|
Objektivní porovnání účinnosti jednokanálových metod potlačení šumu
Koula, Ivan ; Heralt, J.
The contribution tries to evaluate objectively the efficiency of most widely used one-channel noise cancellation methods for a speech record. In this contribution three basic one-channel methods are compared (spectral thresholding, spectral subtraction and RASTA algorithm). The efficiency of designated methods is evaluated by percentage hit rate of isolated words recognition. This test enables an objective evaluation of methods with regard to the intelligibility.
|
|
Konverse hlasu
Vích, Robert ; Vondra, Martin
Different techniques for voice conversion based on nonlinear spectral envelope warping are presented. They enable transformation of an utterance of a source speaker into the utterance of a target speaker. The voice conversion algorithm is based on spectral speech analysis, frequency transformation, spectrum envelope warping and high quality cepstral speech synthesis. The described voice conversion approach also considers the change of speech prosody.
|
|
Zvýraznování řeči pomocí cepstrálního vokodéru
Vondra, Martin
In this paper the cepstral vocoder is utilized in speech enhancement. In the analysis step first an algorithm for suppression of noise is used and then the estimation of the cepstral model follows, which leads to more reliable speech parameters. The speech is then synthesized by the cepstral vocoder, which works in the source-filter mode. The output speech is compared with speech enhanced by standard algorithm like spectral subtraction and Log Spectral Amplitude (LSA) estimator.
|
|
Slepá separace signálů pomocí fuzzy maskování
Čermák, Jan
Paper deals with separation of source signals from their mixtures without any information about mixing environment. This problem is known as blind source separation. We focus on improving a multichannel method called time-frequency binary masking. The proposed approach is based on using a fuzzy mask instead of a binary mask. We show that the separation performance increases by employing fuzzy mask. Moreover our method reduces musical noise introduced by binary masking.
|
| |
|
Algoritmy potlačení šumu v řeči zkreslené telekomunikační sítí
Koula, Ivan ; Esposito, A.
This paper aims to provide an evaluation of the effectiveness of three different speech enhancement algorithms. The evaluation of their efficiency was based on the hit rate recognition by computer speech recognizer. These speech enhancement algorithms are based on the method of spectral subtraction and differ by noise spectrum estimation system. The first algorithm estimates noise spectrum on the basis of its statistical characters. Next two algorithms estimate noise spectrum by nonlinear adaptive models.
|
| |
|
Modifikace metody mapování spektrogramu
Koula, Ivan
Algorithms enhancing the useful audio signal at the expense of background interference noise have become a topical problem. Most important of all is to improve methods working in real time. The spectrogram mapping method is a comparatively simple and for some kinds of additive noise very effective method.
|
|
Real time noise reduction
Koula, Ivan
Algorithms enhancing the useful audio signal at the expense of background interference noise have become a topical problem. Most important of all is to improve methods working in real time. These methods come in useful especially in area of mobile communications.
|
|
Pseudokepstrogram a jeho použití při analýze řeči
Vích, Robert ; Vondra, Martin
Generalized cepstral speech analysis is briefly summarized and applied to the construction of cepstrograms. By means of cepstrograms and pseudo cepstrograms of clean and noise corrupted speech it is shown that pseudo cepstrograms are more robust to background noise. The mel-scale representation of the speech spectrum is further used as a pre-processing step in pseudo cepstrum visualization. This combination of nonlinear approaches proved to be a very effective tool in speech processing.
|