VSE knjižnice (vzajemna bibliografsko-kataložna baza podatkov COBIB.SI)
  • Missing feature detection in various noisy environments
    Vlaj, Damjan
    The problem of the automatic speech recognition is that whenever noise is present, the accuracy level of the recognition proces is significantly reduced. Owing to the fact that in the world of real ... applications nois almost always exists in the background several techniques have been developed to reduce the influence of noise on the speech recognition process. One of the possibilities to reduce the influence of noise on speech signal is using pre-processing speech enchancement techniques. In our research we used the Missing Feature Theory. Missing feature problem occurs when spech segments are filtered, interrupted or masked by noise. Our research procedure was as follows. In the first step we detected noise by the amount of energy present in the speech signal. Based on our energy calculation we eliminated the noisy part of the signal from our further research. We continued by calculating the MFCC_E_D_A feature vector for that part of signal, which contained speech. The studio database TIDIGITS, to which we added different noise types, used for our work. Our research was concluded by another test on the connected digits, for which we used the Slovenian telephone database SpeechDat II, which incorporated noise from natural environment.
    Vir: Advances in speech technology : seventh International Workshop, Maribor, Slovenia, July 4-5, 2000 : proceedings (Str. 147-155)
    Vrsta gradiva - prispevek na konferenci
    Leto - 2001
    Jezik - angleški
    COBISS.SI-ID - 6308118

Vnos na polico