In this paper, traditional noise reduction algorithms such as spectral subtraction, Wiener, MMSE and logMMSE filtering algorithms, and two less known Wiener-TSNR and Wiener-HRNR filtering algorithms had been compared with the use of a set of quality measures. It is found that excessive noise reduction leads to insignificant degradation of the speech signals quality, but significantly reduces the accuracy of the automatic speech recognition (ASR). It is shown the existence of the speech quality measures which satisfactorily are matching with the accuracy of automatic speech recognition. This result is useful for practice because of speech recognition accuracy can be predicted by means of speech quality measures. In addition, it is found that there is no single algorithm among the considered noise reduction algorithms, which is the best in terms of maximum recognition accuracy for a wide range of input signal-to-noise ratio from minus 10 dB to plus 30 dB.

  Повний текст PDF - 508.075 Kb    Зміст випуску     Цитування публікації

  Якщо, ви не знайшли інформацію про автора(ів) публікації, маєте бажання виправити або відобразити більш докладну інформацію про науковців України запрошуємо заповнити "Анкету науковця"