approaches especially in nonstationary noise environment where most conventional algorithms fail to perform.
Best Signal Selection with Automatic Delay Compensation in VoIP Environment
"... Selecția semnalelor și compensarea automată a întârzierii în mediul VoIP ..."
On the Use of Adaptive Fuzzy Wavelet Filter in the Speech Enhancement
"... Abstract—This paper proposes an adaptive fuzzy wavelet filter that is based on a fuzzy inference system for enhancing speech signals and improving the accuracy of speech recognition. In the last two decades, the basic wavelet thresholding algorithm has been extensively used for noise filtering. In t ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract—This paper proposes an adaptive fuzzy wavelet filter that is based on a fuzzy inference system for enhancing speech signals and improving the accuracy of speech recognition. In the last two decades, the basic wavelet thresholding algorithm has been extensively used for noise filtering. In the proposed method, adaptive wavelet thresholds are generated and controlled according to the fuzzy rules about the presence of speech in contaminated signals. In this adaptive fuzzy wavelet filter, the relationships between speech and noise are summarized into seven fuzzy rules using four linguistic variables, which are used to determine the state of a signal. A hybrid filter is proposed here, which combines an adaptive fuzzy wavelet filter and the spectral subtraction method to filter contaminated signals. An amplified voice activity detector in the proposed hybrid filter is designed to improve performance when the signal-to-noise ratio (SNR) is lower than 5 dB. The filtering that is performed using the adaptive fuzzy wavelet filter and the spectral subtraction method is controlled by support vector machines. Experimental results demonstrate that the proposed system effectively increases the SNR and the speech recognition rate. Index Terms—speech enhancement; wavelet thresholding; fuzzy; voice activity detection, spectral subtraction, support vector machines; I.
